Peter Swimm of Toilville Makes AI Actually Work for Businesses

Peter Swimm has seen enough AI hype cycles to know where the bodies are buried. Long before ChatGPT turned “AI” into a household buzzword, he was building chatbots, automating enterprise systems, and watching Fortune 500 companies light piles of money on fire chasing shiny tech that promised miracles and delivered chaos. 

In this 70th episode of Bald Ambition, Mookie Spitz sits down with the founder of Toilville to cut through the hysteria surrounding artificial intelligence, workplace automation, and the endless parade of SaaS tools that have turned modern business operations into a spaghetti bowl of redundancy, meetings, subscriptions, and digital busywork.

What follows is part AI debate, part business therapy session, and part philosophical cage match over the future of work itself. Peter argues that most companies approach AI backwards: they start with magical thinking about super-bots replacing humans instead of first understanding how their businesses actually function. He breaks down why so many AI pilots fail, why organizations drown in “technical debt,” and why businesses often spend fortunes automating processes that were already dysfunctional in the first place. Instead of selling miracle cures, Peter advocates ruthless discovery, operational simplification, and using AI only where it genuinely creates leverage.

Mookie pushes the conversation into deeper territory: the psychology of AI hype, corporate layoffs disguised as “innovation,” why employees fear change management, and the growing tension between giant cloud-based frontier models and smaller local AI systems running privately on personal hardware. Peter explains why he believes the future belongs to leaner, more customized AI ecosystems rather than endless dependence on trillion-dollar data-center empires. The two also spar over AGI fantasies, universal basic income, worker anxiety, SaaS bloat, and why some executives seem more interested in replacing humans than helping them work better.

Along the way, Peter shares real-world examples from healthcare, enterprise contact centers, and small business consulting engagements where the real breakthrough wasn’t “AI magic” — it was simply eliminating pointless friction. Instead of flashy demos and investor bait, he makes the case for practical systems that reduce drudgery, preserve human connection, and stop employees from wasting half their lives copy-pasting information between five different apps that should never have existed separately in the first place.

The Guest

Peter Swimm is a conversational AI technologist, product strategist, and the founder of Toilville, a consultancy focused on helping organizations implement AI and workplace automation without losing the human expertise and institutional knowledge that actually make businesses work. 

Over the past two decades, Swimm has worked across startups and enterprise tech, including roles connected to conversational AI systems at Microsoft, Walmart, LivePerson, and Botkit, the chatbot framework later acquired by Microsoft. His work has centered on chatbot design, workflow automation, contact-center optimization, and AI governance, with a recurring emphasis on practicality over hype.

His Company

https://www.toilville.com/

Send the host a text! Let him know what you think

Support the show

Episode Transcript

SPEAKER_00 (00:03):
Hello and welcome to the Bald Ambition Podcast.
I'm still a very bald host,Mookie Spitz, and the one with
all the ambition today is Mr.
Peter Swim.
He is the founder of Toilville.
No accident, I think, in theselection of that startup name.

(00:36):
But I'll let the maestrohimself, Peter, describe
describe his baby.
You're probably the fifth guestI've had over the last few weeks
related to AI optimization inthe workplace.
I've had a guest optimizingsales forces and sales
management, the dispensing ofbeverages in the mobile

(01:00):
Starbucks-like setting,operationalizing AI for everyday
business tasks.
How are you different?
What do you offer?
Peter, we're we're we're allears.

SPEAKER_02 (01:12):
So I I'm a person who kind of fell into the AI
industry, uh, but I did it 15years ago.
And so this is probably my thirdor fourth AI revolution so far
in my career.
And I think like what a lot ofpeople uh hope to happen and
have been describing is whatwould happen in this has only
become true very recently.

(01:33):
So in my consultancy, um, I tookmy experience of working for big
companies like Walmart andMicrosoft of how AI pilots start
and fail.
And it's largely because a lotof people build towards the hype
window as opposed to thepractical floor.
So I help a lot with my clients,one kind of shaping their
expectations of technology andcreating a set of rubrics that

(01:56):
allows you to evaluate whatwould cost, you know,
irregardless of like if it's AIor not, and how you can like
find low-hanging fruit andtransform that stuff with or
without AI.

SPEAKER_00 (02:08):
So it sounds like your optimization first, with or
without AI.
And now that you've got the bigguns of AI, especially after the
chat GPT revolution, what wasit, November 2022?
The world seemed to flip almostovernight.
Tell us a little bit about someof your background in AI prior

(02:29):
to that, because I think mostfolks think that AI is a new
thing, and it's been with us forquite some time, even in
application.
Yeah, everything fromrecommendation engines on your
favorite streaming and shoppingservice to um all sorts of
back-end stuff that's been goingon, uh, without most people ever

(02:51):
really understanding it orknowing it's there.

SPEAKER_02 (02:53):
I think everything like that we think of as the
algorithm is basically what AIhas become today.
So when you you know you'relooking for carpet tiles, and
then every single ad around youis carpet tiles, and uh, all
that stuff has been in motionfor a long time, and now it's
being under the brand of AIbecause tools like ChatGPT and
other large large languagemodels has much fuzzier edges.

(03:16):
So, like you remember back inthe day, you would go to a
website and you you type in thewrong URL and it's file not
found, doesn't load.
We've now have the ability tokind of like correct in situ
with, oh, you meant this, orhere's something similar.
And when you put that inapplication, your application no
longer says, you know, yourchatbot doesn't say, Oh, I don't

(03:38):
understand that, because Iunderstand everything.
And so we've kind of solved thatunderstanding problem without
really understanding whatunderstanding is in a
technological sense.
And so my background, uh uh adecade ago, uh application
called Slack came out and theyannounced their API.
And uh the company that I wasin, botkit, was uh launched with

(04:01):
the API chatbots.
And chatbots were basically dumbversions of things you you use
on the telephone when you pressone, you get a thing and all
that.
Our chatbot ran on regex, whichis uh for people who know, regex
is basically like matchingpatterns.
So like if I say it's a name,it's gonna be capitalized and
have one or more extra letters,and that's probably a name.

(04:24):
And so our chatbot, whichprobably can run on you could
probably run an entire serviceon the power in your cell phone
in your pocket right now, um,was a state of the art in 2016.
Right?
That's the best you can do.
And the vision and the value ofwhat people are selling sounds a
lot like what ChatGPT isoffering today.

(04:45):
And the difference is that it'sfinally fast enough and close
enough that you don't need tolike have a data scientist set
it up for your company.
You can like pay for a servicewith credit card and put it down
and you kind of haveunderstanding.
But that's where the problemsrun in because something that
demos very well, like you canrun a demo and get people really

(05:07):
excited.
When you run it across a largerpopulation of people, you run
into problems.
So later on, I went to work uhautomating contact centers at
Walmart.
And you know, the contactcenters are an interesting
problem because most peopledon't want to call in to
anything.
You know, they would love tojust get the answer.
So if they're calling in, it'salready a problem.

(05:28):
You've already made a mistakesomewhere along the line where
the information wasn't findable.
And so their problem is everytime a person picks up a phone
for another person, it costsfive dollars, and that person
picks up 150 calls a day, andthere's a room of 200 of them.
And so it's money on the tablethat you see there.
And so their goal was toautomate this contact center.

(05:50):
And there are some things youcan't automate, right?
Like Mookie, you're gonna inventa problem and you're gonna be
like, Wow, I've never heard thatone before.
And a chatbot wouldn'tdefinitely know what to do with
it.
So that's when I started to kindof see the patterns of like, oh
in the five-minute phone callthat costs five dollars, you

(06:10):
spend two minutes on verifying aperson's identity, right?
You just find out who they areand what they are, and all that.
A computer could do that, butlike it's not a question about
like my personal life orwhatever.
So let's find the low-hangingfruit to do things.
And that's kind of kind of putyou down like this rabbit hole
of like, okay, if everything isa series of like l levers and

(06:31):
pistons and things you canautomate, and then things that
require a personal touch, how doyou arbitrate that kind of
working place to get people todo stuff?
And so a lot of the tooling Idid, I was working at Microsoft
when my product became co-pilotstudio, which was the thing to
design co-pilot experiences.
And so I was working directlywith customers every day of like

(06:51):
turning legacy applications intoAI applications.
And a a thought occurred to me,and my clients would, you know,
my people I work with wouldpoint it out too, is just like,
oh, we're automating somethingthat was originally meant to
automate something that wassomething real, right?
So you you don't invent acontact center because you like

(07:12):
contact centers.
You vent a contact centerbecause you have questions and
people need to answer them.
And so instead of put stickingAI in something that's already
an abstraction, what if youbring it closer to source and
make it so the person who knowsall the answers is distributed
better and and not wasting timeon what time does the office
open?
Because that's on the website.
You know, what's the phonenumber that's on the website?

(07:34):
Then it's like, oh, my airplaneticket got eaten by my dog.
You know, something weird thathas a person has to intervene.
Okay, that needs to talk to aperson.
So instead of like wasting allthe time that you're paying for
five dollars a call on answeringquestions that read the manual,
uh let's get people into moredeep conversations that

(07:56):
highlight your business and thepersonality and your culture of
your company.

SPEAKER_00 (08:01):
Well, you're breaking the need, the business
need into two distinctcategories.
The one is deterministicquestion response.
Need to check uh customers'information or background.
Uh, and even remedial machinecan do this because it's
basically if then.

(08:23):
And you mentioned that that'sthe lowest hanging fruit, and
that's stuff that you canautomate.
And there are other functionsbetter left to a real human,
which are much more complex,interwoven, and unanticipated.
But ChatGPT and their kind, allthe frontier models, have shown
us that it goes beyond justif-then, and they have very

(08:45):
sophisticated machine learning,deep learning machinations that
mimic human behavior to theextent where they could they
could take the role of anoperator, at least to a certain
extent, they can pass theso-called Turing test for quite
some time, and this is where alot of the hype comes in.

(09:08):
So, if I'm hearing youcorrectly, if you really want to
optimize your business and evenlook at it through the lens of
this AI stuff, start with thebasics, build up from there, and
don't let yourself getbamboozled by a lot of the
hysteria about plugging in a botthat's gonna solve all your

(09:29):
problems.
Am I hearing you properly?

SPEAKER_02 (09:32):
You got it right up right on.
It's just like if I paint theface on a rock, it's not a pet,
you know?
And I think a lot of people getreally uh parasocial with their
applications because this thingthat is juiced to be agreeable
and to think that you are makingstunning observations often, uh
it's like wow, everything I saythis bot, I'm a genius.

(09:54):
I'm I'm smarter than Einsteinand Lee Yakoka combined, and uh
this is gonna be the best thingsince sliced bread.
And then they go into court andthey present their case, and
everything they cited is madeup, you know.

SPEAKER_00 (10:07):
And we were talking about or even if it's not made
up, it's prioritized to blowsmoke up their butt.
It's programmed to besycophantic.
And many folks fall in love withtheir bot.
They conclude that not only isit sentient, akin to their own
consciousness, but the botactually cares and they create
an emotional reaction.
And I think it's that same kindof sentiment that I'm hearing

(10:30):
from a lot of business owners,business managers, where they
want these kind of miracles tohappen operationally in their
businesses, and they expectfingers to be snapped, magic
happens, and then all theirproblems go away.
But where I think you're you'rebasing your startup on is this

(10:50):
understanding that you need atleast a remedial understanding
of the core business issues,which are typically, if you look
at the lowest of low-hangingfruit, very deterministic and
operational.
We need to just figure some ofthe nuts and bolts stuff out.
And as we increase in ourrelationship with you, we get to
know you better and we get toplug in increasingly

(11:12):
sophisticated technology,inclusive of this AI stuff, we
can work together to do that.
But most folks do this backasswards.
They start with this delusionalnotion that these super bots
will save you immediately.
And I think what you're gettingat again is no, no, no, no, no.
Let's start with understandingwhat you're doing, what some of

(11:35):
your issues are, and let'scrawl, walk, run into more
sophisticated solutions ifthey're appropriate.
Yeah.

SPEAKER_02 (11:43):
Yeah.
And I and I think it is for alot of people, it's a seed
change.
It's not ever overnightrevolution.
Like uh a lot of my clients findout that things they've been
carrying as like technical debtand things that they're part of
their job are not part of theirjob.
It's part of being a client ofSalesforce or Slack or any other
tooling that makes you do itthat way.
And so if you had your brothersto go back in time to when you

(12:05):
started your business and youdid it from scratch, what would
it look like?
And uh for a lot of people, it'slike, oh, I can fire three of my
SaaS services because I nolonger need to copy paste an
email into a Salesforce thingthat sends a thing to a Slack
that does it this, that doesthat, this orchestration.
And so a lot of these tools thatare middleware that are designed

(12:25):
to like facilitate theorchestration between things
becomes uh relevant.
And so I you know, a lot of myclients they uh increase people
and de and using the money theysave from technology to get more
people because really they justneed five more people of them.
And if the tooling is that muchcheaper, then great, I can send

(12:47):
the virtual self with me alongon the calls and record the
notes, and uh I don't have to goto every meeting and I have a
good idea of the pulse ofthings, and uh delegation
becomes trivial.
And you don't hear a lot ofstories like that because these
companies are spending$18billion quarter training this
stuff, and you're not gonna makethat back unless you get$18 of

(13:08):
value out of the labor pool.

SPEAKER_00 (13:11):
And so go ahead and the SaaS Mageddon, yeah, which
is hit investors in Wall Street,which is software as a service,
Salesforce, even DocuSign.
Big companies, they need a lotof subscriptions, and businesses
have become increasingly relianton these to perform all these

(13:32):
functions, some of which areredundant, they're
interconnected, overlycomplicated.
And what you're recommending isif you deconstruct your
operations, in a sense, from theground up, you'll realize that
A, you might not even need allof this, and B, your your whole
setup is probably a spaghettibowl mess of redundancy,

(13:53):
interoperability challenges.
And you're recommending they goback to basics and build it back
up again in a way that makesmore sense and if necessary uses
some of the some of thetechnology that you could bring
to bear.

SPEAKER_02 (14:08):
Yeah, I mean, I've worked for companies that had
everything.
They've had, you know, they'vehad Slack and Teams and WebEx,
and and it's just like eachdivision has to like so you have
to, if you were in a positionwhere you worked with everyone,
you had six clients and fouremail applications and six video
calls, and it's silly, right?

(14:28):
It's like you should have aphone number, and the phone on
your desk rings and you pick itup and you answer it.
Like everything else is justlike unnecessary cruff.
And then you have to like gointo a meeting like, well, this
organization, our artifactdirector is PowerPoint, and this
one, the artifact director, is aticket in Jira.
And so now people are copyingpasting stuff from Jira into

(14:50):
PowerPoint, into Word documents,and all this like domain
switching and stuff is like therhythm of business, and no one
ever talks about that cost.
No one ever talks about, youknow, they you always talk about
like shadow quitting andemployee responsibility and
being in the office or not inthe office, but they never talk
about the fact that I had tohave three meetings about one

(15:11):
email.
And so we're kind of at thisopportunity now, like, okay,
what if we would just redo theentire social contract where um
all my employees workautonomously up to the point
where it's time for a decisionto be made, and I thumbs up and
thumbs down after I read theirreport.
And I think that is a muchbetter way to work.

(15:31):
Like at my company internally,we effectively uh four-day work
week because uh we budget fivedays of work, and if you find a
cool way to do it and you onlyneed to work three days, that's
congratulations, you won theweek.
You know, we I'm not like tryingto like track my burndown rates

(15:52):
and and seeing what people aredoing every hour a day.
I just like I want them to behappy and healthy and ready to
work when we need work done.
And I think this is structurallya much better way to work than
you know in China, they have the669 six days a week, uh you
know, nine to nine every day.
And it's just like if you haveto work more than 40 hours a

(16:14):
week, you don't know how to plantime, you don't know how to like
work effectively, and thesolution for technical problems
isn't more work, it's smarterwork.

SPEAKER_00 (16:26):
Yeah, and it sounds like central to that is once
again the discovery that you do.
You you analyze how businessesdo what they do before you make
any kind of recommendation.
Can you take us through how thisrelationship begins and what's
involved?
Like, let's say I'm Mookie Spitzand I have the bald ambition

(16:48):
LLC, right?
And I've got all this all thisnonsense that's going around
with processing my videos anddistributing them and responding
to customer service.
And I've got to your point, I'musing with my my 10 employees,
I'm using Slack, and I've gotemail, and I've got Teams, and I

(17:12):
just put out podcasts.
Yeah, yeah.

SPEAKER_02 (17:16):
And I think like a lot of people are like that.
So you I think a lot of thepeople I'm in competition with
for contracts, you'll get SOW,and the SOW is just like, here's
what I want.
It's like, sure, I'll build youwhat I want, but I really want
to understand why you'rebuilding what you want and why
isn't what you have today notworking for you.
Because it it's easier for me tobring the price of SOW down from

(17:40):
understanding how muchredundancy you've built into it
than from cutting my hours.
And so I rather like, you know,if I'm gonna slash 40 hours, I
want to slash 40 hours becauseit actually is 40 hours less
award versus like me likeworking nights and weekends and
helping people out and stuff.
And so that kind of goes down toI have to understand what keeps

(18:00):
you up at night.
So you're told me all the thingsyou want, so I understand why
you want it, but what willhappen to you if you don't get
it?
You know, and what happens thatis uh keeping you from realizing
your vision with these existingtooling because sometimes the
the problem started 18 stepsago.
And so now you've accumulatedall these toolings that are

(18:23):
trying to work together better,but maybe less is more, and you
can cut versus uh add to thatpile.

SPEAKER_00 (18:31):
I work in business consulting too, and it's often
seemed like a cat chasing itstail or a doctor charging for
the treatment before thediagnosis is even conducted.
So to your point, you you give astatement of work which
outlines, let's say, scope andbilling, but you don't yet know
really what you're fixing untilyou do that initial discovery

(18:55):
phase.
So many more responsible andadept consultants will issue two
SOWs.
There's an SOW for what'sconsidered the discovery phase,
where you diagnose the patient,as it were.
And the other one is once wefigure out what the hell is
wrong, then uh then take it tothat, take it to that next step.

(19:15):
So what you're saying is veryrelatable.
It seems inevitable andnecessary to really understand a
business before you can fix it.
But a lot of folks are aredoing, in a sense, what you're
doing.
So citing again some of theguests I've had before, most of
them cite the need for extensivediscovery.
Um, there's some overlap interms of some of the value prop

(19:38):
that you're offering.
What does Peter and company andToilville do which adds that
special sauce to this kind oftechnological infrastructure
optimization?

SPEAKER_02 (19:55):
Um, so I've worked with companies as small as three
or four people, and I workedwith companies that is the
largest employer on the planet.
And I think the major differencebetween all those things is risk
aversion, right?
And what do you like?
So you have your goals, and whatcan you get away with before it
comes to pay the piper?
And so a lot of the times I, youknow, you're talking about

(20:18):
discovery phase.
I make my discovery phases veryexpensive, and I do it as a uh
canard in a way to like forcepeople to think about okay, what
are my non-negotiables?
What do I really need and don'tneed?
Because the better they defineit in the SOW, the lower my
price goes, you know, because Ican I I can say, okay, I can do

(20:39):
this in a day, I could do thisin two days, and because you're
able to book me three solid dayswith all the information, I will
cut you a break versus hangingaround with you and feeling it
out and finding out, like, oh,okay, oh, we're actually
building a hovercraft, not ahelicopter, you know, because I
could do that, you know, we canhang out and vibe and go play
tennis and do whatever it takes.

(20:59):
That's my day rate, you know.
Like, I'll hang out and dothings.
But if you want to like get downto brass tacks and do stuff and
you know what you want and youhave a way to measure the before
and after already, it's very,very affordable.
And I think that kind of arelationship is much more
appreciated in smallentrepreneurs because everyone
wants to have a guy or a gal togo to when they need something,

(21:22):
and then they can go sit on theshelf.
I don't want to be like servicecontracts, I don't want to be
like coming to your Fridaymeetings every week for the rest
of my life.
You know, I want to be like, heyPeter, this guy wants to sell me
this thing.
Is it real?
Like, no.
Okay, thanks.
Bye.
That would be the perfectrelationship for me with a
client.
It was just like, I I talk somany people out of stuff, you

(21:43):
know, and when you when you workfor a big company and you're
selling stuff, it's harder totalk people out of stuff.
But I think like less is alwaysmore.
And uh especially for verypersonality-driven businesses
that are like individuals, theyleft the big guys kind.
Because they were stifled.
And so the last thing you wantto do is create new stiflement

(22:04):
that like interrupts yourability to be, you know, the
special person that you are thatdrives your business.

SPEAKER_00 (22:11):
Let's dig in just a little bit more.
Um, a prior guest was MartinMartinez, Marvin Martinez, a
band saw AI.
Now, I bring him up specificallybecause he has the opposite
approach, diametric oppositeapproach.
He's talking about he makes acall, he shows up, and really

(22:31):
only in a matter of hours he'sdone his discovery and he's
making fixes.
Now, his point of view is hegoes for the lowest of
low-hanging fruit.
Instantly an obviousrecognizable problems that to
your point might or might nottake an AI solution.
And then he kind of burrowshimself in and works up from

(22:54):
there.
Now, the reason I showcased himin contrast to you is to
highlight your own value prob.
And the next question is whatkinds of solutions after you do
this deep discovery dance, whichto your point is is extended in
time, can be expensive, whatkind of solutioning do you do?

(23:19):
In contrast, let's say to well,it's obvious that you know 56%
of a call is wasted on rigmarol.
And if we just put in thislittle thing thingamajig, then
you'll be better off.
Yeah.
How are how are you differentthat way?

SPEAKER_02 (23:39):
So one, I don't think it's either or.
It's like a and but because Ican absolutely take the SFW and
show up to the meeting with aproof of concept now.
And it's almost free for me tojust because I have universal
tooling and my tooling isagnostic to the point where I
can just white label it to yourtooling, and so you give me the

(24:00):
specifications and I could buildmy first guess of what you want
and bring it to the meetingbefore you sign anything.
And so that is a my strategicadvantage and it's your
disadvantage, right?
Because that is such an easything to do and demonstrate that
you may not realize this is onlyever going to be 87% good

(24:21):
because of all that load-bearingstuff.
Like, you know, when you buy anold house and you tear open the
walls for the first time and youkind of hold your breath and see
what the situation is going oninside there with mold or no
mold.
Yeah, yeah.
That that is that is everyone'severyone's stack at their
company right now.
You'll find out, like, oh, thisone server has been online for

(24:42):
22 years.
And if it goes down, who knowsif it'll come back up again.
And so there's a lot of thingslike that that don't get the you
don't even find that indiscovery.
Sometimes you find it six monthsinto the engagement.
And so, how do you arbitrate thethe renegotiation of the
contract or whatever?
You gotta have a goodrelationship and you gotta
understand it's like, okay, thisis structurally very bad, and

(25:03):
here's what we got to do to fixit, and no amount of AI is gonna
fix it because this is afundamental technological
problem.
And then that is more like atraditional software project
than it's ever been.
So what you do is then you makeevery bit of work independent of
the other bit of work.
So if you know ghosts are overhere and gold is over here, we

(25:27):
can keep working on the goldwhile we work at the ghost, and
we'll get to that problem in duecourse of time.
But if you're predicated on thatbecause the SOW forced you into
the path of the ghost, you runinto an issue where all work
stops until you fix it.
And I think that is also likethe biggest problem in software,
is like the side quest kind oflike debt, where it's just like,

(25:50):
oh, I thought we were justbuilding a new deck.
It's like, yeah, but yourfoundation's sinking, so now you
need a new foundation.
You know, and it and I think alot, like, you know, I come from
family contractors, obviously.
So I think a lot about thesethings where it's just like,
yeah, an addition to a houseseems like a simple, cheap thing
until you start digging intostuff you didn't touch and you
didn't architect.

(26:10):
So um I don't want to like Idon't want to, you know, make it
tough for me and overpromisesomething that I can't like in
good conscience promise.
But also I want to be honest andbuild a relationship with you
that hey, maybe this addition issomething you should put off
until you fix that leaky roofand I can help you with the
roof, but that's a differentconversation and that's free

(26:33):
advice.
And I think that is a betterbusiness standpoint than a lot
of technology and servicescompanies have where they're
like they say yes uh to arelationship they may not be
interested in keeping foreverversus me trying to find uh
clients and partners that are init for the long haul and we have
a good relationship and maybe weonly work together for 18 to 24

(26:56):
months.

SPEAKER_00 (26:57):
Let's make that longer-term relationship real to
our listeners and viewers withsome use cases.
Can you share uh at least inabstract or anonymize some of
the some of the stuff thatyou've done based on this deeper
dive, based on creating thatrelationship, and based on

(27:19):
really spending some time underthe hood before you really make
a move?

SPEAKER_02 (27:23):
Yeah, like for example, uh I had a client and
they wanted to have a chatbot ontheir website that answers
emails, right?
And makes sense, you know.
I answering emails is hard.
Um, but they're we're a brokerfor health insurance, so it
quickly became apparent that youcan't have a chatbot ask 167

(27:44):
questions because that'sexcruciating.
You know, that is just like nota good experience.
So instead of having the chatbot ask 160 questions, let's get
you to fill out the form with160 items, and then you come to
the meeting with the form filledout.
And so that kind of like it'slike, oh yeah, fill out a form,

(28:05):
it makes a lot of sense.
But then you run into theproblem of like, oh, let's
answer the questions around theform and all that.
So we take this project that ismeant to sit in front of
customers, and now it's it's uha failure path versus like you
you pick up the phone and youpress one, it's like hey, you're
waiting on hold for 12 minutes.
While you're waiting on hold, doyou want to fill out some

(28:26):
information that'll speed upthis call?
And so when they pick up thephone, uh half of it's filled
out, right?
And so, or maybe everything easyis filled out, and then the
things that require a littledigging are done by the person.
And so one, it helps the user,you know, making better use of
time, but more importantly, ithelps the employees because
they're no longer jugglingrandom systems that change

(28:49):
arbitrarily, they're handlingtheir system.
And now when they pick up thephone, a little baseball card
pops up and says, Hey, this isCindy, she just had a root canal
last week.
And so you can say, Hey Cindy,how's your mouth feeling?
And that little bit of glue andstuff, one calls are cheaper and
they're better, but also likeimproves customer satisfaction

(29:10):
and improves uh employeesatisfaction because they're not
flipping around in eightdifferent windows and they're
not like trying to make you knoweverything happen.
And it kind of surfaces the ideathat a lot of the people, and
people don't find this out untilthey start trying to automate
their business.
There's a lot of load-bearingpeople on site organizations.
You know, you'll have a personin your company who's been there

(29:31):
30 years and they knoweverything, they have speed
dials for everyone and and theyhave everyone's personal home
cell numbers and they makethings happen.
And the reason they have to dothat is because your blind spots
as a leader not being able to uhsimplify the process for them.
And so if I can kind of likepull that out into you know,

(29:53):
visibility for an organization,then that's an entire roadmap of
work, you know, for yearspossibly, of transforming your
business uh into a data-drivenuh unified brain that knows that
the president of the UnitedStates is Donald Trump and not
Joe Biden because a document wasuploaded six six years ago, and

(30:15):
uh can tell you the office hoursand also collect as much
information as it can before itbugs Cindy, who's really quite
busy.
So that's kind of like my mycareer pitch for a long time.
It's just like uh I've had had acontact center say, Can you have
the AI tell jokes?
I'm like, of all the problemsyou have, telling jokes is the

(30:38):
one you focus on.
And it's like, well, what kindof jokes?
And there's all these questionsthat come out of these problems,
and I think technologists sayyes, as opposed to probing
deeper into that question,because they know this is like
the unending banana stand ofmoney for them, where they can
it's like, yeah, this is gonnabe a problem that we're gonna
get to 80% quality, and thenit'll be stuck there for months

(31:00):
or years.
And I'm not interested in doingthat anymore.

SPEAKER_00 (31:03):
So human beings in general, regardless of their
station in life, have thisknee-jerk tendency to be
emotional about stimuli, andthat forces folks into the
tactical bucket almostinstantly.
So if you're showing concepts toa client rather than think about

(31:24):
the brand strategy and theoverarching story you want your
company to tell, they're gonnainstantaneously respond that
they don't like the color of thelogo or this image needs to be
smaller or larger.
That's uh natural.
And in operations, what you'redescribing is analogous, where

(31:46):
uh you know you do you do anevaluation of their
infrastructure, and their firstquestion to you is we want to
make a funnier bot.
Maybe we should put a joke inthere because they're on hold
for 10 minutes, and maybe we canentertain them.
When the missed opportunity isthey're on hold for 10 minutes,
and the series of questions thatthe operator will need to endure

(32:08):
with the customer once they areon could be addressed with the
utility of a bot given today'stechnology, and it could be a
fun, engaging experiencefacilitated by a user interface
that's friendlier and more funto use.
So if I hear you properly, uhstop thinking tactically, stop

(32:31):
thinking about knee-jerk,instantaneous, oftentimes
useless solutions just becausethey seem to be something that
might have a quick fix.
Step back and look at the flowof your information throughout
your organization, go back topain points and need, and let's
build it back up from thevantage point of that kind of

(32:53):
optimization rather than runningaround with a bunch of band-aids
and plugging in holes andrelying on a 30-year-old veteran
employee who grew up with thisinfrastructure and has been her
90% of her job or his job isfilling holes.
Right.

SPEAKER_02 (33:11):
And and I think like it's counterintuitive because a
lot of people come up to andsay, Well, what's the ROI of
this thing that you described?
Like, well, how do you measurework now?
And they don't have an answer,they don't know what the cost is
of like changing the text on thecan of soda is.
Well, it's about 47 meetingsacross six different
organizations and hundreds ofhours of billable legal time and

(33:33):
all that stuff, but you'll shipAI in two weeks because you want
to be fast and agile.
And so, what how do we how do wemake change management for work?
Have as much uh liability andattribution as the text that
goes underneath the logo on thesoda can, I think, is uh is like
a unanswered question in a lotof organizations.

(33:55):
And so this is all stuff you getfor free when you work a toil
bill, because you know, we don'tlike doing work at your behest
because we say so.
And so maybe I I what talkpeople out of surfaces with us,
uh, or maybe I choose not towork with people because I see
no uh daylight to work that waywith them.

(34:15):
And I think that has to be okaybecause you know it it's part of
a two-fold system.
Do there is the change from aplace of wanting to change, or
is it change coming from wantingto rubber stamp a regime of
decisioning and make stuffhappen?
And I think that is alsosomething that like a lot, you
know, I think if people wantedthat, that they would not have

(34:36):
left working for the big guysand started their own companies.
So I think that's a very easierconversation to have outside the
enterprise than it is to likewhen I'm talking to Jimmy, the
insurance broker or Cindy, thethe veterinarian who's starting
a veterinarian application for awebsite, because I think they
realize that yeah, I could havejust bought a SaaS product to do

(34:59):
this if all I wanted was get itdone.
But I think there's a better wayto do it.
And so that is like the littleglimmer of light that I start
this engagement with people.
It's just like, yeah, like it'sall it's all kind of arbitrary,
and you just got to do it theway you like it.

SPEAKER_00 (35:13):
So do you want any way or they could have bought a
joke bot right and so solvetheir problem now?
You're the change managementguy, yeah, and change management
for upper management mightembody some of the
characteristics that you'redescribing uh optimization,
heightened efficiencies, ROI.

(35:35):
They're thinking along thoselines.
They see the spreadsheets,they're the big picture, the
chess master moving the piecesaround.
But folks throughout theorganization, they equate change
management with unemployment.
So, how do you how do you dealwith this notion that I believe

(35:55):
in a recent poll only 27 of thepopulation approves of AI or
thinks that AI will benefitthemselves, society, and the
rest, which I believe if you dothe math is at least 73%, hate
AI and think AI is going todestroy their life and

(36:17):
eventually kill all humanity?
So, so change management has alot to do with, to your point
earlier, building trust.
But this trust building happensat senior management.
They're the folks who hire youand they want the changes to be
managed.
But then how do you assure folksthroughout the organization that
you're actually a knight inshining armor and not an

(36:40):
executioner?

SPEAKER_02 (36:42):
Yeah, you know, and and I think that's a huge part
of, you know, I I've said thisbefore on podcasts where no one
is better at the slow walk thanthe American worker.
You know, if you have anAmerican worker doing something
they don't want to do, they'llbe like, they'll do just enough
of it to keep their job.
And I think AI is in thatposition because its branding

(37:02):
has been horrible, itsimplementation has been
horrible, because no one hasever posited the value inherent
in the process to the averageeveryday information worker or
developer or coder.
And you know, we've seen this intime immemorial automation,
going back to the Luddites,going back, you know, hundreds
of years to you know, the steamengine, um, that people uh are

(37:26):
afraid of technology and it's upto the technologists to uh find
something that works foreveryone.
But in AI and in tech, thecustomer is not people, it's
venture capital and investorsand uh things so they have no
skin in the game whether or notyou like it or or it does your

(37:46):
job better than you or not,because they're trying to sell
it to your boss, and bosses havenever needed a reason to lay
people off in the history ofwork.
And I think like all thesethings kind of point to a
picture of uh if AI is a posterchild of work versus employers,
of course it's gonna have ablind eye.
I I I once had um I worked for ataco company, a a taco chain,

(38:12):
and they created a way for youto order tacos via Facebook, and
it had horrible approvalratings.
And I was on this two-hour call,and then I unmuted about an hour
into and said, What happens whenthe food comes cold?
And they're like, Oh, they haveto call the restaurant.
So there's nothing you can do inFacebook to get more food sent

(38:32):
to you or anything.
It's like, no, that is allbecause it's just like the this
uh Paul Varillas is aphilosopher in and from Paris in
the late 20th century, and hesaid the ship was the invention
of the shipwreck.
So sometimes technology eventsproblems that weren't foreseen
or were very foreseeable, thatdidn't exist before they

(38:53):
invented and as is after theymeant it.
And it's your responsibility asthe inventor to deal with them.
So uh technologists who push AIsystems should have uh solutions
for workers that produce thingslike universal basic income or
other things that supporthumanities and families and
workers.
And if we don't, then there'llbe other societal problems that

(39:14):
will cause AI to be uhinsignificant in comparison.
And so I think like when itcomes down to the end of the
day, I'm not I don't I don'thave stock in Chat GPT, I don't
care if they succeed or fail.
I want people to work well andbe happy at their job and do as
much as they can to do theAmerican dream of owning family

(39:36):
and you know having a home andsending your kids to college and
not having to go fund me whenthey get sick.
And so if you look at thingsfrom that standpoint, there's a
lot of opportunity there.
Um but I think there's no uhmarketing positing that to
people in a way that is likebelievable or acceptable to the
populace.
So until that happens, AI willbe kind of a dirty word.

SPEAKER_00 (40:00):
Tons of mixed signals.
You mentioned UBI, universalbasic income.
To me, this is uh a real mixedbag of tricks.
On the one hand, you're you'reespousing that the technology
will generate so much capitaland heighten efficiencies to the
extent that there'll be allthese trillions of dollars that
the government could tax out ofthese companies, which could

(40:23):
then be redistributed throughoutthis population.
But it begs the question of thepopulation losing their damn
jobs, which would necessitatethis kind of payout.
So I'm I'm always quizzical whenpeople cite UBI as the savior to
this problem, because again, itbegs the question of AI taking

(40:44):
away millions of jobs.
On the flip side, the messagingsucks.
I agree with you, because a lotof people's work is grunt work,
remedial stuff that they hatebut they know they need to do.
And what AI can do, at leastinitially, is automate a lot of
the dumb drudgery and it'll openup more hours for that employee

(41:08):
to do things that are ostensiblymore human, which is connect
with other human beings, performmore strategic functions, and
presumably add more value to thecompany than doing these rote
behaviors that constitute nowmaybe a half of their job.
So I think that's that's amissed opportunity right there,

(41:30):
where we're not here to takeyour jobs, we're here to improve
the quality of life at your job,which we're hoping and designing
for you to keep.
And that that's lost in the inthe mix frequently.
However, the question is againbegged by companies like Stripe.

(41:51):
You got Jack Dorsey with wearinga hat that says love on it and
laying off 90% of his workforceon a Slack call.
And then you've got companieslike Meta who have just unloaded
a huge percentage of theirworkforce, and you have these
big tech players who, on the onehand, are boosting their capex

(42:15):
in AI investment by literallyhundreds of billions of dollars,
and on that same investor call,they are boasting that they're
laying off tens of thousands oftheir workforce.
So it's one big convoluted mess.
I I agree, the messaging is justterrible any way you look at it.

(42:37):
But let's go back to you.
Uh, let's be pragmatic.
When when you come in a company,how how do you onboard in a way
where you're not bringingrevolution?
Right?
You're you're you're you're thedoctor.
We've been you've we've beenusing your role as an analog for
being the physician of businessoperations that are sick due to

(43:01):
all these problems.
Instead, a lot of staff willlook at you as you know, the guy
with the pitchfork and and uhand riot mob coming to burn down
the uh the factory, burn downthe office you know, we don't
have to get into like uh praxisand work stuff here, but I do

(43:25):
say that revolution is theoutcome uh when all else is
failed, right?

SPEAKER_02 (43:30):
And because people see no other choice to do things
that they feel they cannotsurvive without.
And so if we have a problem ofmalaise, I think most companies
suffer from malaise and notrevolution, where uh they don't
feel like their voice matters ornothing changes or uh new system
same problems, um, those arethings that where the worker and

(43:52):
I don't think you know there's alot of things in the workplace
where they want work to be afamily, and I think a lot of
people are happily to have.
Happy to have a job that getsthem their insurance and their
food and money on the table, andthey go home and they don't
think about it until the nextday at nine.
I think that should be fine.

SPEAKER_00 (44:09):
Yeah.
And easily 95 out of a hundred.

SPEAKER_02 (44:13):
But then you have Jack Dorsey's of the world who
has he ever dug a ditch?

SPEAKER_00 (44:19):
You know, like he did he did dig a ditch, and
that's called Twitter.

SPEAKER_02 (44:24):
Oh yeah, right, yeah.
But he he sold it to an evenbigger ditch.
So um and I think like I thinklike that is just like we have a
lot of millionaires andbillionaires right now who don't
have a very strong sense of one,the value of work and the sense
of what uh is the cost of work,right?

(44:46):
Because anyone that says firing80% of your company is the
hardest day in their life issuspect in my eyes, because that
is you it's already you'vealready decisions already
happened.
There's no you're not goingaround each person individually
and looking in the eyes andsaying they're doing I I I've
been I've been not fired as manytimes as I've been fired by

(45:08):
layoffs.
And the not fired, I think, isthe hardest thing because
everyone in the room loses theirjob but you and two other
people.
And so they all line up to gettheir hand and their stuff, and
you're sitting there at yourdesk trying to work.
Like that's crazy.
That's a crazy thing to putpeople through and to and and
then to go on uh your blog andsay, yeah, it's the hardest day

(45:31):
of my life.

SPEAKER_00 (45:32):
It's like the bad news is you're fired.
The worst news is you're not.

SPEAKER_02 (45:39):
Right.
Yeah.
All their jobs are now your job.
You know, like, and oh yeah,you're not gonna get a raise
this year.
There's no bonuses this yeareither.
So but you still have a job.

SPEAKER_00 (45:49):
We're bringing in the AI guy.

SPEAKER_02 (45:51):
Yeah, we're bringing well, so and so I come in this
situation where it's just like,okay, I have the boss who's out
of connection with the workers,and I have the workers who have
been suffering at the hands ofthe boss's whims for many years.
And so I think the the bestthing for me to do in a
situation like that is one toshine light on data problems.

(46:12):
Like we need to understand howthe work is happening, and we
need to have accountability fordecisioning, even from the sea
level down.
And uh and I think like light onshadows is one way to do that.
Uh, I always encourage people touh work as not a family.
Like I have my nieces andnephews are getting in the
workplace, and I tell them to bepromiscuous as they possibly can

(46:34):
with employers, because if youspend 20 years with one company,
you won't learn anything that'llhelp you at the next company
when they lay you off in your40s.
So, you know, you dig as manyditches and you know, paint as
many pumpkins and whatever weirdstuff you can do before you have
the family and all that stuff toworry about where it's harder to
be mobile.
And uh because worker workermobility now, like we went from

(46:59):
the COVID times where workershad a lot of leverage, and now
uh the pendulum swung the otherway, and workers don't have a
lot of leverage, but they do,they can slow roll the AI
projects, they can quit the job.
Uh it's dangerous to quit a jobnow, you know, everything's tied
to health insurance and allthat.
But yeah, that's always on thetable.
So you have to treat people withrespect.

(47:20):
And I think if I come on thetable and I'm Karl Marxing and
I'm banging my shoe on thetable, I'm not gonna do
anything.
But if I can like familiarizemyself with the complexity of
relationship between labor andtheir employers and technology,
there's surely some sort ofthing that makes things a little
better.
And the process isn't to solveit in the day, it's to create a

(47:43):
process where you can startchipping away at the problems
systematic systematically.
And so if we can go in and dothat and we can use L AI to like
it's always easier to see it inmotion than to see it on a
whiteboard or see it in ablueprint.
So if we can like just bring itto life and look at it and say,
oh, this stinks, instead ofspending six months on it, then

(48:06):
we should do that, and then weshould like embrace things that
make work better instead ofmaking work more work.

SPEAKER_00 (48:13):
There's a pragmatic aspect to humans being left in
the loop and humans keepingtheir jobs that the doomsayers
ostensibly say they're all goingto lose.
One aspect is just customerservice.
You could have a touring-testedbot do a lot of it for you, but

(48:36):
at the end of the day, whatgreases the wheels are the
interpersonal relationship andpeer-to-peer functions of so
many companies.
Companies hire people to relateto other people.
Doesn't matter how sophisticatedthe bot or the robot, whether
the predictive or reinforcementlearning will do the shot or
not.
You need people.

(48:57):
It's a people business doingservice, creating products for
other people.
That's one saving grace, Ithink.
And the other is the adoptioncurve where the technology could
be there, but it often takeshuman beings a generation,
sometimes two, to really adaptto the technology.

(49:17):
Remember Friendster, the socialnetwork.
Uh that that that had a lot ofthe functionality of a Facebook,
but people weren't really quiteready for a social network and
sharing.
And the tech might be there fora lot of this AI plugin and

(49:38):
adoption, but people don't getit.
Uh, they they have troublelearning it and implementing it,
and it's a lot slower, I think,than people realize in terms of
plugging in and letting it taketake control of an organization.
So I think we're getting alittle bit ahead of ourselves
with the hype curve that you'vebeen talking about.

(50:00):
Yeah.
Not only from the vantage pointof business optimization, but
actual implementation,utilization, and cultural
transformation.
Yeah.

SPEAKER_02 (50:10):
Well, I mean, if you think about it, everything is
repeated modalities of stufffrom the infancy of computers.
Like if you go back to look atXerox Park in the 70s, it looks
a lot like Windows still.
It looks a lot like OS X still,we're still clicking on icons.
And so people.
So people make people make blogsthat are like an online version

(50:32):
of a magazine, and they makeTwitter, which is online version
of a text message, which is theonline version of a telegram.

SPEAKER_00 (50:37):
Yeah, yeah.
Uh the when we look at ourphones like 500 times a day, you
know what we're really lookingat as far as the UI goes?
Yahoo.com 1998.
It's like, remember email,sports, when people didn't
understand how to use the web orwhat it was, you couldn't like
navigate to anything.
There weren't real searchengines.

(50:58):
So you needed a button to pressto perform a function.
Yep.
And apps on a mobile device arepretty much the same.
To your point about it repeatingand iterating the technology,
the UI, how people access techin the way that's most
functional to them based ontheir learning curve.

(51:21):
Right.

SPEAKER_02 (51:21):
And I think a lot of it at the end of the day, it you
can feel very busy doing allthis, but you don't get anything
done.
Right.
And I think a lot of technologywork feels the same process
where, okay, I'm going betweenSlack and Teams and I'm clicking
all the buttons and respondingto all the things.
It's like, oh, two hours left toactually do my work.
I've had my hours of meetingsand I clicked on all my things

(51:43):
and I responded to everything,and I made my inbox empty.
Now time to write the report Ihad to write.
I'll have AI write it and Iwon't read it and I'll submit
it.
And you know, they'll they'lllaugh me out of high school
because yeah, I said AbrahamLincoln uh was the first man on
the moon or something, right?
And and I think like that's alsothe the thing when we talk about

(52:05):
ROI is like, oh, if just becauseyou can ship code ten times
faster, you shouldn't ship 10times more code.
You should take the time and doall the things you cut because
you couldn't have time to do it,like accessibility and uh making
sure all the bugs are out.
So 1.0 is a real 1.0, you don'thave to ship 1.1 the next day.

SPEAKER_00 (52:24):
Let's flip this conversation around.
We were talking about use casesthat are very pragmatic and uh
and fairly immediate and reallyerupt or emerge from your
discovery in a way that'suseful.
Uh let's say you do yourdiscovery and a company like
Mookie Spitz Bald Ambition, Iactually need a proprietary LLM.

(52:48):
I just you you did the discoveryand you just concluded I need a
deep seek cranking on my laptop,and I need agentic technology to
help me distribute my podcastacross all my channels, and I
need agents to to do theprocessing, the editing of my
videos, and then we worktogether and you envision a

(53:10):
high-tech solution, okay?
Like really cut cutting edge.
Like you suggest I get Nvidia'sNemo on my desk, and you suggest
that I get ClaudeBot to you knowstart automating my functions,
and we get serious, we get deepabout AI in 2026.

(53:32):
You you conclude that thatactually would work for bald
ambition LLC.
Uh how does how does that work?
Where you bring you bring thebig kahuna, let let's let's flip
it around, let's stop talkingabout just plugging in Slack in
a more efficient way and talkabout bringing in the big guns

(53:53):
and doing it in a way thatactually adds value instead of
just blowing smoke up upeveryone's posterior for the
sake of jacking the SOW.

SPEAKER_02 (54:03):
Well, I'll tell you what we do at Toyville.
And so the first week ofToyville, I took the week off
because that's a great way tostart a business, is take the
week off.
So I took the week off and I didnothing but vibe coding for a
week.
And at the end of the week, Ihad 3,000 computer programs on
my computer.
And I did an analysis of thecode and it was five programs.

(54:23):
And when I thought about it alittle harder, it was two
programs.
But that 3,000 computer programswas like 20 gigs of stuff on my
computer taking up a mess, andall it was was change
management.
It was CRUD, which is create,read, update, delete operations.
And so we started building asystem uh is basically uh
governance tool that wrapsaround things like flawed.

(54:45):
And so whenever you use AI likethis, uh one you should own the
model.
And if you can't own the modelbecause it's too expensive and
you don't have$12 billion aquarter to burn, you can rent
it.

SPEAKER_00 (54:57):
That's what Apple did, by the way.
Apple didn't want to do that.

SPEAKER_02 (55:01):
I think Apple's very smart about what they're doing
because you notice they haven'treleased anything yet because
they're company run by designersand it isn't designer ready yet,
this technology.

SPEAKER_00 (55:11):
But I'm all the capex that all these companies
are burning through to createtheir own frontier models.
Apple's like, you guys do allthe heavy lifting, and when you
finally figure it out, we'regonna plug it in.
At first, I thought Tim Cook wasout to lunch about AI, but the
more I think about looking back,he saved Apple a trillion
dollars by doing nothing.

(55:32):
I agree.

SPEAKER_02 (55:33):
I agree because like uh we our internal tooling, we
run Apple computers and they'reready to do this stuff without
buying the stuff you described.
You can any computer that saysApple intelligence is more than
adequate for most people for AItasks of personal nature.
And so our tooling that we buildin-house, um, it starts off with

(55:54):
stupid computer programs andthen it goes to Apple AI and
then it goes to cloud AI whenit's really stumped.
And the the key difference isevery time you go to cloud AI,
we count how much we use andwhat it costs, and we use what
we learn to write computerprograms to avoid doing that the
next time.
And so instead of 3,000 computerprograms, you have one computer
program, and everything becomeslike a practice or praxis or

(56:18):
ritual, whatever you want tocall it.
And so uh then your entirebusiness is you know, it is a a
personal Google that lives inyour computer and they'll never
use it to train Chat GPT.
Um, I remember and you probablydo too, when Amazon first came
up and they had the storeswithin a store.

(56:40):
So all these companies like,well, we can't beat Amazon, so
we'll have Toys R Us.amazon.com.
And look how that worked out forToys R Us, you know, and I think
a lot of people who build theirbusiness on these tools are
gonna run to the same trapagain.
So I teach my clients, yeah,absolutely, if you can afford
getting an NVIDIA computer andrun it locally, you can get most
of what you need done.

(57:01):
And if we need to like do asprint, we'll spend a couple
hundred bucks on Claw this monthand get it to write the program
and leave Claude, and that'swhat they know.
That's why the that's why theywant to own your output.
That's why they want to make youhooked on being paying rent for
AI for the rest of your life.

SPEAKER_00 (57:18):
This is to me the biggest tension in AI if we're
looking at trending and futurescoping.
Uh, I liken it to the IBMservers succumbing to the PC and
the Apple personal computer, oreven the Pro Tools example,
where you had a lot of musicianswho were renting studio time

(57:41):
very expensively from the bigcompanies, and then the software
came and they could just do itall at home on their on their on
their PC.
And that's basically this bigpicture to local.
So you have all these companiesinvesting trillions now,
building these expensive bigfrontier models, and then they
charge you for tokens.

(58:01):
You're renting the computingtime from them, and they own
your content, and they're gonnamonetize your engagement with
them, they're gonna provideadvertising, it's gonna be it's
gonna be jacking you foreverything, including your
subscription dollars.
And the alternative is um get alocal LLM, to your point, on

(58:22):
your own laptop, on your own PC,and you own your own content and
you do everything locally.
And this battle between theservers versus the PC is gonna
be the defining dichotomy, Ithink, in AI as it moves
forward.

(58:42):
And not enough people, I think,are talking about this.
Uh the investors are all aboutthe big frontier models.
I think with the dangling fruitof AGI, if if we keep pumping
this up given the existing LLMarchitecture, we're gonna have a
hell 9,000.
And most computer scientists andeven egghead bald guys,

(59:04):
podcaster consultants like meare like, no way, Jose.
You're not you're not gonna getartificial general intelligence
from uh a large language model.
You're just not going to.

SPEAKER_02 (59:15):
No.
No, I mean, you know, gettingartificial intelligence from a
large language model is likegetting a gravity predictor out
of a pachinko machine, right?
Yeah, it is a is an input andoutput, and all it does is
statistically bias the result.
And so if you look at AI likethat, we have things called
computer programs, and we've hadthem for decades that are biased

(59:36):
result machines, right?
So if you tell a calculator andyou type in a number plus a
number, it gives you an output.
And so that bounded statemachine can be made for hundreds
of a cent now.
So why not?
You know, we have all thistechnology in our house in our
lives, in your phone, that ismore than powerful enough for a

(59:56):
private AI cloud.
And why hasn't uh it happenedyet?
One, because the entire marketis perhaps owned by a oil
company like consortium of RAMand CPU and GPU and tech
companies that have decided thisis the way it has to happen,
maybe if you're in this.

SPEAKER_00 (01:00:18):
They're patting each other on the back and like
eating themselves, you know.
It's the chihuahuas and thecats, you know, it's like you,
you know, one's eating, theother's excrement.
So it's like uh, you know, yeah,open AI buys NVIDIA chips and uh
in in open ASI, uh OpenAI isdependent on the data centers,
and the data centers aredependent on the frontier

(01:00:40):
models, and you know, round andround it goes.

SPEAKER_02 (01:00:42):
Yeah.
I mean, like if you buy Chat GPTPro at the$200 a month, by my
estimation, that's 300% loss forChat GPT on pure compute.
So why would you do this as acompany?
And you know, a company that'slost, you know,$18 billion and a
quarter has never ever succeededin the history of mankind.

(01:01:02):
And I think their thing thatthey hope is either one the
blunder on the AGI, and theywill be the only game in town
and they'll own the AGI.
Um but you know, someone willuse the AGI to train another
AGI, they'll be cheaper, andthat'll happen the next day.
Then take over the world.
So but so I think like it it'sthis it all these do this

(01:01:24):
discussions I think distractfrom am I gonna lose my job?
Right?
Because you know, we're debatingabout is Peter Pan real or not?
Meanwhile tech companies arelaying off people and saying you
have to come in the office anddo all that stuff.
And so these conversations arelike okay, I love to have a
spitball conversation about isAGI real or not?

(01:01:45):
It's not, and uh and do that allday.
But like you're getting paidhourly to do a job, so how do we
get that done today?
And I think that is like a lotof the tension people have at
work where it's like, wait, it'scool to play with this new
thing.
I I have two hours, I will neverget back, and now I have to go

(01:02:05):
back and doing the job in apredictable way because it's due
at a predictable time, so I'mgonna go back to the non-AI way
to do it.
So I think there's a lot of likecontext switching you have to do
until this technology says whatit says on the tin, and also
like a company is prepared tomake the changes they have to
make to make it work better.
And until that happens, I thinkwe're kind of at the whims of

(01:02:27):
like tech companies andfunctionalities and leadership.

SPEAKER_00 (01:02:30):
There's two big developments though that do have
pragmatic import.
And the first is agentictechnology that just blew up
with anthropic.
So you mentioned profitability,anthropic's bottom line
increased by 30x.
They were predicting about abillion in revenue and they
cranked to 30, and they'reactually predicting

(01:02:51):
profitability in maybe a year ortwo.
Woohoo! They're actually gonnamake money, maybe.
And then the the other one isthis local LLM stuff that we've
been talking about.
So agents can do stuff andthey're amazingly adept at it,
and then local proprietary LLMsare walled gardens of content

(01:03:13):
that turns the model upside downfrom these data, data-centered
server minds that charge you forevery token.

SPEAKER_02 (01:03:22):
I I do think there's gonna be a market spread of like
personal AI and industrial AI,right?
And so, like the AI that acompany like Microsoft needs to
do a day of work is structurallydifferent than the work you do
to produce the podcast.
And you'll probably have enoughAI in your machine that you get
for paying the$2,000 for yourlaptop or whatever that merits

(01:03:44):
the ROI and its cost, and it'llbe good enough for many things.
And uh the AI that you take tomake like a Marvel movie or
whatever would have to be baredby the cost by the people who do
it, and that's fine.
But I think like we have to getout of the speculative industry
where like, okay, uh so atToyoville, our AI system, we

(01:04:05):
basically wrap anthropic clawedand we divert it whenever
possible to an application thatdoes what I ask it to do.
So I don't send anything in thecloud that already exists.
And because of that, only 80 13%of clawed things actually make
it to claw when I'm using clawedcode because I'm just rebuilding
stuff on the existing structure.

(01:04:25):
And of that 13%, it's 100% casheffective because I rewrite the
prompt to not be different everytime and only say the stuff we
need to get to the next step.
And so I think the future issomething like that where okay,
uh I'm using$1.30 of Claude aday, and I'm using the rest of
the computers I paid for allday, and instead of spending$500

(01:04:47):
a week on Claude, I get by onlike$15 a day.
And that to me is that's moresustainable as a business, as a
business owner, and I think it'smore sustainable like
environmentally, because like weshouldn't have to like you know,
it's like using Terry Claw fortoilet paper, a lot of these AI
processes, where it's just like,hey, I need to I need to be able
to write an email to someone'slike, okay, I'm gonna write for

(01:05:09):
you an email program first, andthen we're gonna like figure out
how email works.
And all this happens if you lookat the internal memory every
single turn.
And so if if AI is expensive, weshould price it so and we should
make the tooling more efficient,and all that stuff has to
happen.
And I absolutely don't think anyof my clients should be on the
hook to build or write thatcode.

(01:05:30):
So I I I I do it for myselfbecause that's how I retain my
competitive advantage, but Ithink it's not ready for prime
time for a person to just use inlike buying a suit from Walmart.
You still have to go to BrookBrothers and get it measured and
tailored to you in a way thatrequires people like me.

SPEAKER_00 (01:05:48):
If I'm hearing you correctly, it's local, not these
big servers, as best you can.
And then you need an expert likeyou, Mr.
Swim at Toilville to plug it in,to customize it, eliminate
redundancies and overkill, andoptimize it to your work streams

(01:06:11):
that you determine through yourrigorous and detailed discovery
process.
So that's kind of a good bowthat we could put on your value
prop, which is uh come to Peterwith uh your business issues.
He'll look under the hood in away that's detailed and
comprehensive and thoughtful.

(01:06:33):
Only after he determines reallywhat's broken will he try to fix
it from a holistic level.
And then he'll apply the toolsof the trade in a way that's
most expeditious without thehype.
And when he does implement moresophisticated AI solutions,
he'll ensure that you maximizethe control of your data and

(01:06:57):
he'll continue the relationshipwith you to customize these
ever-evolving technologicalmiracles so that uh it it keeps
working, it taps into the pulseof the latest technology and
isn't giving away your data andyour soul to these huge frontier

(01:07:17):
models who are gobbling up theworld.
Is that a decent uh overview?
You know, I'm gonna need to sendmy pitch deck to you for you to
punch it up because that wasgreat.
Thank you.
All right.
That's my that's my free serviceto your guests who come out of
all the ambition.

(01:07:37):
We'll we'll wrap it up intopitch because you know,
essentially this podcast isabout consultative selling,
storytelling, and business.
But you take the complex, youmake it simple, and you make it
resonant in a way that'slogical, inevitable, and
differentiated.
And uh, it sounds to me likeyou've got you've got this going
on at Toilville.

(01:07:59):
Thank you.

SPEAKER_02 (01:08:00):
Yeah, and of course, like all right, I do emails for
free.
So if you have a quick questionor whatever, you can hit us up.
Uh, our website is people makeit better.com.

SPEAKER_00 (01:08:09):
I'll have links in the description below of the
audio and the YouTube video.

SPEAKER_02 (01:08:15):
And I I I work a lot with people who don't think they
can afford someone like me.
So, like not-for-profits, uhreligious groups, community
groups, uh, we can makesomething happen uh to help you
out.
So reach out to me and happy totalk to anyone.

SPEAKER_00 (01:08:28):
Great, thank you, Peter Swim, from founder of
Toilville.
And you should rename it afteryou you initiate a relationship.
It goes from Toilville to uhwell what's what's desirable end
state from Toilville to happy,happy opulist.

SPEAKER_02 (01:08:49):
You know, like I'm I'm a son of the soil.
I I don't I don't turn my nosedown to toil.
So it it it's alright if I ownthe harvest.

SPEAKER_00 (01:08:58):
That's right.
You we we reap what we sow, andit sounds like your farming is
very thoughtful anddeliberative, deliberative, and
good for good for the businessenvironment.
Thanks so much, Peter, for yourtime.
Like, subscribe, comment, andshare, and check his links below
to initiate uh at least thebeginning of a discovery call

(01:09:19):
for Peter to check out what yougot and see if he can help.
Thank you so much.