November 5, 2025 • 47 mins
The source provides an extensive overview of how Artificial Intelligence (AI) and Robotic Process Automation (RPA) are fundamentally transforming the hospitality sector, specifically targeting operational efficiency and profitability. It begins by highlighting a severe margin squeeze driven by rising costs for labor, energy, and food, contrasting this with properties achieving significant EBITDA growth through AI adoption. The episode then details specific applications, such as using AI to reduce the night audit process from hours to seconds and employing generative models like GPT-4o and Claude 3.5 Sonnet to ensure dynamic regulatory compliance and avoid massive fines. Furthermore, the source explains that AI facilitates Revenue Management 3.0 by optimizing unknown demand using real-time external data and addresses labor shortages by augmenting roles, resulting in superior performance in areas like upselling and personalized guest experiences. Ultimately, the analysis concludes that AI-native hotels are securing higher valuations and achieving substantial ROI through strategic technology deployment.
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
Speaker 1 (00:00):
Welcome back to the deep dive. Today, we are focusing
on well, one of the biggest sectors out there, hospitality,
and it's facing some fundamental challenges right now. Yeah. Absolutely,
We're not here today to talk about you know, fancy
lobbies or loyalty points. We are really digging into the
financial nuts and bolts, specifically what our sources are calling
(00:20):
the margin squeeze. No one saw coming.
Speaker 2 (00:23):
And that squeeze, I mean, as of late twenty twenty five,
it's become a real structural threat. It's not just a blip.
When the Bureau of Labor Statistics data dropped in September,
it painted a pretty clear picture and frankly, it was painful.
Operational costs are just surging way faster than revenue.
Speaker 1 (00:38):
How bad are we talking, Well, look.
Speaker 2 (00:40):
At the numbers. Energy costs are up four point eight
percent year over year, food and beverage spiking six point
two percent, and wages you know for the essential staff
everyone needs, still climbing at four point nine percent.
Speaker 1 (00:51):
Okay, so costs are really heading upwards fast. What about
the other side? Is revenue keeping pace at all our
hotels managing to charge more?
Speaker 2 (00:59):
That's the problem. It's not accelerating revpair growth. You know,
revenue per available room, the key metric. It's basically crawled
to a stop across the America's just one point one percent.
Speaker 1 (01:09):
Wow, one point one percent. That's sluggish.
Speaker 2 (01:11):
It's the weakest recovery we've seen since well since the
initial bounce back after the pandemic first hit. So think
about it. If you're running a standard, say two hundred
and fifty room hotel, that cost surge. It means you're
suddenly eating an extra one hundred and eighty seven thousand
dollars an annual operating collage, and you can't just jack
up your rates to cover that, not without losing guests
(01:33):
to the competition. The math, as the sources really hammer home,
it's just it's utterly merciless.
Speaker 1 (01:39):
Okay, so that sounds pretty bleak across the board, universal
distress almost, But this deep dive, it isn't just about
the pain, right, It's about the ones who are figuring.
Speaker 2 (01:48):
It out exactly.
Speaker 1 (01:49):
Our sources highlight this huge contrast. You've got properties really
struggling and then others others are suddenly pulling away financially massively.
Speaker 2 (01:56):
That's the core story here, that distinction because right in
the middle of that tough market, dealing with the same inflation,
the same flat revenue. There's this uh three hundred and
twelve room autographic Collection hotel and they recently posted a wow,
a staggering seven point four percent lift in their ebit
dust seven.
Speaker 1 (02:16):
Point four percent in this climate.
Speaker 2 (02:18):
How not by cutting corners, not by slashing staff or
cheapening the guest experience. Note they did it through very specific,
very integrated deployments of AI artificial intelligence and RPA robotic
process automation. And the result a twenty two point swing
in labor productivity.
Speaker 1 (02:33):
Well, hang on a twenty two point swing. That's not
just tweaking things. That's structural, that's a different operating model exactly.
Speaker 2 (02:37):
It's fundamental change, not just you know, incremental improvement.
Speaker 1 (02:40):
So the mission for you listening today, whether you're an
owner and operator, maybe an investor looking at the space,
is to really get the blueprint here. We need to
unpack how these specific winning hotels are using tools like
GPT four to zero, GROG four, Claude three point five sonnet.
We're moving beyond just automating simple tasks. This is about
building systems that can predict issues they can almost self
(03:04):
heal operationally, turning that what was it, one hundred and
eighty seven thousand dollars drag, Yeah, into a genuine competitive advantage.
Speaker 2 (03:12):
And you have to understand, this isn't just theory anymore.
It's not pilots. When a few properties start showing these
kinds of returns, it immediately resets the bar. It redefines
the baseline expectation for profitability for everyone in the sector.
Speaker 1 (03:25):
All right, so the pressure's on too, Adapter get left
behind pretty much. Yeah, Okay, let's jump straight into the
operational weeds then, because the sources, they really showed that
the first big wave of savings, the really massive ones,
come from tackling that friction in the back office, you know,
the stuff where accuracy used to mean hours and hours
of human effort.
Speaker 2 (03:43):
Yeah, and mistakes, oh absolutely. And if you talk to
any GM, any hotel manager and ask them what the
single most dreaded, high stress but ultimately low value pass is,
they'll all say the same thing. The night audit.
Speaker 1 (03:56):
Ah, Yes, the infamous night audit, the traditional like two
point four or five am fire drill. That's where all
the gremlins crawl out right. Folio mismatches, reservations that just
seem to float in the limbo, those city ledger accounts
that never quite balance.
Speaker 2 (04:09):
Yep, all of it. It's manual. It's incredibly stressful for
the person doing it, and it's so easy to make
mistakes under pressure.
Speaker 1 (04:16):
Our sources actually track this a mid scale chain, I
think it was in the Midwest. They clocked an incredible
three hundred and twelve minutes of manual touches per property
per property during that audit, every single night.
Speaker 2 (04:28):
Think about that. That's over five hours of labor just
poured into this reconciliation black hole, night after night.
Speaker 1 (04:36):
So that huge time sink is exactly what this new
AI stack is designed to just obliterate precisely.
Speaker 2 (04:43):
We're looking at a combination system. Here. You've got the
property management system or PMS, let's say, use, then the
digital guest platform like Canary Technologies, your accounting system quick
books Online is common, and then the brain connecting them
all a large language model or LLM like GPT four
h Break that.
Speaker 1 (05:00):
Down for me for listeners who aren't you know, database wizards,
What does it actually mean when GPT four to oh
writes and executes complex SQL reconciliation scripts. Is the AI
literally coding on the fly.
Speaker 2 (05:13):
That's exactly the power shift. It's not just finding problems,
it's fixing them. So first canaries AI engine pulls in
all the raw transaction data. We're talking maybe fourteen thousand
individual transactions every night.
Speaker 1 (05:26):
Fourteen thousand. A human auditor would have to manually.
Speaker 2 (05:29):
Check all that, or at least sample heavily and hope
for the best. Now GPT four to oh, it's been
trained on millions of past audit logs, all the accounting rules,
all the reconciliation procedures. So when it spots something weird
like an orphaned payment, it doesn't just flag it. It
writes the specific SQL query needed to go into quick books,
compare the relevant fields, find the match, and then write
the correcting entry.
Speaker 1 (05:50):
And it does this how quickly.
Speaker 2 (05:51):
It executes that entire fine diagnosed right code, execute fixed
process in just zero point eight seconds.
Speaker 1 (05:57):
Wait, let me get this straight. Three hundred and twelve minutes.
It's five hours of manual, error prone, high stress work
is now done in less than one second. That sounds
almost unbelievable.
Speaker 2 (06:07):
Too clean, It's the new reality. This is the metric shift.
We're talking about. But you know, you need checks and balances,
so the system doesn't actually achieve zero human time.
Speaker 1 (06:15):
Okay, right, yeah, there has to be some oversight.
Speaker 2 (06:17):
Yeah. If the AI finds an anomaly it can't resolve
with high confidence or something really unusual, it automatically routes
it to a human auditor via slack. But and this
is key, that human only has a very tight, like
twelve second window to veto the AI's proposed fix.
Speaker 1 (06:33):
Twelve seconds that's not much time.
Speaker 2 (06:35):
It forces focus. If the human looks at the proposed
fix and it seems correct, they do nothing. The system
then automatically executes the correction. If they get veto, it
gets flagged for deeper manual review later. So the total
process time it drops from three hundred and twelve minutes
down to about three seconds three.
Speaker 1 (06:52):
Seconds okay, and the financial impact immediate.
Speaker 2 (06:54):
You're looking at annual labor savings of around forty one
two hundred dollars per property and the accuracy practically perfect
ninety nine point nine nine three percent.
Speaker 1 (07:04):
Wow. Okay, so forty one thousand dollars save, huge accuracy jump.
But what happens to the actual night auditor you mentioned
the human veto. But does this mean fewer jobs? The
sources must touch on this right, augmentation versus replacement.
Speaker 2 (07:17):
They do, and it's a critical point we'll revisit later.
The human role fundamentally changes. Instead of spending five hours
chasing ghosts in the data, the auditor now maybe spends
five minutes reviewing the handful of really complex exceptions the
AI couldn't crack, and the rest of their shift they're
freed up for higher value work, maybe preparing reports for
the owners, analyzing trends, even helping out the front desk
(07:40):
during a late rush. It's an elevation of the role,
not just elimination.
Speaker 1 (07:43):
Okay, that makes sense, shifting skills. So moving from saving
labor to avoiding well potentially massive costs. The sources talk
a lot about dynamic compliance, turning compliance from this reactive
headache into a proactive shield.
Speaker 2 (07:57):
This is a really important differentiator. It shows a i'maturity
beyond just basic automation. Think about privacy regulations. In Q
two twenty twenty five, the California Consumer Privacy Act the
CCPA got expanded significantly. It now covers transient guest PII
personally identifiable information.
Speaker 1 (08:16):
Okay, some names, addresses, maybe credit card details, Yeah, that.
Speaker 2 (08:20):
Kind of thing exactly. And the rules around redacting that data,
securing it, managing consent they got much stricter before AI
Manually complying with this for every single guest. Our sources
estimated it cost about twenty eight dollars per reservation just
for the compliance piece.
Speaker 1 (08:36):
Twenty eight dollars per booking. If you're running a big hotel,
say at four hundredroom resort in California, that adds up
incredibly fast.
Speaker 2 (08:43):
That's paralyzing, totally unmanageable, so the resort tracked. In the sources,
they implemented what's called the compliance Fortress. It's a stack
using Orcato for the automation workflow, one Trust for the
governance and policy side, and crucially Claude three point five
sonnet the AI from Entthropic.
Speaker 1 (08:58):
Interesting why Claude specific for this? We've mentioned GBT four
and GROC already. What's different here.
Speaker 2 (09:04):
Claude, particularly the three point five sonnet model, is built
around what Anthropic calls constitutional AI. It's designed to be
very good at adhering to complex strict rule sets and
explaining its reasoning. It excels that classification tasks where accuracy
and auditibility are paramount perfect for compliance.
Speaker 1 (09:23):
So how does this fortress actually work? When does it
kick in?
Speaker 2 (09:26):
The trigger is guest check in the moment that happens
in the PMS, it fires off an encrypted webhook basically
a secure notification. Claude receives the guest data in just
forty milliseconds. That's faster than you can blink. It analyzes
and classifies up to forty two different data fields name, email,
birth date, maybe even loyalty number, payment tourken. It identifies
(09:48):
everything that falls.
Speaker 1 (09:48):
Under CCPA forty milliseconds. Wow.
Speaker 2 (09:51):
Then where CATO takes over automatically redacting or tokenizing the
sensitive PII based on Claude's classification and the rules set
in one trust, and every single action is logged for
a complete audit trail.
Speaker 1 (10:01):
The speed is one thing, but the cost difference.
Speaker 2 (10:03):
Must be huge phenomenal. The cost per reservation for CCPA
compliance drops from twenty eight dollars down to zero zero
zero zero zero four basically zero.
Speaker 1 (10:12):
Okay, that's efficiency. Yeah, but you mentioned avoiding costs too, right.
Speaker 2 (10:16):
This is the kicker. The real financial game changer isn't
just the efficiency saving. It's the avoided penalties. In just
the first ninety days after rolling this out, this system
actively prevented eighty four separate potential regulatory violation.
Speaker 1 (10:30):
Eighty four violations. What's the cost of a.
Speaker 2 (10:32):
Violation under the expanded CCPA, it can be up to
fourteen thousand dollars per violation, So eighty four times fourteen
thousand dollars.
Speaker 1 (10:41):
Let's carry the one over one point one million dollars.
Speaker 2 (10:44):
One point two million dollars. Yeah, in avoided fines in
just three months.
Speaker 1 (10:47):
Okay, that single statistic, one point two million dollars saved
in ninety days. It completely reframes compliance. It's not just
a cost center anymore. It's a massive profit protector exactly.
Speaker 2 (10:58):
It moves from being operational drag to a strategic financial asset.
Speaker 1 (11:02):
All right, let's pivot. We've talked cost cutting compliance shields.
Now let's talk about actually generating more revenue radically more.
Part two in our Sources digs into what they call
revenue management three point zero.
Speaker 2 (11:13):
Yeah, this is exciting stuff. It's about moving beyond the
traditional approach, which, well, traditional revenue management is mostly about
optimizing for known demand, looking at your booking pas historical trends,
compset pricing, trying to price correctly based on what you
think is going to happen.
Speaker 1 (11:29):
Okay, reacting to patterns exactly.
Speaker 2 (11:32):
Revenue management three point zero powered by predictive AI is
about optimizing for unknown or unforeseen demand. It's about pricing
the future before that future even fully materializes.
Speaker 1 (11:43):
Pricing the future. I like that. So what's the big
example here?
Speaker 2 (11:47):
The primary case study is fantastic. It's the seventy two
hour flash storm that hit Boston back in October twenty
twenty five.
Speaker 1 (11:53):
Okay, set the scene for us. This wasn't like a
predictable blizzard everyone knew was coming, No, not at all.
Speaker 2 (11:58):
It was a severe, quite unexpected, and it caused chaos
at Logan Airport, a sudden thirty one percent spike in
flight cancelations, stranded passengers everywhere.
Speaker 1 (12:06):
So normally every revenue manager in Boston would be scrambling
right pulling old storm data, trying to figure out how
to fill rooms suddenly empty from cancelations, while maybe capturing
some walkins precisely.
Speaker 2 (12:17):
But that traditional response is inherently reactive. It relies on
lagging indicators. What has happened the Liberty Hotel in Boston,
though they're running a different kind of system. Was their
stack dueto game changer for the revenue management core, Google
big Query for handling massive data sets, and crucially Grock
four as the AI analysis engine.
Speaker 1 (12:38):
Grock four again. Why Grock here?
Speaker 2 (12:40):
Grock's strength, according to the sources, is its ability to
ingest and synthesize vast amounts of real time, rapidly changing
public data, very very quickly. It excels at connecting disparate
data points.
Speaker 1 (12:51):
Okay, so what happened at the Liberty?
Speaker 2 (12:52):
So at six point one s of am Eastern time,
the storm is rapidly getting worse. Cancelations are mounting at
that moment. Grock four ingested one point eight million rows
of high frequency, constantly changing data.
Speaker 1 (13:04):
One point eight million rows in real time. What kind
of data are we talking about? Not just hotel booking,
surely no way beyond that.
Speaker 2 (13:11):
It was pulling live flight status feeds directly from the FAA,
hyperlocal ride share demand patterns, seeing where people were suddenly
requesting cars indicating immediate need for shelter. Local train schedules
were there running delayed, Plus incredibly granular aciweather microgrid.
Speaker 1 (13:27):
Data microgrid weather data, how detail.
Speaker 2 (13:30):
Showing which specific neighborhoods were losing power or facing immediate flooding.
It wasn't just storm in Boston, it was severe impact here.
Now the AI synthesized all of this to understand not
just that demand existed, but the nature of the demand,
acute immediate need for safe shelter in specific locations.
Speaker 1 (13:47):
Okay, that's an incredible level of real time situational awareness.
How fast could the system turn that insight into actual
pricing decisions?
Speaker 2 (13:55):
Two minutes?
Speaker 1 (13:55):
Two minutes? Yes, By six point one ten am. Just
two minutes after ingesting that data, the system had completely
rewritten every single best available rate that back car levels
for every room type for the next ninety six hours.
Speaker 2 (14:07):
Wow. Was it just a blanket price increase. No. Far
smarter than that. It was surgical recognizing the immediate desperate
need for basic shelter. It dropped the rates for standard
rooms on the shoulder nights tonight and tomorrow night by
twenty nine percent, aggressively capturing that distressed traveler market. Flooding
in from the.
Speaker 1 (14:26):
Airport get captured volume first, But at.
Speaker 2 (14:28):
The same time it recognized that guests who typically book
premium suites. They'd prioritize safety, comfort, and luxury, and price
would be less of an object in this scenario. So
for the upcoming weekend, premium suites ejected the rates up
by a massive forty.
Speaker 1 (14:43):
Forty one percent increase on suites while dropping standard rates.
That's that's playing the elasticity perfectly based on the specific situation,
true dynamic pricing.
Speaker 2 (14:51):
Exactly exploiting the perceived value of safety and immediate availability
for different segments.
Speaker 1 (14:56):
So what was the bottom line?
Speaker 2 (14:57):
Did it work amazingly well? Nine point zero am that morning,
just one hundred and sixty eight minutes after the AI
kicked into gear, the hotel was one hundred percent sold out,
completely full, and their average daily rate the ADR it
ended up being forty two points above the average for
their direct competitors in the str report for that period, forty.
Speaker 1 (15:16):
Two points above the comp set while being completely sold
out during a chaotic event. That translates to serious money.
Speaker 2 (15:21):
Incremental revenue directly attributable to this AI driven strategy. During
that seventy two hour window, four hundred and eighty seven
thousand dollars.
Speaker 1 (15:30):
Nearly half a million dollars pulled out of thin Air
essentially from what could have easily been a disaster with mass.
Speaker 2 (15:36):
Cancelations, exactly salvaged from a potential cancelation hole and turned
into a massive win.
Speaker 1 (15:41):
Okay, I have to ask the compute cost running Rock
for processing one point eight million data rows, rewriting all
those rates, What did that cost the hotel?
Speaker 2 (15:51):
This is the punchline verifiable cloud compute costs for that
entire operation, eleven dollars and forty.
Speaker 1 (15:57):
Cent eleven dollars and forty cents to make nearly half
a million.
Speaker 2 (16:00):
Yep, that cost to profit ratio. It just demonstrates the
monumental shift. It's not a human guessing anymore. It's an
algorithm seeing demand signals your competitors literally cannot see yet,
and acting instantly.
Speaker 1 (16:11):
Incredible. Okay, let's shift gears again from revenue generation to
arguably the biggest constraint facing the industry people staffing. The
ahla's twenty twenty five report, it confirms sixty eight percent
of operators cite staffing shortages as their number one headache.
Speaker 2 (16:28):
Yeah, it's a persistent, major challenge.
Speaker 1 (16:30):
So Part three of our dive looks at how AI
addresses this, not just by replacing people, but by moving
past the idea of simply finding bodies to fill.
Speaker 2 (16:39):
Roles exactly, the strategy shift is towards multiplying minds, using
AI to augment the capabilities of the existing team, making
them more effective, more productive, and maybe even making their
jobs better.
Speaker 1 (16:50):
Okay, multiply minds. Let's start with something traditionally very human
driven up selling, especially hyper personalized up selling. How does
AI change that game? Our sources talk about Connie two
point zero, right.
Speaker 2 (17:02):
Connie two point zero is Hilton's next gen AI concierge,
and the sophistication level here is really key. It's not
just a chatbot. It uses eleven Labs for the voice generation.
Speaker 1 (17:11):
Eleven Labs they do those hyperrealistic AI voices.
Speaker 2 (17:14):
Right exactly, makes the interraction feel much more natural, builds trust.
Then it uses the Lama three point two nine zero
B model, which includes a vision component.
Speaker 1 (17:22):
He communicates via in room tablets. Though why does the
vision part matter?
Speaker 2 (17:27):
Good question? The vision allows Connie to understand the context
within the room, even if it's just communicating via the
tablet screen. Is the TV on? Are the lights dimmed?
Has the guest repeatedly swiped away notifications? It uses these
visual cues to tailor the timing and style of its approach.
It's not just about language, it's about context.
Speaker 1 (17:45):
Okay, that's clever, and it speaks multiple.
Speaker 2 (17:47):
Languages fluently in fourteen languages, which vastly expands its reach
and effectiveness.
Speaker 1 (17:53):
So the sources actually put this to the test human
versus AI.
Speaker 2 (17:56):
Yep a direct ad test over a fourteen day period
cross twelve hundred rooms. They tracked conversion rates specifically for
late checkout up cells human concierge attempts versus Connie two.
Speaker 1 (18:07):
Point erh attempts and the results. What was the gap?
Speaker 2 (18:09):
It was significant. The human concierges, who were doing a
perfectly fine job by traditional standards, achieved a four point
one percent conversion rate on those late checkout offers.
Speaker 1 (18:18):
Okay, four point one percent seems reasonable. And Connie two point.
Speaker 2 (18:22):
Zero twenty eight point seven percent conversion rate whoa.
Speaker 1 (18:24):
Four point one percent versus twenty eight point seven percent.
That's almost seven times.
Speaker 2 (18:29):
Higher, nearly seven times yes, and it gets better. On
top of that massive lift in the primary upsell, Connie
two point zero also managed an additional two point four
percent conversion on food and beverage cross cells. How did
it do that seamlessly? It would confirm the late checkout
and then add something like, since you'll be enjoying the
room a bit longer, may I assist you with a
Brench reservation downstairs, natural relevant timely.
Speaker 1 (18:51):
Seven times higher conversion. It almost suggests the machine's consistency,
it's perfect memory, its lack of distraction. Maybe that outweighs
the human type for this kind of specific transactional up cell.
Speaker 2 (19:02):
That's certainly what the data implies. The AI is always on,
always polite, never having a bad day, and has instant
access to the guest's entire profile and preferences. The annualized
revenue lift from just this one upsell initiative across only
twenty two properties testing Connie two point zero two point
nine million dollars two.
Speaker 1 (19:17):
Point nine million dollars. Okay. That transforms a task that
was maybe a marginal benefit into a serious, scalable profit center.
Speaker 2 (19:23):
Precisely, it takes a necessary but often low yield human
interaction and makes it highly profitable and digitally scalable.
Speaker 1 (19:31):
All right, let's talk housekeeping. This is a huge labor department,
physically demanding work hard to optimize schedules. How does AI
help here? We're talking robots predicting dirt.
Speaker 2 (19:43):
Kind of Yeah. It's the combination of SoftBank's Whiz cleaning robot.
You might have seen these autonomous vacuum cleaners paired with
brain cores AI software, which acts as the brain for
the robot and the overall cleaning operation.
Speaker 1 (19:55):
So the AI isn't just driving the robot.
Speaker 2 (19:57):
No, it's much smarter than that. The AI's core. Our
task here is to optimize the entire cleaning schedule and
route plan by predicting the soil load of each room
before the human housekeeping team even goes near it.
Speaker 1 (20:08):
Hold on predict soil load? How on Earth does an
AI predict how dirty room is going to be? That
sounds like sci fi.
Speaker 2 (20:15):
It uses a few data streams. First, historical data from
the wizrobox themselves, surface reflectance data from previous cleans. It
learns which rooms tend to get dirtier faster based on
where and tear patterns.
Speaker 1 (20:26):
Okay, historical patterns.
Speaker 2 (20:28):
Makes sense, But here's the really fascinating bit. The source
is highlighted. It also analyzes aggregated anonymized guest activity data,
including believe it or not social media tags associated with
the hotel or area.
Speaker 1 (20:41):
It's looking at guest Instagram tags.
Speaker 2 (20:42):
Seriously anonymized and aggregated. Yes, it knows that guests tagging
things like hashtag party weekend, hashtag family reunion, maybe even
hashtag pet friendly stay are statistically far more likely to
generate a high soil load than say a solo business
traveler tagging hashtag conference life.
Speaker 1 (20:59):
Wow. So it's using external behavior signals to predict internal
cleaning needs exactly.
Speaker 2 (21:05):
This allows the system to dynamically prioritize housekeeping labor, send
more resources to the rooms predicted to be dirtiest, maybe
a just cleaning frequency for others.
Speaker 1 (21:14):
Okay, what was the actual impact at a large property.
Speaker 2 (21:16):
They tested this at a big eleven hundred room casino
hotel in Las Vegas. The result they were able to
cut the number of turndown services that evening refreshed by
nineteen percent just by optimizing based on predicted need.
Speaker 1 (21:27):
Nineteen percent fewer turndowns. That must free up a lot
of staff time.
Speaker 2 (21:31):
It freed up the equivalent of fourteen full time employees
fourteen fifties. But here's the crucial part. Guess satisfaction scores
actually went up by six points.
Speaker 1 (21:41):
Satisfaction went up even though they cut service.
Speaker 2 (21:43):
How because the AI wasn't just cutting randomly, it was
redirecting that freed up human labor. Staff spent more time
providing detailed, high quality cleaning and service to the high
touch areas, the premium suites, the rooms that really needed
that human attention, instead of wasting time on lately used rooms.
Speaker 1 (22:01):
Uh okay. So it's a perfect example of augmentation. The
machine handles the optimization and prediction, freeing up humans for
the high value, high touched tasks that guests really notice.
Speaker 2 (22:12):
Precisely, the robot vacuums the standard floor. The human adds
the fold and towel.
Speaker 1 (22:16):
Elephant moving from the physical room to the guest's memory,
or rather the hotel's memory of the guests. This idea
of the guest as chief memory officer. The sources mentioned
as Skift survey what was that finding?
Speaker 2 (22:27):
Yeah, A twenty twenty five Skift survey found that sixty
percent six zero percent of millennials said they would abandon
booking with a hotel brand if the hotel forgot a
preference they had previously stated, like their pillow type or
room temperature.
Speaker 1 (22:41):
Sixty percent would walk away over a forgotten pillow preference. Wow,
forgetfulness is now a direct business.
Speaker 2 (22:47):
Risk, a huge one. Loyalty is incredibly fragile. So to
combat this, leading edge hotels are building what the sources
call a memory graph.
Speaker 1 (22:55):
Okay, memory graph sounds complex. Explain that for someone deep
into database tech. How is it different from just a
regular customer database.
Speaker 2 (23:04):
Think of it less like a spreadsheet listing facts about
a guest and more like a dynamic, interconnected web of
their history and preferences with the brand. It's often built
on graph database technology like Neo four J, so it
links things together exactly. It doesn't just store attributes and
maps relationships and behaviors over time. It ingests data from
literally every single touch point. The preferences you tick on
(23:26):
a pre stay email form that goes in the command
you gave to the Alexa device in your room last
time in the system, reading your license plate as you
arrive in, even the tap from your RFID wristband ordering
a drink at the pool bar that gets added to
the graph too.
Speaker 1 (23:41):
So it's building this incredibly rich, multidimensional picture of every
guest interaction and Grock four is involved here too.
Speaker 2 (23:48):
Yeah. GROCK four acts as the continuous processing engine. It
constantly analyzes all this incoming data and updates what the
sources call a seven dimension preference vector for each guest.
Seven dimensions what things like their preferred climate settings, sleep
patterns do not disturb times, dietary needs or allergies noted previously,
leisure interests shown through activity bookings, even privacy sensitivity. GROCK
(24:12):
refines this vector constantly, so the system doesn't just store
your preference for feather pillows. It understands your overall sleep profile.
Speaker 1 (24:19):
The goal then is truly seamless, almost anticipatory service. Give
us that check in example again, what actually happens?
Speaker 2 (24:26):
Right, So you walk up to the desk before you
even get your key, the system has already acted. Your
guest folio is populated, but it's also triggered actions Room
fourteen twenty two. It's automatically being set to sixty eight
degrees fahrenheit because the graph knows that's your sweet spot.
A notification has pinged housekeeping ensure hypoallergenic duvet is on
(24:46):
bed for incoming guests. Two bottles of still water are
dispatched by room service and the do not disturb on
the phone system is pre activated from ten pm to
eight am based on your past behavior.
Speaker 1 (24:57):
And the guest hasn't asked for any of this during
this specific to check in.
Speaker 2 (25:00):
Nope, it just happens based on the memory graph, it's
anticipating needs before they're even expressed.
Speaker 1 (25:05):
That level of personalization of being truly remembered must have
a huge impact on loyalty. What's the metric.
Speaker 2 (25:10):
It's dramatic. The sources looked at repeat guest net promoter
scores MPs. The chain average was around sixty seven, which
is decent kind of satisfied properties. Using this memory graph system,
their repeat guest MPs jumped to ninety one ninety one.
Speaker 1 (25:26):
That's not just satisfied. That's deep into promoter or evangelist.
Speaker 2 (25:30):
Territory exactly, and that kind of loyalty shift it drastically
cuts your customer acquisition costs down the line and maximizes
lifetime value. It's a massive competitive mote.
Speaker 1 (25:42):
Okay, Part four, Let's dig into the financial core. Maybe
less sexy than predicting dirt or cloning voices, but absolutely
critical to the bottom line. We're talking energy, fraud and
attracting capital let's start with energy utilities. The sources call
the silent thirty two percent of op X.
Speaker 2 (25:59):
Yeah, spence line that never sleeps.
Speaker 1 (26:01):
It's relentless and notoriously hard to control, right because it's
driven by guest behavior, which is all over the place exactly.
Speaker 2 (26:07):
Traditional HVAC systems and hotels are often pretty dumb. They
rely on basic motion sensors, passive infrared or PR. They
only know if someone is in the room right now
or was recently. Lots of wasted heating and cooling and
empty rooms.
Speaker 1 (26:20):
So what's the AI powered upgrade? The TEXTAC mentioned is
Verdant's ZX thermostat paired with AWS in front of two chips.
What's special there.
Speaker 2 (26:27):
It's about moving from simple detection to sophisticated prediction. These
new sensors are hyper sensitive. They're sampling data, temperature, humidity, motion,
even tiny changes in magnetic fields from balcony doors opening
or closing two hundred.
Speaker 1 (26:41):
Times per second, two hundred times a second.
Speaker 2 (26:43):
Okay, that fire hose of data is processed right there
on the device. That's the edge machine learning part. Using
those AWS inferentiate chips, it's not just looking for simple motion,
it's analyzing patterns in that high frequency data stream to
predict occupancy and guests intent. And it can do this
on average twenty three minutes earlier than a traditional PIR
(27:04):
sensor figures out the room is truly.
Speaker 1 (27:06):
Vacant twenty three minutes earlier. Why does that twenty three
minute hit start matter so much for energy savings?
Speaker 2 (27:11):
It allows the HVAC system to start coasting much sooner.
If the AI detects subtle patterns, maybe the guest is
packing based on movement, or the balcony door sensor shows
repeated short opening suggesting they're about to leave. It begins
to slowly relax the temperature controls. It lets the room
drift slightly towards the unoccupied set point before the guest
actually walks out the door, and the old PIR sensor
(27:32):
finally notices.
Speaker 1 (27:34):
I see it's squeezing out savings in that twenty three
minute window. Room by room, day by day predictive climate
control based on behavioral micro signals. What's the financial impact
when you scale that up?
Speaker 2 (27:44):
The sources looked at a large real estate investment trust
a rate with one hundred and eighty properties using this system.
On average, it shaved one point eight killowat hours kill
what hours per occupied room night?
Speaker 1 (27:56):
One point eight killowaters per room night doesn't sound like
a lot on.
Speaker 2 (27:59):
Its own, but multiply that across one hundred and eighty hotels,
thousands of rooms, three hundred and sixty five days a year.
It added up to eleven point four million dollars in
annual utility savings for the REAP.
Speaker 1 (28:10):
Eleven point four million dollars.
Speaker 2 (28:11):
Wow, And for the ESG folks that also meant an
eighty two hundred metric ton reduction in their carbon footprint.
Huge co benefit.
Speaker 1 (28:20):
And the kicker for the finance team the payback.
Speaker 2 (28:22):
Kayback period on the capital investment for the Verdana WS
system just nine point two months less than the.
Speaker 1 (28:28):
Year to get their money back, and then pure savings afterwards.
That makes it almost a financial noe brainer.
Speaker 2 (28:32):
Doesn't Absolutely it shifts from a nice to have green
initiative to a core financial imperative.
Speaker 1 (28:36):
Okay, let's talk payments fraud compliance Another area where regulations
are getting tougher. PCI DSS four point zero point one
was mentioned, demanding continuous behavioral compliance. What does that mean?
In practice?
Speaker 2 (28:51):
It means the old way of just tokenizing car data
isn't enough anymore. Regulators now expect merchants, including hotels to
continuously monitor the bhaavior around transactions, especially for card not
present payments, to detect anomalies and potential risks in real time.
It's a much higher bar.
Speaker 1 (29:08):
So how are hotels meeting this? What's the AI stack
look like? Here?
Speaker 2 (29:11):
It's often a combination basis theory vaults for the secure
tokenization and handling of the rock hard data. Then anthropics
Claude constitutional AI again remember good for rule adherence and classification,
and data dog row user monitoring to capture the behavioral
signals during the online booking process.
Speaker 1 (29:27):
Okay, so Claude is analyzing the transaction details and the
user behavior and then what scoring it?
Speaker 2 (29:33):
Exactly? Every card NOTT present transaction gets a risk score
save from zero to one hundred based on hundreds of factors.
IPHS reputation velocity, checks device fingerprint, how the user interacted
with the payment.
Speaker 1 (29:44):
Form, and if the score is too low, if it
looks dodgy.
Speaker 2 (29:47):
Transaction scoring below a certain threshold. The example given was
ninety two are automatically quarantined. They don't get processed immediately.
They're flagged for human review.
Speaker 1 (29:57):
How do they speed up that human review is a
list of block transactions?
Speaker 2 (30:02):
No, much smarter. The system automatically generates a short like
nine second loom style video recording of the anomalous user session.
It shows the reviewer exactly what looked suspicious. Maybe rapid
changes and address input, attempts to use multiple cards, connections
from known risky networks. Gives them instant context to make
a decision.
Speaker 1 (30:22):
That's efficient. So what was the big surprise finding when
they implemented this? Was it catching lots of external hackers?
Speaker 2 (30:28):
Actually? No, The biggest finding, the one that generated significant
unexpected revenue recovery, was an external fraud.
Speaker 1 (30:35):
Wait was it? Then?
Speaker 2 (30:36):
It was identifying what the industry politely calls friendly.
Speaker 1 (30:40):
Fraud, friendly fraud like what?
Speaker 2 (30:42):
In this case, the system identified that about zero point
seven percent of all reservations were being made by corporate
travel bookers using their personal credit cards, often to romly
claim loyalty points or benefits, or maybe booking non compliant
rooms or rates. Effectively, it was internal policy violation and
misuse bleeding money.
Speaker 1 (31:01):
Wow. And the system caught this pattern, yes.
Speaker 2 (31:04):
By analyzing booking patterns, card types, loyalty linkages, and corporate
account rules. The recoveries from clamping down on just this
specific type of friendly fraud totaled nine hundred and forty
thousand dollars in the first year for the hotel groups started.
Speaker 1 (31:17):
Nearly a million dollars recovered just from applying behavioral AI
to the payment stream to catch internal misuse. That's pure
cash flow straight to the bottom line.
Speaker 2 (31:26):
Absolutely unexpected, but highly valuable.
Speaker 1 (31:28):
Okay, So this level of operational tightening, compliance hardening, fraud reduction,
energy saving, it has to change how investors look at
these assets, right. The sources made a really strong point
about an AI premium in capital markets.
Speaker 2 (31:41):
It's huge investor demand is now actively dictating the pace
of AI adoption and hospitality. We're talking about forty one
billion dollars in private equity dry powder earmarked for the sector.
But the limited partners, the LPs who provide that money,
they are now explicitly asking for AI readiness scores on
potential hotel acquisition before they'll even approve a term sheet.
Speaker 1 (32:02):
So if you're running an older hotel portfolio without these
AI systems, you're basically seen as riskier, we less attractive significantly,
So you're carrying operational inefficiency, compliance risk, potentially higher energy costs,
things that AI demonstrably fixes.
Speaker 2 (32:17):
Is there a quantifiable gap invaluation?
Speaker 1 (32:20):
Yes, and it's substantial. The sources indicate that sophisticated buyers
like PE firms are currently paying around eight point two
times EBITA for hotel assets they deem AI native or
properly AI integrated eight.
Speaker 2 (32:32):
Point two x okay, and for legacy portfolios without.
Speaker 1 (32:34):
This tech, the multiple drops to around six point one
x ebitda.
Speaker 2 (32:38):
Wow, that's a two point one turn difference in valuation
multiple just based on AI integration. That premium is very real,
very real, and likely growing because the returns from implementing
AI are so fast and so predictable. As we've seen,
investors see it as de risking the asset while simultaneously
unlocking significant upside.
Speaker 1 (32:56):
We really need to dig into those rapid returns. The
sources laid out this almost like an investor's ninety day playbook,
a hyper accelerated plan designed specifically to boost that AI
readiness score and capture that premium. Walk us through it.
Speaker 2 (33:10):
Yeah, it's designed for speed and impact, focused entirely on
the highest ROI fastest payback initiatives okay.
Speaker 1 (33:17):
First steps weeks one and two.
Speaker 2 (33:18):
Deploy an AI powered expense crawler think tools like RAMP,
maybe brex integrated with something like GPT four to ROH.
It automatically scans the general ledger, categorizes spending, benchmarks it
against industry data, and instantly flags outliers, potential savings or
non compliant spending.
Speaker 1 (33:36):
Just automated expense review. What kind of savings does that
typically find?
Speaker 2 (33:39):
On average? That step alone identifies around four point two
percent in immediate systemic savings across various expense lines quick
win number one.
Speaker 1 (33:47):
Okay, four point two percent savings in two weeks. What's next?
Weeks three through six?
Speaker 2 (33:52):
Attack the operational friction points we discussed earlier. Automate the
night audit process using tools like museflow or similar integrations.
And automate accounts payable invoicing using platforms like userbill dot com.
Speaker 1 (34:05):
The back office stuff. What's the ROI on that module?
Speaker 2 (34:07):
Based on the labor savings and accuracy gains we talked about,
this typically yields around a three hundred and forty percent
ROI within the first year.
Speaker 1 (34:14):
Massive payback Wow, okay, So expense crawl, then back office automation.
What about revenue? When does that come in?
Speaker 2 (34:20):
That's the focus for weeks seven through twelve. The playbook
calls for spinning up a revenue war room slack bot.
Speaker 1 (34:27):
A slack bot for revenue. How does that work?
Speaker 2 (34:29):
It uses AI to constantly monitor market signals, competitor rate changes,
flight loads, event calendars, even sentiment analysis from reviews. It
live scores the hotel's own pricing decisions, and pushes proactive
recommendations and alerts directly to the revenue management team via
slack comp set just drop weekend rates, recommend holding firm,
or sudden spike in searches for dates x toy consider
(34:51):
raising bar by twenty dollars. It creates this constant feedback.
Speaker 1 (34:55):
Loop, so continuous AI powered decision support for the revenue team.
What's the typical profit lift from that?
Speaker 2 (35:01):
Generally looking at it, eight percent to eleven percent lift
in GOP gross operating profit.
Speaker 1 (35:06):
Okay, so let's tell this up, this whole ninety day
three phase playbook. What's the estimated cost to implement these
initial systems.
Speaker 2 (35:14):
The sources cite a total cost of around thirty eight
thousand dollars for the software. An initial setup for these
specific high impact.
Speaker 1 (35:21):
Plays thirty eight thousand dollars and the typical financial impact
in the first year.
Speaker 2 (35:25):
The typical first year EBITDA lift generated by this thirty
eight K investment is estimated at one point one million dollars.
Speaker 1 (35:32):
One point one million dollars from a thirty eight K spend.
Speaker 2 (35:35):
Which works out to a what is that roughly a
twenty nine hundred percent return on investment in your one.
Speaker 1 (35:41):
Twenty nine hundred percent ROI If you're an investor or
an owner looking to sell or refinance, that kind of
math is just undeniable. It makes AI integration absolutely critical.
Speaker 2 (35:50):
It becomes the single most important factor in maximizing asset
value in today's market period.
Speaker 1 (35:56):
All Right, our final section, and it's a crucial one.
The human element, the creative side, because there's obviously a
lot of fear around AI taking jobs, but the data,
like that McKenzie report from twenty twenty five, seems to
paint a different picture.
Speaker 2 (36:10):
It really does. The data is pretty consistent on this.
McKinsey predicted that about fifty four percent of tasks within
hotel jobs could be augmented by AI, meaning AI helps
the human do the job better or faster, but only
around twelve percent of current hotel jobs were projected to
be fully automated away in the medium term.
Speaker 1 (36:28):
So augmentation, not replacement, is the main story.
Speaker 2 (36:30):
Overwhelmingly, Yes, the real focus for smart organizations isn't just
cutting heads. It's about retaining their best talent, upscaling them,
and fundamentally changing their roles to work with the AI,
giving them better tools to do more interesting work.
Speaker 1 (36:43):
Marriott's AI Solmelia program sounds like a prime example of this.
Tell us about that, what does an AI so milier learn.
Speaker 2 (36:49):
It's a really interesting initiative. It's a dedicated four week
training program designed for high performing frontline staff, concierges, front
desk agents, supervisors. They're not learning hospitality basics, they're learning
how to manage and leverage the hotel's AI systems.
Speaker 1 (37:07):
Often's the curriculum things like.
Speaker 2 (37:09):
Prompt engineering basically how to talk to the large language
models effectively to get the best results, bias auditing learning
how to spot and mitigate potential biases in the algorithms
to ensure fairness, and even robot chaperoning. The practical skills
needed to manage and troubleshoot physical robots like those whiz
cleaners we discussed.
Speaker 1 (37:28):
So they're training staff to become supervisors of the AI
in robotics exactly.
Speaker 2 (37:32):
They're teaching them to own the technology, not be replaced
by it, turning them into masters of the new operational infrastructure.
Speaker 1 (37:38):
That's a powerful shift in mindset. What were the results?
Did it help with retention?
Speaker 2 (37:43):
The results were profound according to the sources, and amazing.
Ninety two percent of the employees who graduated from the
Aismeliar program were promoted within nine months.
Speaker 1 (37:52):
Ninety two percent promoted.
Speaker 2 (37:53):
Wow, and maybe even more importantly, overall, staff attrition turnover
in the properties participating in the program fell by thirty one.
Speaker 1 (38:01):
Percent, down by nearly a third. Yeah.
Speaker 2 (38:04):
When employees feel empowered by technology, when they see it
creating new career paths for them, when they feel like
they own the algorithm instead of being victims of it,
their loyalty to the company's skyrockets, that's huge.
Speaker 1 (38:16):
Yeah. Okay. Beyond optimizing standard operations, AI seems uniquely good
at spotting and exploiting these weird, unexpected micro demand opportunities
things humans might miss entirely, these edge cases that become
new best practices absolutely.
Speaker 2 (38:30):
The speed of AI allows for incredibly opportunistic revenue capture.
The source material had a great example, the K pop pop.
Speaker 1 (38:37):
Up K Pop pop Up tell Me More.
Speaker 2 (38:39):
Okay, so a major K pop ban announces a surprise
concert in Atlanta just seventy two hours notice, Chaos ensues
among fans. The Omnia Hotel at CNN Center looks at
its inventory. They only have forty one rooms left for
those dates, a tiny number.
Speaker 1 (38:53):
So normally they just sell those out quickly at a
decent rate. Maybe maybe.
Speaker 2 (38:58):
But their AI cluster this time, a combo of Duato
for pricing, grock four for the real time context analysis,
and crucially, an API feed from ticket Master spotted something else.
It detected an immediate spike of one hundred and eighty
thousand searches on secondary markets just for tickets to that
specific consuit in the Atlanta area.
Speaker 1 (39:16):
One hundred and eighty thousand searches. Okay, that signal's intense
niche demand exactly.
Speaker 2 (39:22):
The AI understood the specific rabid fan base being activated,
so it didn't just raise rates slightly. It automatically created
and priced a unique package, the the IPK Pop pre
Show Experience or something similar bundling the room with some
themed amenities, and it priced this package at a jaw
dropping three hundred and eighty percent above their normal average
daily rate for those room types plus three.
Speaker 1 (39:43):
Hundred and eighty percent eight R Did anyone actually buy it?
Speaker 2 (39:46):
Those forty one rooms sold out in eleven minutes. Eleven minutes,
generating one hundred and eighty four thousand dollars in gross
revenue from just forty one rooms, purely because the AI
spotted the micro demand signal and priced it hyper aggressively instantly,
that kind of.
Speaker 1 (40:00):
Speed, integrating ticketing data, creating a package repricing selling out
in minutes. A human revenue team, it couldn't do that.
Speaker 2 (40:06):
Not a chance. It perfectly exploits that short term volatility.
Speaker 1 (40:09):
Okay. Another example was about capacity on demand, the insurance
adjuster swarm.
Speaker 2 (40:14):
Right, different scenario, same principle of AI speed and scale.
Hurricane Leela hits Mobile, Alabama, major damage. Suddenly about fourteen
hundred insurance claims. Adjusters descend on the city, all needing rooms,
often for extended stays.
Speaker 1 (40:29):
Huge sudden influx of very specific demand. How did one
hotel capitalize.
Speaker 2 (40:35):
A local hampton and realized they needed to reach these
adjusters immediately and personally before competitors lock them up. But
calling fourteen hundred people is impossible. Sending a generic email
blast gets lost.
Speaker 1 (40:46):
So what did they do?
Speaker 2 (40:47):
They got creative with AI. They used eleven Labs the
voice cloning tech again to create a message using the
actual voice of their general manager, a familiar, trusted local voice.
Speaker 1 (40:57):
Okay, clone the GM's voice. Yeah, then what?
Speaker 2 (40:59):
They then use an AI platform to spin up forty
thousand personalized voicemails using that cloned voice. Each voicemail perhaps
slightly tweaked with the adjuster's name or company if they
had that data from registries. These voicemails offered special extended
stay rates maybe day rates for using the lobby to work,
options tailored specifically to what adjusters need during a catastrophe resports,
(41:21):
and they mass delivered these voicemails instantly.
Speaker 1 (41:24):
Forty thousand personalized voicemails delivered.
Speaker 2 (41:26):
Instantly captured sixty two percent of that incoming adjuster market
share almost immediately, because the outreach felt personal, timely, and
directly addressed their specific needs using a trusted voice at
a scale no human team could ever achieve.
Speaker 1 (41:40):
Again, it's using AI not just for internal efficiency, but
as a tool for creative, scalable outreach and revenue generation.
These examples are powerful so for our listener who's maybe
feeling a bit overwhelmed but also inspired, thinking, Okay, I
need to get started. What's the practical roadmap? Where do
you begin? On Monday morning?
Speaker 2 (41:58):
The sources actually lay out four pretty concrete first steps,
things that separate the hotels actually doing this from the
ones just talking about it. Step one, audit your data
mode before you even think about AI tools. Look at
your own data. Export the last ninety days of data
from your core systems, your PMS, your point of sale pos,
your central reservation system crs. Really examine it. How complete
(42:22):
is it? How clean is it? Are there gaps, missing fields, inconsistencies?
Score it. If your data completeness and quality score is
below about eighty five percent, the sources say, stop, fix
your data first. AI runs on data garbage in, garbage out,
but with AI, it's exponential.
Speaker 1 (42:37):
Garbage app foundational. Get your data house in order first. Okay.
Speaker 2 (42:41):
Step two pick one bleeding wound, don't try to boil
the ocean, and implement AI across the entire hotel at once.
Focus find the one area that's causing the most financial
pain right now. Is it those night audit labor overages?
Is it spiking utility costs you can't explain? Is it
terrible upsell conversion rates? Pick the single biggest, clearest pain
point with a measure of financial impact.
Speaker 1 (43:01):
Focus for impact makes sense. Step three sounds counterintuitive for
big companies right.
Speaker 2 (43:05):
Step three is buy, don't build. Resist the urge to
create your own bespoke AI platform from scratch. The vast
majority of the sources, say eighty four percent of hotels
successfully implementing AI are using established off the shelf solutions
from vendors, maybe with some fine tuning or integration help.
Speaker 1 (43:25):
Why buy instead of build?
Speaker 2 (43:26):
Speed to value? Buying a proven solution gets you results
much faster. The average time to value for these off
the shelf AI tools is around forty one days. Building
your own takes years, huge investment, high risk of failure.
In this market, speed matters. Capture that AI premium.
Speaker 1 (43:42):
Now, okay, buy, don't build for speed, and the final
step step.
Speaker 2 (43:46):
Four ring fence the first win. Once you've picked your
bleeding wound step two and implemented a solution step three,
you absolutely must track its impact obsessively and communicate the success.
Create a simple dashboard could be Google Data Studio, could
be power Yi. It doesn't have to be fancy showing
the daily or weekly KPIs for that one specific initiative,
track the ROI and share that dashboard relentlessly with department heads,
(44:08):
maybe even all staff.
Speaker 1 (44:09):
Why is that communication so important?
Speaker 2 (44:11):
Because momentum is oxygen for change. That first undeniable, measurable
win builds credibility, It generates excitement, its cures buy in
for the next AI project. You need to celebrate and
broadcast that initial success to fuel the ongoing transformation.
Speaker 1 (44:28):
This has been well quite the deep dive. It feels
like we've fundamentally rewritten the operating manual for hospitality in
the last half hour to quickly recap the core findings.
The industry's financial foundations are definitely under pressure from costs.
But AI isn't just a potential solution. It's an immediate, proven,
and highly measurable escape route.
Speaker 2 (44:47):
Yeah, the evidence is compelling.
Speaker 1 (44:49):
Now we're seeing these integrated AI systems delivering what was
it seven percent to fourteen percent swings in EBITDA, and
not over years, but within quarters. The hotels that are
truly winning in twenty twenty five and beyond. They're not
necessarily the ones running cheaper, They're the ones running profoundly smarter.
Speaker 2 (45:04):
And this internal transformation within hotels it's about to get
a major boost from the outside, a macro tailwind. Remember
the Federal Reserve their November twenty twenty five meeting minutes
strongly signaled that interest rate cuts are likely coming, probably
by early twenty twenty six.
Speaker 1 (45:19):
Okay, cheaper money is coming.
Speaker 2 (45:21):
Exactly, So combine cheaper debt with this AI amplified cash flow.
Think back to that incredible twenty nine hundred percent ROI
from the ninety day Investor Playbook, and what do you get?
Speaker 1 (45:32):
A feeding frenzy, massive acquisition fever.
Speaker 2 (45:36):
You got it. The valuation premium for these AI native
high performing assets is only going to increase. It creates
this really urgent window right now for operators to get
their house in order, implement these systems and capture that value,
whether they plan a hold or sell.
Speaker 1 (45:50):
A real gold rush for operational efficiency and intelligence. Okay,
here's where my mind goes though, and maybe something for
you the listener, to really chew on as we wrap up,
I can predict with high accuracy that a room will
be empty twenty three minutes from now, allowing the HVAC
to power down preemptively, and if AI can access that
seven emission preference dector we talked about and automatically configure
(46:12):
your room perfectly. Temperature pillows, water do not disturb before
you even swipe your key.
Speaker 2 (46:17):
What becomes obsolete? What old ways of doing things.
Speaker 1 (46:20):
Just disappear exactly? What traditional way do we even measure
customer satisfaction? When this system anticipates needs so perfectly? Does
a paper comment card left on the desk have any
meeting anymore? Does even a standard post day email survey
asking how is your stay? Capture anything useful? When the
goal shifts from reactive problem solving to proactive perfect anticipation, It's.
Speaker 2 (46:44):
A fundamental shift. The source material had a provocative line.
Profit is the new loyalty. Maybe loyalty isn't about points anymore.
It's about the seamless, effortless, perfectly remembered and anticipated experience
that AI enables.
Speaker 1 (46:57):
So the final thought for you, what happens to the
deaf nition of good service when every single manual click,
every operational data point, every guest interaction instantly becomes part
of a learning loop for a machine that never forgets
and constantly optimizes, What does performance measurement look like when
the ultimate goal is that the customer never actually has
to ask for anything because it's already been anticipated and delivered.
That's the future AI is building in hospitality right now.
Welcome back to the deep dive. Today, we are focusing
on well, one of the biggest sectors out there, hospitality,
and it's facing some fundamental challenges right now. Yeah. Absolutely,
We're not here today to talk about you know, fancy
lobbies or loyalty points. We are really digging into the
financial nuts and bolts, specifically what our sources are calling
(00:20):
the margin squeeze. No one saw coming.
Speaker 2 (00:23):
And that squeeze, I mean, as of late twenty twenty five,
it's become a real structural threat. It's not just a blip.
When the Bureau of Labor Statistics data dropped in September,
it painted a pretty clear picture and frankly, it was painful.
Operational costs are just surging way faster than revenue.
Speaker 1 (00:38):
How bad are we talking, Well, look.
Speaker 2 (00:40):
At the numbers. Energy costs are up four point eight
percent year over year, food and beverage spiking six point
two percent, and wages you know for the essential staff
everyone needs, still climbing at four point nine percent.
Speaker 1 (00:51):
Okay, so costs are really heading upwards fast. What about
the other side? Is revenue keeping pace at all our
hotels managing to charge more?
Speaker 2 (00:59):
That's the problem. It's not accelerating revpair growth. You know,
revenue per available room, the key metric. It's basically crawled
to a stop across the America's just one point one percent.
Speaker 1 (01:09):
Wow, one point one percent. That's sluggish.
Speaker 2 (01:11):
It's the weakest recovery we've seen since well since the
initial bounce back after the pandemic first hit. So think
about it. If you're running a standard, say two hundred
and fifty room hotel, that cost surge. It means you're
suddenly eating an extra one hundred and eighty seven thousand
dollars an annual operating collage, and you can't just jack
up your rates to cover that, not without losing guests
(01:33):
to the competition. The math, as the sources really hammer home,
it's just it's utterly merciless.
Speaker 1 (01:39):
Okay, so that sounds pretty bleak across the board, universal
distress almost, But this deep dive, it isn't just about
the pain, right, It's about the ones who are figuring.
Speaker 2 (01:48):
It out exactly.
Speaker 1 (01:49):
Our sources highlight this huge contrast. You've got properties really
struggling and then others others are suddenly pulling away financially massively.
Speaker 2 (01:56):
That's the core story here, that distinction because right in
the middle of that tough market, dealing with the same inflation,
the same flat revenue. There's this uh three hundred and
twelve room autographic Collection hotel and they recently posted a wow,
a staggering seven point four percent lift in their ebit
dust seven.
Speaker 1 (02:16):
Point four percent in this climate.
Speaker 2 (02:18):
How not by cutting corners, not by slashing staff or
cheapening the guest experience. Note they did it through very specific,
very integrated deployments of AI artificial intelligence and RPA robotic
process automation. And the result a twenty two point swing
in labor productivity.
Speaker 1 (02:33):
Well, hang on a twenty two point swing. That's not
just tweaking things. That's structural, that's a different operating model exactly.
Speaker 2 (02:37):
It's fundamental change, not just you know, incremental improvement.
Speaker 1 (02:40):
So the mission for you listening today, whether you're an
owner and operator, maybe an investor looking at the space,
is to really get the blueprint here. We need to
unpack how these specific winning hotels are using tools like
GPT four to zero, GROG four, Claude three point five sonnet.
We're moving beyond just automating simple tasks. This is about
building systems that can predict issues they can almost self
(03:04):
heal operationally, turning that what was it, one hundred and
eighty seven thousand dollars drag, Yeah, into a genuine competitive advantage.
Speaker 2 (03:12):
And you have to understand, this isn't just theory anymore.
It's not pilots. When a few properties start showing these
kinds of returns, it immediately resets the bar. It redefines
the baseline expectation for profitability for everyone in the sector.
Speaker 1 (03:25):
All right, so the pressure's on too, Adapter get left
behind pretty much. Yeah, Okay, let's jump straight into the
operational weeds then, because the sources, they really showed that
the first big wave of savings, the really massive ones,
come from tackling that friction in the back office, you know,
the stuff where accuracy used to mean hours and hours
of human effort.
Speaker 2 (03:43):
Yeah, and mistakes, oh absolutely. And if you talk to
any GM, any hotel manager and ask them what the
single most dreaded, high stress but ultimately low value pass is,
they'll all say the same thing. The night audit.
Speaker 1 (03:56):
Ah, Yes, the infamous night audit, the traditional like two
point four or five am fire drill. That's where all
the gremlins crawl out right. Folio mismatches, reservations that just
seem to float in the limbo, those city ledger accounts
that never quite balance.
Speaker 2 (04:09):
Yep, all of it. It's manual. It's incredibly stressful for
the person doing it, and it's so easy to make
mistakes under pressure.
Speaker 1 (04:16):
Our sources actually track this a mid scale chain, I
think it was in the Midwest. They clocked an incredible
three hundred and twelve minutes of manual touches per property
per property during that audit, every single night.
Speaker 2 (04:28):
Think about that. That's over five hours of labor just
poured into this reconciliation black hole, night after night.
Speaker 1 (04:36):
So that huge time sink is exactly what this new
AI stack is designed to just obliterate precisely.
Speaker 2 (04:43):
We're looking at a combination system. Here. You've got the
property management system or PMS, let's say, use, then the
digital guest platform like Canary Technologies, your accounting system quick
books Online is common, and then the brain connecting them
all a large language model or LLM like GPT four
h Break that.
Speaker 1 (05:00):
Down for me for listeners who aren't you know, database wizards,
What does it actually mean when GPT four to oh
writes and executes complex SQL reconciliation scripts. Is the AI
literally coding on the fly.
Speaker 2 (05:13):
That's exactly the power shift. It's not just finding problems,
it's fixing them. So first canaries AI engine pulls in
all the raw transaction data. We're talking maybe fourteen thousand
individual transactions every night.
Speaker 1 (05:26):
Fourteen thousand. A human auditor would have to manually.
Speaker 2 (05:29):
Check all that, or at least sample heavily and hope
for the best. Now GPT four to oh, it's been
trained on millions of past audit logs, all the accounting rules,
all the reconciliation procedures. So when it spots something weird
like an orphaned payment, it doesn't just flag it. It
writes the specific SQL query needed to go into quick books,
compare the relevant fields, find the match, and then write
the correcting entry.
Speaker 1 (05:50):
And it does this how quickly.
Speaker 2 (05:51):
It executes that entire fine diagnosed right code, execute fixed
process in just zero point eight seconds.
Speaker 1 (05:57):
Wait, let me get this straight. Three hundred and twelve minutes.
It's five hours of manual, error prone, high stress work
is now done in less than one second. That sounds
almost unbelievable.
Speaker 2 (06:07):
Too clean, It's the new reality. This is the metric shift.
We're talking about. But you know, you need checks and balances,
so the system doesn't actually achieve zero human time.
Speaker 1 (06:15):
Okay, right, yeah, there has to be some oversight.
Speaker 2 (06:17):
Yeah. If the AI finds an anomaly it can't resolve
with high confidence or something really unusual, it automatically routes
it to a human auditor via slack. But and this
is key, that human only has a very tight, like
twelve second window to veto the AI's proposed fix.
Speaker 1 (06:33):
Twelve seconds that's not much time.
Speaker 2 (06:35):
It forces focus. If the human looks at the proposed
fix and it seems correct, they do nothing. The system
then automatically executes the correction. If they get veto, it
gets flagged for deeper manual review later. So the total
process time it drops from three hundred and twelve minutes
down to about three seconds three.
Speaker 1 (06:52):
Seconds okay, and the financial impact immediate.
Speaker 2 (06:54):
You're looking at annual labor savings of around forty one
two hundred dollars per property and the accuracy practically perfect
ninety nine point nine nine three percent.
Speaker 1 (07:04):
Wow. Okay, so forty one thousand dollars save, huge accuracy jump.
But what happens to the actual night auditor you mentioned
the human veto. But does this mean fewer jobs? The
sources must touch on this right, augmentation versus replacement.
Speaker 2 (07:17):
They do, and it's a critical point we'll revisit later.
The human role fundamentally changes. Instead of spending five hours
chasing ghosts in the data, the auditor now maybe spends
five minutes reviewing the handful of really complex exceptions the
AI couldn't crack, and the rest of their shift they're
freed up for higher value work, maybe preparing reports for
the owners, analyzing trends, even helping out the front desk
(07:40):
during a late rush. It's an elevation of the role,
not just elimination.
Speaker 1 (07:43):
Okay, that makes sense, shifting skills. So moving from saving
labor to avoiding well potentially massive costs. The sources talk
a lot about dynamic compliance, turning compliance from this reactive
headache into a proactive shield.
Speaker 2 (07:57):
This is a really important differentiator. It shows a i'maturity
beyond just basic automation. Think about privacy regulations. In Q
two twenty twenty five, the California Consumer Privacy Act the
CCPA got expanded significantly. It now covers transient guest PII
personally identifiable information.
Speaker 1 (08:16):
Okay, some names, addresses, maybe credit card details, Yeah, that.
Speaker 2 (08:20):
Kind of thing exactly. And the rules around redacting that data,
securing it, managing consent they got much stricter before AI
Manually complying with this for every single guest. Our sources
estimated it cost about twenty eight dollars per reservation just
for the compliance piece.
Speaker 1 (08:36):
Twenty eight dollars per booking. If you're running a big hotel,
say at four hundredroom resort in California, that adds up
incredibly fast.
Speaker 2 (08:43):
That's paralyzing, totally unmanageable, so the resort tracked. In the sources,
they implemented what's called the compliance Fortress. It's a stack
using Orcato for the automation workflow, one Trust for the
governance and policy side, and crucially Claude three point five
sonnet the AI from Entthropic.
Speaker 1 (08:58):
Interesting why Claude specific for this? We've mentioned GBT four
and GROC already. What's different here.
Speaker 2 (09:04):
Claude, particularly the three point five sonnet model, is built
around what Anthropic calls constitutional AI. It's designed to be
very good at adhering to complex strict rule sets and
explaining its reasoning. It excels that classification tasks where accuracy
and auditibility are paramount perfect for compliance.
Speaker 1 (09:23):
So how does this fortress actually work? When does it
kick in?
Speaker 2 (09:26):
The trigger is guest check in the moment that happens
in the PMS, it fires off an encrypted webhook basically
a secure notification. Claude receives the guest data in just
forty milliseconds. That's faster than you can blink. It analyzes
and classifies up to forty two different data fields name, email,
birth date, maybe even loyalty number, payment tourken. It identifies
(09:48):
everything that falls.
Speaker 1 (09:48):
Under CCPA forty milliseconds. Wow.
Speaker 2 (09:51):
Then where CATO takes over automatically redacting or tokenizing the
sensitive PII based on Claude's classification and the rules set
in one trust, and every single action is logged for
a complete audit trail.
Speaker 1 (10:01):
The speed is one thing, but the cost difference.
Speaker 2 (10:03):
Must be huge phenomenal. The cost per reservation for CCPA
compliance drops from twenty eight dollars down to zero zero
zero zero zero four basically zero.
Speaker 1 (10:12):
Okay, that's efficiency. Yeah, but you mentioned avoiding costs too, right.
Speaker 2 (10:16):
This is the kicker. The real financial game changer isn't
just the efficiency saving. It's the avoided penalties. In just
the first ninety days after rolling this out, this system
actively prevented eighty four separate potential regulatory violation.
Speaker 1 (10:30):
Eighty four violations. What's the cost of a.
Speaker 2 (10:32):
Violation under the expanded CCPA, it can be up to
fourteen thousand dollars per violation, So eighty four times fourteen
thousand dollars.
Speaker 1 (10:41):
Let's carry the one over one point one million dollars.
Speaker 2 (10:44):
One point two million dollars. Yeah, in avoided fines in
just three months.
Speaker 1 (10:47):
Okay, that single statistic, one point two million dollars saved
in ninety days. It completely reframes compliance. It's not just
a cost center anymore. It's a massive profit protector exactly.
Speaker 2 (10:58):
It moves from being operational drag to a strategic financial asset.
Speaker 1 (11:02):
All right, let's pivot. We've talked cost cutting compliance shields.
Now let's talk about actually generating more revenue radically more.
Part two in our Sources digs into what they call
revenue management three point zero.
Speaker 2 (11:13):
Yeah, this is exciting stuff. It's about moving beyond the
traditional approach, which, well, traditional revenue management is mostly about
optimizing for known demand, looking at your booking pas historical trends,
compset pricing, trying to price correctly based on what you
think is going to happen.
Speaker 1 (11:29):
Okay, reacting to patterns exactly.
Speaker 2 (11:32):
Revenue management three point zero powered by predictive AI is
about optimizing for unknown or unforeseen demand. It's about pricing
the future before that future even fully materializes.
Speaker 1 (11:43):
Pricing the future. I like that. So what's the big
example here?
Speaker 2 (11:47):
The primary case study is fantastic. It's the seventy two
hour flash storm that hit Boston back in October twenty
twenty five.
Speaker 1 (11:53):
Okay, set the scene for us. This wasn't like a
predictable blizzard everyone knew was coming, No, not at all.
Speaker 2 (11:58):
It was a severe, quite unexpected, and it caused chaos
at Logan Airport, a sudden thirty one percent spike in
flight cancelations, stranded passengers everywhere.
Speaker 1 (12:06):
So normally every revenue manager in Boston would be scrambling
right pulling old storm data, trying to figure out how
to fill rooms suddenly empty from cancelations, while maybe capturing
some walkins precisely.
Speaker 2 (12:17):
But that traditional response is inherently reactive. It relies on
lagging indicators. What has happened the Liberty Hotel in Boston,
though they're running a different kind of system. Was their
stack dueto game changer for the revenue management core, Google
big Query for handling massive data sets, and crucially Grock
four as the AI analysis engine.
Speaker 1 (12:38):
Grock four again. Why Grock here?
Speaker 2 (12:40):
Grock's strength, according to the sources, is its ability to
ingest and synthesize vast amounts of real time, rapidly changing
public data, very very quickly. It excels at connecting disparate
data points.
Speaker 1 (12:51):
Okay, so what happened at the Liberty?
Speaker 2 (12:52):
So at six point one s of am Eastern time,
the storm is rapidly getting worse. Cancelations are mounting at
that moment. Grock four ingested one point eight million rows
of high frequency, constantly changing data.
Speaker 1 (13:04):
One point eight million rows in real time. What kind
of data are we talking about? Not just hotel booking,
surely no way beyond that.
Speaker 2 (13:11):
It was pulling live flight status feeds directly from the FAA,
hyperlocal ride share demand patterns, seeing where people were suddenly
requesting cars indicating immediate need for shelter. Local train schedules
were there running delayed, Plus incredibly granular aciweather microgrid.
Speaker 1 (13:27):
Data microgrid weather data, how detail.
Speaker 2 (13:30):
Showing which specific neighborhoods were losing power or facing immediate flooding.
It wasn't just storm in Boston, it was severe impact here.
Now the AI synthesized all of this to understand not
just that demand existed, but the nature of the demand,
acute immediate need for safe shelter in specific locations.
Speaker 1 (13:47):
Okay, that's an incredible level of real time situational awareness.
How fast could the system turn that insight into actual
pricing decisions?
Speaker 2 (13:55):
Two minutes?
Speaker 1 (13:55):
Two minutes? Yes, By six point one ten am. Just
two minutes after ingesting that data, the system had completely
rewritten every single best available rate that back car levels
for every room type for the next ninety six hours.
Speaker 2 (14:07):
Wow. Was it just a blanket price increase. No. Far
smarter than that. It was surgical recognizing the immediate desperate
need for basic shelter. It dropped the rates for standard
rooms on the shoulder nights tonight and tomorrow night by
twenty nine percent, aggressively capturing that distressed traveler market. Flooding
in from the.
Speaker 1 (14:26):
Airport get captured volume first, But at.
Speaker 2 (14:28):
The same time it recognized that guests who typically book
premium suites. They'd prioritize safety, comfort, and luxury, and price
would be less of an object in this scenario. So
for the upcoming weekend, premium suites ejected the rates up
by a massive forty.
Speaker 1 (14:43):
Forty one percent increase on suites while dropping standard rates.
That's that's playing the elasticity perfectly based on the specific situation,
true dynamic pricing.
Speaker 2 (14:51):
Exactly exploiting the perceived value of safety and immediate availability
for different segments.
Speaker 1 (14:56):
So what was the bottom line?
Speaker 2 (14:57):
Did it work amazingly well? Nine point zero am that morning,
just one hundred and sixty eight minutes after the AI
kicked into gear, the hotel was one hundred percent sold out,
completely full, and their average daily rate the ADR it
ended up being forty two points above the average for
their direct competitors in the str report for that period, forty.
Speaker 1 (15:16):
Two points above the comp set while being completely sold
out during a chaotic event. That translates to serious money.
Speaker 2 (15:21):
Incremental revenue directly attributable to this AI driven strategy. During
that seventy two hour window, four hundred and eighty seven
thousand dollars.
Speaker 1 (15:30):
Nearly half a million dollars pulled out of thin Air
essentially from what could have easily been a disaster with mass.
Speaker 2 (15:36):
Cancelations, exactly salvaged from a potential cancelation hole and turned
into a massive win.
Speaker 1 (15:41):
Okay, I have to ask the compute cost running Rock
for processing one point eight million data rows, rewriting all
those rates, What did that cost the hotel?
Speaker 2 (15:51):
This is the punchline verifiable cloud compute costs for that
entire operation, eleven dollars and forty.
Speaker 1 (15:57):
Cent eleven dollars and forty cents to make nearly half
a million.
Speaker 2 (16:00):
Yep, that cost to profit ratio. It just demonstrates the
monumental shift. It's not a human guessing anymore. It's an
algorithm seeing demand signals your competitors literally cannot see yet,
and acting instantly.
Speaker 1 (16:11):
Incredible. Okay, let's shift gears again from revenue generation to
arguably the biggest constraint facing the industry people staffing. The
ahla's twenty twenty five report, it confirms sixty eight percent
of operators cite staffing shortages as their number one headache.
Speaker 2 (16:28):
Yeah, it's a persistent, major challenge.
Speaker 1 (16:30):
So Part three of our dive looks at how AI
addresses this, not just by replacing people, but by moving
past the idea of simply finding bodies to fill.
Speaker 2 (16:39):
Roles exactly, the strategy shift is towards multiplying minds, using
AI to augment the capabilities of the existing team, making
them more effective, more productive, and maybe even making their
jobs better.
Speaker 1 (16:50):
Okay, multiply minds. Let's start with something traditionally very human
driven up selling, especially hyper personalized up selling. How does
AI change that game? Our sources talk about Connie two
point zero, right.
Speaker 2 (17:02):
Connie two point zero is Hilton's next gen AI concierge,
and the sophistication level here is really key. It's not
just a chatbot. It uses eleven Labs for the voice generation.
Speaker 1 (17:11):
Eleven Labs they do those hyperrealistic AI voices.
Speaker 2 (17:14):
Right exactly, makes the interraction feel much more natural, builds trust.
Then it uses the Lama three point two nine zero
B model, which includes a vision component.
Speaker 1 (17:22):
He communicates via in room tablets. Though why does the
vision part matter?
Speaker 2 (17:27):
Good question? The vision allows Connie to understand the context
within the room, even if it's just communicating via the
tablet screen. Is the TV on? Are the lights dimmed?
Has the guest repeatedly swiped away notifications? It uses these
visual cues to tailor the timing and style of its approach.
It's not just about language, it's about context.
Speaker 1 (17:45):
Okay, that's clever, and it speaks multiple.
Speaker 2 (17:47):
Languages fluently in fourteen languages, which vastly expands its reach
and effectiveness.
Speaker 1 (17:53):
So the sources actually put this to the test human
versus AI.
Speaker 2 (17:56):
Yep a direct ad test over a fourteen day period
cross twelve hundred rooms. They tracked conversion rates specifically for
late checkout up cells human concierge attempts versus Connie two.
Speaker 1 (18:07):
Point erh attempts and the results. What was the gap?
Speaker 2 (18:09):
It was significant. The human concierges, who were doing a
perfectly fine job by traditional standards, achieved a four point
one percent conversion rate on those late checkout offers.
Speaker 1 (18:18):
Okay, four point one percent seems reasonable. And Connie two point.
Speaker 2 (18:22):
Zero twenty eight point seven percent conversion rate whoa.
Speaker 1 (18:24):
Four point one percent versus twenty eight point seven percent.
That's almost seven times.
Speaker 2 (18:29):
Higher, nearly seven times yes, and it gets better. On
top of that massive lift in the primary upsell, Connie
two point zero also managed an additional two point four
percent conversion on food and beverage cross cells. How did
it do that seamlessly? It would confirm the late checkout
and then add something like, since you'll be enjoying the
room a bit longer, may I assist you with a
Brench reservation downstairs, natural relevant timely.
Speaker 1 (18:51):
Seven times higher conversion. It almost suggests the machine's consistency,
it's perfect memory, its lack of distraction. Maybe that outweighs
the human type for this kind of specific transactional up cell.
Speaker 2 (19:02):
That's certainly what the data implies. The AI is always on,
always polite, never having a bad day, and has instant
access to the guest's entire profile and preferences. The annualized
revenue lift from just this one upsell initiative across only
twenty two properties testing Connie two point zero two point
nine million dollars two.
Speaker 1 (19:17):
Point nine million dollars. Okay. That transforms a task that
was maybe a marginal benefit into a serious, scalable profit center.
Speaker 2 (19:23):
Precisely, it takes a necessary but often low yield human
interaction and makes it highly profitable and digitally scalable.
Speaker 1 (19:31):
All right, let's talk housekeeping. This is a huge labor department,
physically demanding work hard to optimize schedules. How does AI
help here? We're talking robots predicting dirt.
Speaker 2 (19:43):
Kind of Yeah. It's the combination of SoftBank's Whiz cleaning robot.
You might have seen these autonomous vacuum cleaners paired with
brain cores AI software, which acts as the brain for
the robot and the overall cleaning operation.
Speaker 1 (19:55):
So the AI isn't just driving the robot.
Speaker 2 (19:57):
No, it's much smarter than that. The AI's core. Our
task here is to optimize the entire cleaning schedule and
route plan by predicting the soil load of each room
before the human housekeeping team even goes near it.
Speaker 1 (20:08):
Hold on predict soil load? How on Earth does an
AI predict how dirty room is going to be? That
sounds like sci fi.
Speaker 2 (20:15):
It uses a few data streams. First, historical data from
the wizrobox themselves, surface reflectance data from previous cleans. It
learns which rooms tend to get dirtier faster based on
where and tear patterns.
Speaker 1 (20:26):
Okay, historical patterns.
Speaker 2 (20:28):
Makes sense, But here's the really fascinating bit. The source
is highlighted. It also analyzes aggregated anonymized guest activity data,
including believe it or not social media tags associated with
the hotel or area.
Speaker 1 (20:41):
It's looking at guest Instagram tags.
Speaker 2 (20:42):
Seriously anonymized and aggregated. Yes, it knows that guests tagging
things like hashtag party weekend, hashtag family reunion, maybe even
hashtag pet friendly stay are statistically far more likely to
generate a high soil load than say a solo business
traveler tagging hashtag conference life.
Speaker 1 (20:59):
Wow. So it's using external behavior signals to predict internal
cleaning needs exactly.
Speaker 2 (21:05):
This allows the system to dynamically prioritize housekeeping labor, send
more resources to the rooms predicted to be dirtiest, maybe
a just cleaning frequency for others.
Speaker 1 (21:14):
Okay, what was the actual impact at a large property.
Speaker 2 (21:16):
They tested this at a big eleven hundred room casino
hotel in Las Vegas. The result they were able to
cut the number of turndown services that evening refreshed by
nineteen percent just by optimizing based on predicted need.
Speaker 1 (21:27):
Nineteen percent fewer turndowns. That must free up a lot
of staff time.
Speaker 2 (21:31):
It freed up the equivalent of fourteen full time employees
fourteen fifties. But here's the crucial part. Guess satisfaction scores
actually went up by six points.
Speaker 1 (21:41):
Satisfaction went up even though they cut service.
Speaker 2 (21:43):
How because the AI wasn't just cutting randomly, it was
redirecting that freed up human labor. Staff spent more time
providing detailed, high quality cleaning and service to the high
touch areas, the premium suites, the rooms that really needed
that human attention, instead of wasting time on lately used rooms.
Speaker 1 (22:01):
Uh okay. So it's a perfect example of augmentation. The
machine handles the optimization and prediction, freeing up humans for
the high value, high touched tasks that guests really notice.
Speaker 2 (22:12):
Precisely, the robot vacuums the standard floor. The human adds
the fold and towel.
Speaker 1 (22:16):
Elephant moving from the physical room to the guest's memory,
or rather the hotel's memory of the guests. This idea
of the guest as chief memory officer. The sources mentioned
as Skift survey what was that finding?
Speaker 2 (22:27):
Yeah, A twenty twenty five Skift survey found that sixty
percent six zero percent of millennials said they would abandon
booking with a hotel brand if the hotel forgot a
preference they had previously stated, like their pillow type or
room temperature.
Speaker 1 (22:41):
Sixty percent would walk away over a forgotten pillow preference. Wow,
forgetfulness is now a direct business.
Speaker 2 (22:47):
Risk, a huge one. Loyalty is incredibly fragile. So to
combat this, leading edge hotels are building what the sources
call a memory graph.
Speaker 1 (22:55):
Okay, memory graph sounds complex. Explain that for someone deep
into database tech. How is it different from just a
regular customer database.
Speaker 2 (23:04):
Think of it less like a spreadsheet listing facts about
a guest and more like a dynamic, interconnected web of
their history and preferences with the brand. It's often built
on graph database technology like Neo four J, so it
links things together exactly. It doesn't just store attributes and
maps relationships and behaviors over time. It ingests data from
literally every single touch point. The preferences you tick on
(23:26):
a pre stay email form that goes in the command
you gave to the Alexa device in your room last
time in the system, reading your license plate as you
arrive in, even the tap from your RFID wristband ordering
a drink at the pool bar that gets added to
the graph too.
Speaker 1 (23:41):
So it's building this incredibly rich, multidimensional picture of every
guest interaction and Grock four is involved here too.
Speaker 2 (23:48):
Yeah. GROCK four acts as the continuous processing engine. It
constantly analyzes all this incoming data and updates what the
sources call a seven dimension preference vector for each guest.
Seven dimensions what things like their preferred climate settings, sleep
patterns do not disturb times, dietary needs or allergies noted previously,
leisure interests shown through activity bookings, even privacy sensitivity. GROCK
(24:12):
refines this vector constantly, so the system doesn't just store
your preference for feather pillows. It understands your overall sleep profile.
Speaker 1 (24:19):
The goal then is truly seamless, almost anticipatory service. Give
us that check in example again, what actually happens?
Speaker 2 (24:26):
Right, So you walk up to the desk before you
even get your key, the system has already acted. Your
guest folio is populated, but it's also triggered actions Room
fourteen twenty two. It's automatically being set to sixty eight
degrees fahrenheit because the graph knows that's your sweet spot.
A notification has pinged housekeeping ensure hypoallergenic duvet is on
(24:46):
bed for incoming guests. Two bottles of still water are
dispatched by room service and the do not disturb on
the phone system is pre activated from ten pm to
eight am based on your past behavior.
Speaker 1 (24:57):
And the guest hasn't asked for any of this during
this specific to check in.
Speaker 2 (25:00):
Nope, it just happens based on the memory graph, it's
anticipating needs before they're even expressed.
Speaker 1 (25:05):
That level of personalization of being truly remembered must have
a huge impact on loyalty. What's the metric.
Speaker 2 (25:10):
It's dramatic. The sources looked at repeat guest net promoter
scores MPs. The chain average was around sixty seven, which
is decent kind of satisfied properties. Using this memory graph system,
their repeat guest MPs jumped to ninety one ninety one.
Speaker 1 (25:26):
That's not just satisfied. That's deep into promoter or evangelist.
Speaker 2 (25:30):
Territory exactly, and that kind of loyalty shift it drastically
cuts your customer acquisition costs down the line and maximizes
lifetime value. It's a massive competitive mote.
Speaker 1 (25:42):
Okay, Part four, Let's dig into the financial core. Maybe
less sexy than predicting dirt or cloning voices, but absolutely
critical to the bottom line. We're talking energy, fraud and
attracting capital let's start with energy utilities. The sources call
the silent thirty two percent of op X.
Speaker 2 (25:59):
Yeah, spence line that never sleeps.
Speaker 1 (26:01):
It's relentless and notoriously hard to control, right because it's
driven by guest behavior, which is all over the place exactly.
Speaker 2 (26:07):
Traditional HVAC systems and hotels are often pretty dumb. They
rely on basic motion sensors, passive infrared or PR. They
only know if someone is in the room right now
or was recently. Lots of wasted heating and cooling and
empty rooms.
Speaker 1 (26:20):
So what's the AI powered upgrade? The TEXTAC mentioned is
Verdant's ZX thermostat paired with AWS in front of two chips.
What's special there.
Speaker 2 (26:27):
It's about moving from simple detection to sophisticated prediction. These
new sensors are hyper sensitive. They're sampling data, temperature, humidity, motion,
even tiny changes in magnetic fields from balcony doors opening
or closing two hundred.
Speaker 1 (26:41):
Times per second, two hundred times a second.
Speaker 2 (26:43):
Okay, that fire hose of data is processed right there
on the device. That's the edge machine learning part. Using
those AWS inferentiate chips, it's not just looking for simple motion,
it's analyzing patterns in that high frequency data stream to
predict occupancy and guests intent. And it can do this
on average twenty three minutes earlier than a traditional PIR
(27:04):
sensor figures out the room is truly.
Speaker 1 (27:06):
Vacant twenty three minutes earlier. Why does that twenty three
minute hit start matter so much for energy savings?
Speaker 2 (27:11):
It allows the HVAC system to start coasting much sooner.
If the AI detects subtle patterns, maybe the guest is
packing based on movement, or the balcony door sensor shows
repeated short opening suggesting they're about to leave. It begins
to slowly relax the temperature controls. It lets the room
drift slightly towards the unoccupied set point before the guest
actually walks out the door, and the old PIR sensor
(27:32):
finally notices.
Speaker 1 (27:34):
I see it's squeezing out savings in that twenty three
minute window. Room by room, day by day predictive climate
control based on behavioral micro signals. What's the financial impact
when you scale that up?
Speaker 2 (27:44):
The sources looked at a large real estate investment trust
a rate with one hundred and eighty properties using this system.
On average, it shaved one point eight killowat hours kill
what hours per occupied room night?
Speaker 1 (27:56):
One point eight killowaters per room night doesn't sound like
a lot on.
Speaker 2 (27:59):
Its own, but multiply that across one hundred and eighty hotels,
thousands of rooms, three hundred and sixty five days a year.
It added up to eleven point four million dollars in
annual utility savings for the REAP.
Speaker 1 (28:10):
Eleven point four million dollars.
Speaker 2 (28:11):
Wow, And for the ESG folks that also meant an
eighty two hundred metric ton reduction in their carbon footprint.
Huge co benefit.
Speaker 1 (28:20):
And the kicker for the finance team the payback.
Speaker 2 (28:22):
Kayback period on the capital investment for the Verdana WS
system just nine point two months less than the.
Speaker 1 (28:28):
Year to get their money back, and then pure savings afterwards.
That makes it almost a financial noe brainer.
Speaker 2 (28:32):
Doesn't Absolutely it shifts from a nice to have green
initiative to a core financial imperative.
Speaker 1 (28:36):
Okay, let's talk payments fraud compliance Another area where regulations
are getting tougher. PCI DSS four point zero point one
was mentioned, demanding continuous behavioral compliance. What does that mean?
In practice?
Speaker 2 (28:51):
It means the old way of just tokenizing car data
isn't enough anymore. Regulators now expect merchants, including hotels to
continuously monitor the bhaavior around transactions, especially for card not
present payments, to detect anomalies and potential risks in real time.
It's a much higher bar.
Speaker 1 (29:08):
So how are hotels meeting this? What's the AI stack
look like? Here?
Speaker 2 (29:11):
It's often a combination basis theory vaults for the secure
tokenization and handling of the rock hard data. Then anthropics
Claude constitutional AI again remember good for rule adherence and classification,
and data dog row user monitoring to capture the behavioral
signals during the online booking process.
Speaker 1 (29:27):
Okay, so Claude is analyzing the transaction details and the
user behavior and then what scoring it?
Speaker 2 (29:33):
Exactly? Every card NOTT present transaction gets a risk score
save from zero to one hundred based on hundreds of factors.
IPHS reputation velocity, checks device fingerprint, how the user interacted
with the payment.
Speaker 1 (29:44):
Form, and if the score is too low, if it
looks dodgy.
Speaker 2 (29:47):
Transaction scoring below a certain threshold. The example given was
ninety two are automatically quarantined. They don't get processed immediately.
They're flagged for human review.
Speaker 1 (29:57):
How do they speed up that human review is a
list of block transactions?
Speaker 2 (30:02):
No, much smarter. The system automatically generates a short like
nine second loom style video recording of the anomalous user session.
It shows the reviewer exactly what looked suspicious. Maybe rapid
changes and address input, attempts to use multiple cards, connections
from known risky networks. Gives them instant context to make
a decision.
Speaker 1 (30:22):
That's efficient. So what was the big surprise finding when
they implemented this? Was it catching lots of external hackers?
Speaker 2 (30:28):
Actually? No, The biggest finding, the one that generated significant
unexpected revenue recovery, was an external fraud.
Speaker 1 (30:35):
Wait was it? Then?
Speaker 2 (30:36):
It was identifying what the industry politely calls friendly.
Speaker 1 (30:40):
Fraud, friendly fraud like what?
Speaker 2 (30:42):
In this case, the system identified that about zero point
seven percent of all reservations were being made by corporate
travel bookers using their personal credit cards, often to romly
claim loyalty points or benefits, or maybe booking non compliant
rooms or rates. Effectively, it was internal policy violation and
misuse bleeding money.
Speaker 1 (31:01):
Wow. And the system caught this pattern, yes.
Speaker 2 (31:04):
By analyzing booking patterns, card types, loyalty linkages, and corporate
account rules. The recoveries from clamping down on just this
specific type of friendly fraud totaled nine hundred and forty
thousand dollars in the first year for the hotel groups started.
Speaker 1 (31:17):
Nearly a million dollars recovered just from applying behavioral AI
to the payment stream to catch internal misuse. That's pure
cash flow straight to the bottom line.
Speaker 2 (31:26):
Absolutely unexpected, but highly valuable.
Speaker 1 (31:28):
Okay, So this level of operational tightening, compliance hardening, fraud reduction,
energy saving, it has to change how investors look at
these assets, right. The sources made a really strong point
about an AI premium in capital markets.
Speaker 2 (31:41):
It's huge investor demand is now actively dictating the pace
of AI adoption and hospitality. We're talking about forty one
billion dollars in private equity dry powder earmarked for the sector.
But the limited partners, the LPs who provide that money,
they are now explicitly asking for AI readiness scores on
potential hotel acquisition before they'll even approve a term sheet.
Speaker 1 (32:02):
So if you're running an older hotel portfolio without these
AI systems, you're basically seen as riskier, we less attractive significantly,
So you're carrying operational inefficiency, compliance risk, potentially higher energy costs,
things that AI demonstrably fixes.
Speaker 2 (32:17):
Is there a quantifiable gap invaluation?
Speaker 1 (32:20):
Yes, and it's substantial. The sources indicate that sophisticated buyers
like PE firms are currently paying around eight point two
times EBITA for hotel assets they deem AI native or
properly AI integrated eight.
Speaker 2 (32:32):
Point two x okay, and for legacy portfolios without.
Speaker 1 (32:34):
This tech, the multiple drops to around six point one
x ebitda.
Speaker 2 (32:38):
Wow, that's a two point one turn difference in valuation
multiple just based on AI integration. That premium is very real,
very real, and likely growing because the returns from implementing
AI are so fast and so predictable. As we've seen,
investors see it as de risking the asset while simultaneously
unlocking significant upside.
Speaker 1 (32:56):
We really need to dig into those rapid returns. The
sources laid out this almost like an investor's ninety day playbook,
a hyper accelerated plan designed specifically to boost that AI
readiness score and capture that premium. Walk us through it.
Speaker 2 (33:10):
Yeah, it's designed for speed and impact, focused entirely on
the highest ROI fastest payback initiatives okay.
Speaker 1 (33:17):
First steps weeks one and two.
Speaker 2 (33:18):
Deploy an AI powered expense crawler think tools like RAMP,
maybe brex integrated with something like GPT four to ROH.
It automatically scans the general ledger, categorizes spending, benchmarks it
against industry data, and instantly flags outliers, potential savings or
non compliant spending.
Speaker 1 (33:36):
Just automated expense review. What kind of savings does that
typically find?
Speaker 2 (33:39):
On average? That step alone identifies around four point two
percent in immediate systemic savings across various expense lines quick
win number one.
Speaker 1 (33:47):
Okay, four point two percent savings in two weeks. What's next?
Weeks three through six?
Speaker 2 (33:52):
Attack the operational friction points we discussed earlier. Automate the
night audit process using tools like museflow or similar integrations.
And automate accounts payable invoicing using platforms like userbill dot com.
Speaker 1 (34:05):
The back office stuff. What's the ROI on that module?
Speaker 2 (34:07):
Based on the labor savings and accuracy gains we talked about,
this typically yields around a three hundred and forty percent
ROI within the first year.
Speaker 1 (34:14):
Massive payback Wow, okay, So expense crawl, then back office automation.
What about revenue? When does that come in?
Speaker 2 (34:20):
That's the focus for weeks seven through twelve. The playbook
calls for spinning up a revenue war room slack bot.
Speaker 1 (34:27):
A slack bot for revenue. How does that work?
Speaker 2 (34:29):
It uses AI to constantly monitor market signals, competitor rate changes,
flight loads, event calendars, even sentiment analysis from reviews. It
live scores the hotel's own pricing decisions, and pushes proactive
recommendations and alerts directly to the revenue management team via
slack comp set just drop weekend rates, recommend holding firm,
or sudden spike in searches for dates x toy consider
(34:51):
raising bar by twenty dollars. It creates this constant feedback.
Speaker 1 (34:55):
Loop, so continuous AI powered decision support for the revenue team.
What's the typical profit lift from that?
Speaker 2 (35:01):
Generally looking at it, eight percent to eleven percent lift
in GOP gross operating profit.
Speaker 1 (35:06):
Okay, so let's tell this up, this whole ninety day
three phase playbook. What's the estimated cost to implement these
initial systems.
Speaker 2 (35:14):
The sources cite a total cost of around thirty eight
thousand dollars for the software. An initial setup for these
specific high impact.
Speaker 1 (35:21):
Plays thirty eight thousand dollars and the typical financial impact
in the first year.
Speaker 2 (35:25):
The typical first year EBITDA lift generated by this thirty
eight K investment is estimated at one point one million dollars.
Speaker 1 (35:32):
One point one million dollars from a thirty eight K spend.
Speaker 2 (35:35):
Which works out to a what is that roughly a
twenty nine hundred percent return on investment in your one.
Speaker 1 (35:41):
Twenty nine hundred percent ROI If you're an investor or
an owner looking to sell or refinance, that kind of
math is just undeniable. It makes AI integration absolutely critical.
Speaker 2 (35:50):
It becomes the single most important factor in maximizing asset
value in today's market period.
Speaker 1 (35:56):
All Right, our final section, and it's a crucial one.
The human element, the creative side, because there's obviously a
lot of fear around AI taking jobs, but the data,
like that McKenzie report from twenty twenty five, seems to
paint a different picture.
Speaker 2 (36:10):
It really does. The data is pretty consistent on this.
McKinsey predicted that about fifty four percent of tasks within
hotel jobs could be augmented by AI, meaning AI helps
the human do the job better or faster, but only
around twelve percent of current hotel jobs were projected to
be fully automated away in the medium term.
Speaker 1 (36:28):
So augmentation, not replacement, is the main story.
Speaker 2 (36:30):
Overwhelmingly, Yes, the real focus for smart organizations isn't just
cutting heads. It's about retaining their best talent, upscaling them,
and fundamentally changing their roles to work with the AI,
giving them better tools to do more interesting work.
Speaker 1 (36:43):
Marriott's AI Solmelia program sounds like a prime example of this.
Tell us about that, what does an AI so milier learn.
Speaker 2 (36:49):
It's a really interesting initiative. It's a dedicated four week
training program designed for high performing frontline staff, concierges, front
desk agents, supervisors. They're not learning hospitality basics, they're learning
how to manage and leverage the hotel's AI systems.
Speaker 1 (37:07):
Often's the curriculum things like.
Speaker 2 (37:09):
Prompt engineering basically how to talk to the large language
models effectively to get the best results, bias auditing learning
how to spot and mitigate potential biases in the algorithms
to ensure fairness, and even robot chaperoning. The practical skills
needed to manage and troubleshoot physical robots like those whiz
cleaners we discussed.
Speaker 1 (37:28):
So they're training staff to become supervisors of the AI
in robotics exactly.
Speaker 2 (37:32):
They're teaching them to own the technology, not be replaced
by it, turning them into masters of the new operational infrastructure.
Speaker 1 (37:38):
That's a powerful shift in mindset. What were the results?
Did it help with retention?
Speaker 2 (37:43):
The results were profound according to the sources, and amazing.
Ninety two percent of the employees who graduated from the
Aismeliar program were promoted within nine months.
Speaker 1 (37:52):
Ninety two percent promoted.
Speaker 2 (37:53):
Wow, and maybe even more importantly, overall, staff attrition turnover
in the properties participating in the program fell by thirty one.
Speaker 1 (38:01):
Percent, down by nearly a third. Yeah.
Speaker 2 (38:04):
When employees feel empowered by technology, when they see it
creating new career paths for them, when they feel like
they own the algorithm instead of being victims of it,
their loyalty to the company's skyrockets, that's huge.
Speaker 1 (38:16):
Yeah. Okay. Beyond optimizing standard operations, AI seems uniquely good
at spotting and exploiting these weird, unexpected micro demand opportunities
things humans might miss entirely, these edge cases that become
new best practices absolutely.
Speaker 2 (38:30):
The speed of AI allows for incredibly opportunistic revenue capture.
The source material had a great example, the K pop pop.
Speaker 1 (38:37):
Up K Pop pop Up tell Me More.
Speaker 2 (38:39):
Okay, so a major K pop ban announces a surprise
concert in Atlanta just seventy two hours notice, Chaos ensues
among fans. The Omnia Hotel at CNN Center looks at
its inventory. They only have forty one rooms left for
those dates, a tiny number.
Speaker 1 (38:53):
So normally they just sell those out quickly at a
decent rate. Maybe maybe.
Speaker 2 (38:58):
But their AI cluster this time, a combo of Duato
for pricing, grock four for the real time context analysis,
and crucially, an API feed from ticket Master spotted something else.
It detected an immediate spike of one hundred and eighty
thousand searches on secondary markets just for tickets to that
specific consuit in the Atlanta area.
Speaker 1 (39:16):
One hundred and eighty thousand searches. Okay, that signal's intense
niche demand exactly.
Speaker 2 (39:22):
The AI understood the specific rabid fan base being activated,
so it didn't just raise rates slightly. It automatically created
and priced a unique package, the the IPK Pop pre
Show Experience or something similar bundling the room with some
themed amenities, and it priced this package at a jaw
dropping three hundred and eighty percent above their normal average
daily rate for those room types plus three.
Speaker 1 (39:43):
Hundred and eighty percent eight R Did anyone actually buy it?
Speaker 2 (39:46):
Those forty one rooms sold out in eleven minutes. Eleven minutes,
generating one hundred and eighty four thousand dollars in gross
revenue from just forty one rooms, purely because the AI
spotted the micro demand signal and priced it hyper aggressively instantly,
that kind of.
Speaker 1 (40:00):
Speed, integrating ticketing data, creating a package repricing selling out
in minutes. A human revenue team, it couldn't do that.
Speaker 2 (40:06):
Not a chance. It perfectly exploits that short term volatility.
Speaker 1 (40:09):
Okay. Another example was about capacity on demand, the insurance
adjuster swarm.
Speaker 2 (40:14):
Right, different scenario, same principle of AI speed and scale.
Hurricane Leela hits Mobile, Alabama, major damage. Suddenly about fourteen
hundred insurance claims. Adjusters descend on the city, all needing rooms,
often for extended stays.
Speaker 1 (40:29):
Huge sudden influx of very specific demand. How did one
hotel capitalize.
Speaker 2 (40:35):
A local hampton and realized they needed to reach these
adjusters immediately and personally before competitors lock them up. But
calling fourteen hundred people is impossible. Sending a generic email
blast gets lost.
Speaker 1 (40:46):
So what did they do?
Speaker 2 (40:47):
They got creative with AI. They used eleven Labs the
voice cloning tech again to create a message using the
actual voice of their general manager, a familiar, trusted local voice.
Speaker 1 (40:57):
Okay, clone the GM's voice. Yeah, then what?
Speaker 2 (40:59):
They then use an AI platform to spin up forty
thousand personalized voicemails using that cloned voice. Each voicemail perhaps
slightly tweaked with the adjuster's name or company if they
had that data from registries. These voicemails offered special extended
stay rates maybe day rates for using the lobby to work,
options tailored specifically to what adjusters need during a catastrophe resports,
(41:21):
and they mass delivered these voicemails instantly.
Speaker 1 (41:24):
Forty thousand personalized voicemails delivered.
Speaker 2 (41:26):
Instantly captured sixty two percent of that incoming adjuster market
share almost immediately, because the outreach felt personal, timely, and
directly addressed their specific needs using a trusted voice at
a scale no human team could ever achieve.
Speaker 1 (41:40):
Again, it's using AI not just for internal efficiency, but
as a tool for creative, scalable outreach and revenue generation.
These examples are powerful so for our listener who's maybe
feeling a bit overwhelmed but also inspired, thinking, Okay, I
need to get started. What's the practical roadmap? Where do
you begin? On Monday morning?
Speaker 2 (41:58):
The sources actually lay out four pretty concrete first steps,
things that separate the hotels actually doing this from the
ones just talking about it. Step one, audit your data
mode before you even think about AI tools. Look at
your own data. Export the last ninety days of data
from your core systems, your PMS, your point of sale pos,
your central reservation system crs. Really examine it. How complete
(42:22):
is it? How clean is it? Are there gaps, missing fields, inconsistencies?
Score it. If your data completeness and quality score is
below about eighty five percent, the sources say, stop, fix
your data first. AI runs on data garbage in, garbage out,
but with AI, it's exponential.
Speaker 1 (42:37):
Garbage app foundational. Get your data house in order first. Okay.
Speaker 2 (42:41):
Step two pick one bleeding wound, don't try to boil
the ocean, and implement AI across the entire hotel at once.
Focus find the one area that's causing the most financial
pain right now. Is it those night audit labor overages?
Is it spiking utility costs you can't explain? Is it
terrible upsell conversion rates? Pick the single biggest, clearest pain
point with a measure of financial impact.
Speaker 1 (43:01):
Focus for impact makes sense. Step three sounds counterintuitive for
big companies right.
Speaker 2 (43:05):
Step three is buy, don't build. Resist the urge to
create your own bespoke AI platform from scratch. The vast
majority of the sources, say eighty four percent of hotels
successfully implementing AI are using established off the shelf solutions
from vendors, maybe with some fine tuning or integration help.
Speaker 1 (43:25):
Why buy instead of build?
Speaker 2 (43:26):
Speed to value? Buying a proven solution gets you results
much faster. The average time to value for these off
the shelf AI tools is around forty one days. Building
your own takes years, huge investment, high risk of failure.
In this market, speed matters. Capture that AI premium.
Speaker 1 (43:42):
Now, okay, buy, don't build for speed, and the final
step step.
Speaker 2 (43:46):
Four ring fence the first win. Once you've picked your
bleeding wound step two and implemented a solution step three,
you absolutely must track its impact obsessively and communicate the success.
Create a simple dashboard could be Google Data Studio, could
be power Yi. It doesn't have to be fancy showing
the daily or weekly KPIs for that one specific initiative,
track the ROI and share that dashboard relentlessly with department heads,
(44:08):
maybe even all staff.
Speaker 1 (44:09):
Why is that communication so important?
Speaker 2 (44:11):
Because momentum is oxygen for change. That first undeniable, measurable
win builds credibility, It generates excitement, its cures buy in
for the next AI project. You need to celebrate and
broadcast that initial success to fuel the ongoing transformation.
Speaker 1 (44:28):
This has been well quite the deep dive. It feels
like we've fundamentally rewritten the operating manual for hospitality in
the last half hour to quickly recap the core findings.
The industry's financial foundations are definitely under pressure from costs.
But AI isn't just a potential solution. It's an immediate, proven,
and highly measurable escape route.
Speaker 2 (44:47):
Yeah, the evidence is compelling.
Speaker 1 (44:49):
Now we're seeing these integrated AI systems delivering what was
it seven percent to fourteen percent swings in EBITDA, and
not over years, but within quarters. The hotels that are
truly winning in twenty twenty five and beyond. They're not
necessarily the ones running cheaper, They're the ones running profoundly smarter.
Speaker 2 (45:04):
And this internal transformation within hotels it's about to get
a major boost from the outside, a macro tailwind. Remember
the Federal Reserve their November twenty twenty five meeting minutes
strongly signaled that interest rate cuts are likely coming, probably
by early twenty twenty six.
Speaker 1 (45:19):
Okay, cheaper money is coming.
Speaker 2 (45:21):
Exactly, So combine cheaper debt with this AI amplified cash flow.
Think back to that incredible twenty nine hundred percent ROI
from the ninety day Investor Playbook, and what do you get?
Speaker 1 (45:32):
A feeding frenzy, massive acquisition fever.
Speaker 2 (45:36):
You got it. The valuation premium for these AI native
high performing assets is only going to increase. It creates
this really urgent window right now for operators to get
their house in order, implement these systems and capture that value,
whether they plan a hold or sell.
Speaker 1 (45:50):
A real gold rush for operational efficiency and intelligence. Okay,
here's where my mind goes though, and maybe something for
you the listener, to really chew on as we wrap up,
I can predict with high accuracy that a room will
be empty twenty three minutes from now, allowing the HVAC
to power down preemptively, and if AI can access that
seven emission preference dector we talked about and automatically configure
(46:12):
your room perfectly. Temperature pillows, water do not disturb before
you even swipe your key.
Speaker 2 (46:17):
What becomes obsolete? What old ways of doing things.
Speaker 1 (46:20):
Just disappear exactly? What traditional way do we even measure
customer satisfaction? When this system anticipates needs so perfectly? Does
a paper comment card left on the desk have any
meeting anymore? Does even a standard post day email survey
asking how is your stay? Capture anything useful? When the
goal shifts from reactive problem solving to proactive perfect anticipation, It's.
Speaker 2 (46:44):
A fundamental shift. The source material had a provocative line.
Profit is the new loyalty. Maybe loyalty isn't about points anymore.
It's about the seamless, effortless, perfectly remembered and anticipated experience
that AI enables.
Speaker 1 (46:57):
So the final thought for you, what happens to the
deaf nition of good service when every single manual click,
every operational data point, every guest interaction instantly becomes part
of a learning loop for a machine that never forgets
and constantly optimizes, What does performance measurement look like when
the ultimate goal is that the customer never actually has
to ask for anything because it's already been anticipated and delivered.
That's the future AI is building in hospitality right now.
Popular Podcasts
Stuff You Should Know
If you've ever wanted to know about champagne, satanism, the Stonewall Uprising, chaos theory, LSD, El Nino, true crime and Rosa Parks, then look no further. Josh and Chuck have you covered.
Las Culturistas with Matt Rogers and Bowen Yang
Ding dong! Join your culture consultants, Matt Rogers and Bowen Yang, on an unforgettable journey into the beating heart of CULTURE. Alongside sizzling special guests, they GET INTO the hottest pop-culture moments of the day and the formative cultural experiences that turned them into Culturistas. Produced by the Big Money Players Network and iHeartRadio.
Dateline NBC
Current and classic episodes, featuring compelling true-crime mysteries, powerful documentaries and in-depth investigations. Follow now to get the latest episodes of Dateline NBC completely free, or subscribe to Dateline Premium for ad-free listening and exclusive bonus content: DatelinePremium.com