Building an AI Integrated Workforce

Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
Speaker 1 (00:00):
We are back here on the Restaurant Masterminds podcast, and
today we are going to be diving into some fun
stuff and that is going to be a I the
see AI is going off in the background already, AI
and hr is it a collision or a match made
in heaven? Joining me, of course is mister Paul Mullinari
and Stacy Kine.

Speaker 2 (00:19):
How are you guys, fantastic, doing great?

Speaker 3 (00:23):
I'm just preparing, you know, for a party weekend where
hopefully we keep all of our digits, nothing gets blown
up and gets Do you do a.

Speaker 1 (00:33):
Big firework thing at your house?

Speaker 2 (00:35):
We blow up some stuff, we blow up.

Speaker 1 (00:39):
This is the first year I'm gonna let my eight
and nine year old actually use fireworks the first I'm like, hey, listen,
only one limb or two you could? You know you
have others that you know, don't worry about it.

Speaker 2 (00:53):
Yeah, you know you got to touch the stove baron.

Speaker 4 (00:57):
Guys, you know that I am in the Hall of
the Star Spangled Banner.

Speaker 1 (01:01):
That's right, Yeah, that's right.

Speaker 4 (01:03):
So we do it up here in Baltimore, that's right.

Speaker 1 (01:08):
See you're back in the pink rooms. Are you finally
off traveling?

Speaker 5 (01:11):
Now?

Speaker 1 (01:11):
What's going on.

Speaker 4 (01:12):
I am off travel until Tuesday when I go back
to Colorado. It spends a week and a half in
Portugal and it was fantastic. That is why I am
delightfully tan.

Speaker 1 (01:25):
Where was this Lisbon or Porto?

Speaker 4 (01:27):
Where no I went to the southern Portugal to the Augart.

Speaker 1 (01:32):
Oh, just sat.

Speaker 4 (01:34):
On the beach by myself.

Speaker 5 (01:36):
It was fantastic.

Speaker 1 (01:37):
Portugal is a great country, that's for sure. Hey, listen,
we are going to get into it today. You guys
stick around. We've got a great guest that's going to
be talking about hr AI integration for your business. You
don't want to miss it.

Speaker 2 (01:49):
We'll be right back.

Speaker 1 (02:13):
My name is Paul Baron. As the early pioneer in
fast casual, I've seen the industry evolve from just a
few operators to the most sought after segment by consumers
around the world. Now we're planning to shape its future.
Tap into decades of my expertise identifying the emerging brands
and tech winners in the space, Saber Capital will be

(02:37):
fueling the next generation of fast casual innovation. All right,
we are back here, of course, with mister Paul Mulinari
and Stacy Kine and joining us today is a special
guest that's going to be breaking into the topic of
AI and human resources and how does this all mix together.
Joining us is Soniche Monkar coming over from Legion Technologies.

(03:01):
How are you, Soniche I'm good.

Speaker 5 (03:04):
Thank you for having me on the podcast.

Speaker 1 (03:06):
Yeah, give us a okay, So let's go first into
Legion Technologies and learn a little bit about what your
core model is around the products and services you guys
offer the industry. Give us a rundown.

Speaker 5 (03:19):
Yeah. So Legion builds workforce management products. So for those
of you are not familiar with workforce management as a
software category, it helps businesses in labor intensive industries like
restaurants better manage more efficiently manage your labor cost, but
also serve your employees through things that they need like

(03:42):
scheduling and time and attendance, time capture, accurate payroll and
all these capabilities. So you can think about workforce management
as the operating system for your labor And what we
do at Legion is we have incorporated two main outcomes
workforce management which we're really missing historically. One is automation

(04:04):
and we're going to talk a lot about that with AI.
We use AI for demand forecasting, knowing how many guests
you will have in your restaurants as an example, or
accurately scheduling the right people at the right time, or
driving employee experience. But we also deliver a lot of
employee value. So that's what Legend does.

Speaker 1 (04:24):
So automation, you guys have. Would this be considered back
office when you think about the different modules that reside
inside a restaurant or retail operation, is this back office
or something in HQ side?

Speaker 5 (04:41):
It's actually both, So it is on the HQ side
from a reporting analytics standpoint. It's also in the in
the restaurants. It's also employee facing because employees use the
mobile apps. So typically your guests don't use it, your
customers are not you directly interfacing with with Legion, but

(05:04):
everybody else.

Speaker 1 (05:05):
Is okay, all right, cool. I know, Paul, you have
a lot of background in this area, so lead it
off for us.

Speaker 5 (05:11):
Yeah.

Speaker 3 (05:12):
You know, this is something that historically has been really
you know, workforce management has historically failed kind of balance
well seriously to balance well that you know, to automate
labor planning and scheduling and considering the business needs and
employee preferences. Workforce management has in the past been really

(05:36):
really difficult to accomplish this, but now when they're bringing
in AI, it feels like, now this is the golden
moment I think for for workforce and management. So okay,
a about how it how it balances that uh labor planning,
scheduling and the business needs.

Speaker 5 (05:56):
Yeah, So firstly, Paul, I want to say, you're absolutely
right that workforce management has it was not considered successful
in doing the balancing act, and you're absolutely right. And
if workforce management historically was successful in doing that, Lesion
would not be around right now. The Lesion was founded

(06:16):
with that, with that whole premise that, hey, it should
not be a zero sum game. Doing the right thing
for your employees and running an efficient labor operation should
not be either or you can have both, and let's
figure out how to do that. And the real breakthrough
really is, are you know, advanced technologies like AI where

(06:40):
the decision making around what employees need and what your
customers need if you're running a restaurant, and how do
you find that middle ground? Because there are you know,
hundreds of thousands of possibilities of how and when you
can deploy labor who should be scheduled at what time.
But you can be much more smarter than that. You

(07:02):
can say, Okay, I'm going to very accurately using AI
figure out exactly at what time how many people will
be in my restaurant for example, right. And furthermore, what
are the typical things like during lunch time, during the
afternoon evening. You know, what is the mix of items
that are changing. And I'm going to predict these things.

(07:24):
Once I predict these things with a very high degree
of accuracy, which is possible today, and that's what we
do at Legion. Now you put yourself in a position
to know exactly how much labor you need, what skills
do you need, and how many people do you need
at what time. Once you figure that out, the next
piece of the puzzle is hey, employees have a say

(07:47):
in when and how much they want to work. Everybody
gets paid by the hour, and some employees want, for example,
a stable schedule because there were other things going on
in their lives. There may be part time students, or
they may have a second job. Other employees may be
trying to maximize their weekly pay by working as much
as possible because they have the time to do and

(08:09):
and they may and they may need the pay more.
So every employee is different. So if you want to
personalize the schedule for every employee, now you've got to
capture their preferences, and you've got to match the demand
from customers to your employees preferences. And you can do
all of these things automatically using AI. So that has

(08:30):
been the breakthrough where let's not look at just one
side of the equation. Let's bring in all these data
points and let's use AI to figure out what is
the optimal plan for both sides.

Speaker 4 (08:43):
I have a question, Yeah, I have a question, Like
this is all incredibly fascinating. It's not directly related to
kind of what I do on a daily basis, but
I am curious. This is clearly bround breaking. You didn't
wake up when you were like a ten year old
kid and say I would have built workforce management platform

(09:08):
using AI.

Speaker 5 (09:09):
How did you?

Speaker 4 (09:11):
How did you get to this point in your career
up with legions, well legend at this point, what's your
origin story?

Speaker 5 (09:19):
Yeah? You know it was absolutely right. It wasn't. One
day there was a lightning strike and as aid workforce
management that that's what did not happen. I was in
enterprise software for a long time. I was working with
you know, some really great companies like SAP. I was
chief for a officer there. I was with Ariba. There

(09:40):
was all areas of enterprise softwahich is not related to labor.
It was you know, supply to order management. So just different, different,
different categories. At one point I left those jobs at
a great you know, fifteen year run with a Riba
and SAP, and I was taken a lot of time off.
I was you know, I was kind of figure out

(10:00):
what I'm going to do next. And I live by
the way in the San Francisco Bay area, which is
very tech focused. So one of the things I did
which was you know, very very impactful for me, was
I em worked on this very long road trip. I
put the dogs in the car and I went through
like I'm going to spend next several months in national

(10:21):
parks and you know, drive across the country and things
like that. And that was very eye opening to be
out of certain valley's bubble, you know, it's the tech bubble,
just to see like you know, Heartland America and the
jobs and the small businesses and you know, all the
things that kind of, as I call them, real jobs.

(10:45):
And one of the things that I was just fascinated.
I was on the road for two months, and you know,
I was fascinated. Every place I go there's you know,
hiring signs outside like they're looking for that they just
can't have enough people. And it wasn't because there were
so many jobs. The turnover was so high in retail

(11:06):
and restaurants over one hundred percent. You're constantly hiring because
you cannot keep people. And I'm just you know, just
being curious about all these things. I just started talking
to a lot of people. I started talking to the
waiting staff, and I started talking to retail employees and
sales associates and cashiers, but also their managers. And what

(11:28):
I realized I started like a picture started forming, which is,
you know, employees are leaving these jobs for really basic reasons.
It's not something sophisticated. Yes, the pay is could be
better in all these jobs, no doubt about it, but
that's not why they're leaving. They're not leaving because they
get better pay somewhere else. In fact, pay is generally

(11:50):
normalized across you know, metropolitan areas. Restaurants pay the same
and retail pill. So if they're not leaving for pay,
why are they leaving over one hundred percent turnover? Typically, Well,
they are leaving because of experience. They are leaving because
of friction with their schedule. Like I'm working Thursday, Friday, Saturday.
I joined this job, but two months later my school

(12:13):
schedule changes. Now Thursday is no longer possible. Path of
least resistance is to leave this job and find another
job where I don't have to work Thursdays. These types
of things we're driving turnover, and I felt like as
a as an innovator, as a technologist, wow, we got
to help. I mean, this is a solvable problem. This

(12:35):
can be if you can bring in some real sort
of you know, high tech solution around decision making and
figure out what works for the employee. For the employees,
both sides should be in theory happy. Like that was
my you know, kind of naive simple thinking at that point.
What could be so hard about this? That's very hard
to do, but it is extremely impactful once you figure

(12:59):
that out. Once you build that, businesses really appreciate every
almost every manager I spoke to, they wanted to do
the right thing for the employees. It was just very
hard because every employee needs are changing every few weeks,
and it's the possibilities for them to be on top
of was solized. And they also have a business to run.

(13:19):
They have to take care of customers, they have to
take care of budgets, they have to take care of
the clients. So it's a very hard problem, but it's
the ideal problem for technology and particularly for air. So
that was a conclusion about nine years ago. Since then,
we are just focused on what is that middle ground,
and that middle ground. As we start focusing on that,
that middle ground starts looking bigger and bigger and bigger.

(13:40):
There's a lot of middle ground, as it turns out,
and that's that's where we are.

Speaker 1 (13:44):
Now, what are you okay? So Sonnish, when you look
at the AI landscape, the models that are out there
being used right now, you know we utilize it in
our business. I will say that AI has kept us
from hiring people. It's actually because of the fact that
we've integrated AI into so many of the tasks that
were existing jobs that have just made them more efficient

(14:08):
at the work that they do. Now, whether it's a developer, designer, editors,
you know, uniem IT researchers all are able to do
a lot more with you know, less people. Do you
find that we are going to see that kind of
move because we haven't necessarily seen that happen in the
restaurant industry just yet. It's still a very you know,

(14:30):
kind of elbow grease industry.

Speaker 5 (14:32):
That's a great question. So at least my hypothesis, and
you know, everybody at hypothesis at this point because the
world is changing so quickly, but AI, and it's really
hard to pint it down. My hypothesis is that service industry,
where the core skills are communication, empathy, service skills, treating

(14:54):
customers in a certain way, and hospitality. These are these
are I wouldn't say fully insulated by air disruption, but
they are largely you know that those are not the
areas that AI is advancing. AI is advancing on cognitive skills.
And so there is another set of job categories knowledge

(15:15):
workers like legal and things like that, including software, where
you are seeing a lot more disruption, where AI is
potential in service industries to drive efficiency and actually to
improve those jobs. In my opinion, For example, one of
the things that we are using generative AI now is
to drive conversational you know, interfaces with our software. So

(15:40):
almost all of our users of Legion, they are what
we call deskluss workers. They are not in front of
a computer setting, you know, learning software skills and things
like that, or even learning how the software works. They
would rather be out in front of customers or doing
their jobs. So for them to best use software like Legion, hey,

(16:03):
it's easier to say what you want and talk to
it and get your actions and answers done. Like for example,
a manager can say, oh, you know, I need one
more person this afternoon, seems like there are more people
in my restaurant. Lesion, go find go find an additional waiter.
Right great, that's an instruction. That instruction can be captured

(16:27):
by AI and could be translated into a scheduling action,
and that action is then basically relayed into the software executed.
Open shift offers are sent out to eligible employees, the
people who pick the with the right skills and the
right you know everything else, all the rules preserved. That

(16:48):
open shift is claimed, the schedule is automatically updated and
the person shows up Like that type of automation can
be driven by AI, but the person you still need
the manager, you still need the very stuff. It's not
a niche. Jobs are not going away.

Speaker 3 (17:03):
You know, you touched on something earlier that I think
is kind of the root cause analysis, you know, the
root cause for why something like legion really is required,
and that's retention. You know, our industry has a massive
retention problem. You know, something anywhere between seventy and eighty
percent uh turnover is what is the you know, the

(17:27):
typical stats that are thrown out there. So so it's
not just a scheduling situation. It's an engagement absolutely not right.
So you need to find the right balance between okay,
giving them empowering employees to not only have a say
in their schedule, but also create opportunities for engagement so

(17:52):
that way they feel like they're part of, you.

Speaker 1 (17:55):
Know, the something.

Speaker 5 (17:56):
I know.

Speaker 1 (17:57):
There's one thing out work. I think it hits your point, Paul,
is one thing on their website. I don't want to
kind of zone in this. You said give employees gig
like schedule flexibility. That goes back to your point, Paul
on that ability to kind of integrate. How can you
do that on I mean, I get it when it's
uber or it's Postmates where there's not fixed hours in

(18:21):
that and you don't have customer. Sure you have a
flow coming in, but the restaurant industry, I know, do
you feel like this is? How successful has this been
in this gig like model.

Speaker 5 (18:33):
It's been very successful. And I'll tell you and I
understand the question, which is, Hey, restaurants have constraints that
Uber doesn't. For example, operating hours it's open from ear
nine am to ten pm or.

Speaker 1 (18:46):
Whatever, the zero in on just lunch hour and zero in.

Speaker 5 (18:50):
On lunch hours and things like that right there are
So let me take a step back here. What do
employees want? Why do employees why? Why is there this
turnover problem? Let's define the employee experience problem in restaurants, right.
We do, by the way Legions since our founding days,
every year we do a state of hourly worker. You know,

(19:13):
there's a document. We do a lot of research, we
publish that. It's great data in there and you can
see it right there exactly so that consistently shows there
are three things that employees are looking for, and that's
the that's the primary reason for heighth turnover. One is
obviously scheduled flexibility, as we call it. You know, I

(19:34):
want stable hours, I want more hours, I want less hours,
and my needs may change from two week. That's number one,
and that's why the by far number eighty something percent
of people are making stay or go decisions based on schedule.
Fraction second is a modern experience. We're talking about millennials
gen zs that toalk generation, they are no longer coming

(19:57):
to the restaurant or any any place of work and
looking at a paper schedule pictures of it. All those
days are gone. You're now used to door dash and
Uber and all that those experiences where you're dealing with
an app. You're not negotiating everything with managers. You are
talking to an app and you are basically putting your
preferences in there and figuring out if you're going to

(20:19):
get the schedule based on those preferences. So we enable
very rich personalization of schedule. Employees on Legion can say, okay,
which days of the week, how many hours a week,
which times of the day, which locations. They can set
all their preferences and we will match them with work.
The third thing which has immerged very rapidly in the

(20:42):
past five years is instant pay right. That's another attribute
of gig like experience. After my shift is over, I
don't want to wait for a week or two weeks
for my paycheck. Just pay me right now, because that's
what Uber does, That's for door dashers, that's what everybody.
So these are the type of things that are very
very important these experienced drivers. Now now to answer a question, Paul,

(21:06):
we understand there are constraints, but even within the constraints,
employees sign up. They join restaurants knowing that this restaurant
is going to be open from you know, what their
operating hours are, and lunchtime is more important. But they
still want some flexibility on Hey, I really want to
avoid Mondays and Tuesdays because that's something else for me.
Or I really want to avoid Saturday evenings because that

(21:28):
is my you know, time with my with my children
at home or whatever it might be, and those things
when they can do that without having a negotiation. Every
single time go to an app and AI takes care
of them by saying, Okay, I'm going to personalize the
schedule to the extent possible for all all of you

(21:49):
and and try to try to make your experience better.
And that goes a long way.

Speaker 1 (21:54):
This episode is brought to you in part by Gusto,
the number one rated HR platform for payroll, benefits and more.
With Gusto's easy to use platform, you can empower your
people and push your business forward. Over four hundred thousand
businesses choose Gusto every day. Now let's get into it, guys.
There's a couple of things you can do with Gusto

(22:15):
that you should check out. Some of the solutions that
you're just absolutely going to want to know about is,
of course, their business type, new businesses and startups. You
guys are welcome coming in small businesses. Maybe you have
a mid sized business that needs an all in one
payroll and benefit program as well as HR all of
this scaling. The cool thing about this is it's an

(22:36):
all in one platform. They can also select and punch
in right to your accountants, So check it out. You've
got a Gusto Pro platform. You can become a partner
with them if you're an accountant, So if you have
a CPA already, this is the place for you. And
of course the best thing is pricing. The thing about
Gusto is flexible plans and features honest price, no hidden fees.

(23:01):
This is the plan that we use, which is the
plus plan sixty bucks a month. Guys, you cannot go wrong.
It's about nine bucks a person, so you guys can
definitely afford it. Get in there and choose Gusto with
a full suite of tax services, HR services, time tracking, scheduling, expenses, reimbursements.
You get the picture. Gusto is the place for you.

(23:23):
Check it all out. Just go over to Gusto dot
com use our link down below to get started. See
you there, You hit on something there. Paul and I
have talked about this quite a bit, and that is
this idea of instant pay. We have another technology platform
that focuses on blockchain. Blockchain and crypto and streaming payments

(23:46):
is now becoming a thing, at least on the creator side,
So streaming payments real time on the blockchain, those kind
of things. If we see platforms getting to the level
of streaming payments, instant pay, is that going to become
kind of one of the key things that starts to
draw workers, especially the gen Z and millennials to brands

(24:08):
that are doing this. Are you already starting to see
a separation of brands? Okay?

Speaker 5 (24:13):
Absolutely, I mean this is uh. So, our point of
view on this is that this is you know, conversion
towards becoming a core feature that is just required for
anybody having shift jobs in the future. The notion of
a payroll cycle for shift job does not make sense.

Speaker 1 (24:36):
How does the payroll companies integrate? How does Because ADP,
GUSTO those kind of companies, I haven't seen anything come
out of them for that, I know.

Speaker 5 (24:44):
SA.

Speaker 6 (24:44):
Yeah, so I think that there are companies that have
innovated around the traditional paycheck, right so, meaning so, so
what we do at Legion instant pay just to kind
describe how the feature works.

Speaker 5 (25:02):
Employees, the shift is over and we will match employees
clock in, clock out and automatically approve the timesheet because
that that is that is the proof that it did work.
It's earned a wage access technically speaking, so you need
to prove that wages are work was done and wages
are earned. Otherwise it's paid a loan and there's some
some some technicalities and regulation around that. So we take

(25:24):
care of that stuff. But after an employee clocks out
and the timeshet is automatically approved using our our AI
now UH, employees will will be allowed to drop up
to a certain percentage of that shift's pay instantly after

(25:44):
the pay and we leave the remainder for taxes and
garnishments on the paycheck day. So from the payroll standpoint,
they are still providing the same entire pay for the
employer the day of the paycheck, but we intercept the
portion that we've advanced, okay, and we take that part

(26:04):
and send the rest of the employee. So it works
out very well in the way that pyrol process doesn't
have to change. That means all these old processes that
have been laid out with like old school companies there,
they can stay as is while we run a workflow
on top of it.

Speaker 1 (26:19):
Yeah, tell me how this works. Then I'm going to
show you something on screen. This is on your website,
seventy nine percent reduction in premium pay. This is on
QSR that's nine and a half hours saved a week
on scheduling. This is on fast casual that is huge.
Explain that data.

Speaker 5 (26:40):
Yeah, so you know, once you start accurately fore casting
demand and once you start accurately scheduling people based on
demand and work preferences and things like that. Now restaurants
have a lot of you know, restaurants who don't do
this stuff have a lot of extra pay in their
labor cost Like they are paying for penalties, they're paying

(27:03):
for overtime, they're paying for all of these things because
they they are guessing when they were They're guessing not
just when guests would be in the restaurant, but they're
also guessing when employees want to work. So they're guessing
on both sides, and they just incur all these extra
costs no controls in place. So when you put these
controls in place, when you do demand for casting, when

(27:24):
you actually figure out when employees want to work, when
you factor in all these data points and budgets and
things like that, now all the pay extra pay that
you are you are incurring week after week on over
times and all these labor penalties that restaurant industry and
QS are industries in particular, they're highly regulated. With all
these rules, all those penalties and other things drops significantly.

(27:48):
And that's our data that for our customers it drops
step on average by seventy nine percent.

Speaker 2 (27:55):
So to NICHE.

Speaker 3 (27:57):
So for all of this to work right, in order
to get these demand forecasts to be accurate, this is
going to require a tremendous amount of integration, right, you need,
you're gonna need to plug in with other you know,
you have to play nicely with others in the tech stack.
So typically, what does that look like?

Speaker 5 (28:15):
You know?

Speaker 2 (28:16):
How how are you fitting in?

Speaker 3 (28:17):
Are you are you plugging into the POS You're plugging
into payroll accounting, you know, and so on and so forth.

Speaker 2 (28:24):
How does that work?

Speaker 5 (28:26):
You precisely mentioned the three things we plug into, which
is the point of sale, the HR for employee records,
and the pay for sending timesheet from a demand forecast standpoint,
which is the.

Speaker 3 (28:40):
Data things like weather and you know, like Stacey has
to deal all the time with uh, you know at
the store level, you know what's going on in the community,
what's happening?

Speaker 2 (28:51):
How you know, how.

Speaker 3 (28:52):
Can she schedule appropriately with with those outside factors?

Speaker 5 (28:56):
Yeah? So so yes, one of the things which is
our special set legion, we not only integrate with with
point of sales, so we get historical data from at
every location level. Right, So if you've got let's say
QUSR with five hundred locations, each location has its own
data feed and a model is trained for each location

(29:17):
very uniquely because every location is different, even though you
may be selling the same things, but the factors that
drive traffic, for example, could be very different from location
to location.

Speaker 1 (29:27):
So we have hey, general, but how long does it
take you to build that model? You know, because you're yeah,
you're pulling in data right from each location you're mentioning.

Speaker 5 (29:38):
Okay, So during implementation of Legion, when we're working with
our customers to go live and all that stuff, each
model will take about, you know, generally about two to
four hours for the first time training, right, So it
takes it takes a couple of hours to train every location.
We do all this in parallel, and we the in
clouds so we can we can our compute can be

(29:59):
searched and all that stuff. So overall the whole process
would would let's say, take several hours. But that's the
first time. And during that time, not only are we're
training it with customers historical data, but to Paul's point,
we also bring from our data syndication partners, weather data,
local events, holidays, factors, school calendars, all of these things.

(30:19):
Because all these are demand influencers, they can influence what
is your is it driving traffic or taking away traffic?
Sometimes there will be an event center right next to uh,
you know, a place of business, and that could actually
take away traffic because these are you know, football games
or whatever, but during certain times, the traffic search before
and after a game. So we we we capture all

(30:43):
these data and we train these models. So that's one
very unique thing. Every location, every data set gets its
own model. But the other unique thing that we do
is we retrain this model every week because now new
data comes in. There is a one week of new
data that that is added for every single location, and
that's very important to do because over a period of time,

(31:05):
your traffic patterns may change. For example, in front of
your restaurant, there may be a construction and that's that's
going on for several weeks, and now suddenly your traffic
pattern looks very different because people can impart your cars.
How do you how do you continuously learning and continuously
get better. That is basically the retaining plant. The retaining

(31:25):
process is very very fast because it's like one week
of extra data that is added, so that happens in minutes.
But the models are trained initially with several hours with
all the historical data and patterns and things like that,
and they're ready to go and then every week we
retrain those models. So today in Legion's production infrastructure in

(31:45):
our cloud, we train about over three hundred thousand models
every week. That's basically at AI infrastructure at scale, and
that is what's needed. That's the type of infrastructure that
needed to keep these accuracy of these predictive models really shark.

Speaker 1 (32:02):
All right, So kind of one hundred million dollar question
for you is the AI the models. We've been using
AI in our own business for eighteen months now, I
can tell you this about every six months we see
a significant level up in terms of its tool set,
what we're able to do with it. Now, the potential

(32:24):
here in twenty four months, thirty six months, why would
we not see general open AI type models where a
restaurant could just come in and build this on their own,
just say, hey, here's my data of my employees, my
labor from my ADP or wherever. Here's my sales numbers,

(32:44):
which gives me all the data by time stamping. Here's
all of my HR information, create me a scheduling platform.
Is there any risk of AI solving this problem to
the general public.

Speaker 5 (32:58):
Yeah, that's a great question. So we divide AI into
three three buckets of problems, right, And there are different
types of AI, right. One is the predictive AI, all
the predictive problems. You're trying to predict something based on
the past. That's machine learning deep learning. That's where it's
a very different type of models than the genitive AI

(33:19):
models that these lms like open AI uses, these are
very different models. Those models are fine tuned for time
series calculations and things like that, which is really doing
the predictive math, and generative AI models cannot handle that
really well. You have to build very very specific models

(33:39):
to do prediction right. The second type of model is
AI models are what we call optimization models, where when
you generate a schedule, for example, if there are five
inputs and there is only one right answer, every single
time you run the five inputs, you should get the
same answer. Genitive AI does not. These are not deterministic models.

(34:02):
You could get different answer a different times, right, So
those models are not great for optimization problem like schedule optimization.
Generative AI is great for conversational interfaces, for taking a
large body of data for example UH employee handbooks, figuring

(34:23):
out the policy if training employees for what is the
right policy for filing time of request. Well, you can
get maybe one a month or two a month. Then't
have to read the employee handbook. They can just ask
our agent what is the right policy and we will.
So those types of taking large bodies of text and

(34:43):
uh and you know, turning that into information that is
very helpful at the right time, but also enabling conversation
back and forth. That is that's the applicative genitive AI.
So we have in Legion. We use all three types
of AAR for different couple of statements. And it's it's
extremely complicated for anybody to just build this on their own,

(35:07):
especially with generative AI and trying to do that for
predicuve problems.

Speaker 1 (35:11):
Okay, all right, so what about the next generation of
workers gen Z's coming. We're seeing now very small use
cases on a gentic AI I work. We invested in
a small company that's doing a gentic AI right now
in the food service space. They've plugged into door dash
APIs Now you can use voice control to basically get

(35:31):
an order done. They are telling me, you know, within
their group. I don't think I'm spilling the beans here,
but they're telling me that the potential for a gentic
AI is eventually this next generation, especially gen Z, is
going to be able to just tell whatever device they're
taking using headset, eyeglasses, phone, whatever it might be, find

(35:54):
me the best gigwork this week, find me the best
brand based on my schedule. You know, it's basically going
to be a Q a Q master or what are
they are? Prompt master. These these kids are going to
be prompt jockeys, you know, and figure out how to
ask these questions to get the best result because AI
is going to be talking to other AI. That's I

(36:15):
guess when you get into how does this future layout
for jobs in the future.

Speaker 5 (36:22):
I think there are two points you made and I
want to touch on both of them. First is the
way to engage with software is through conversations, like just saying, prompting, talking, speaking,
all those things. We strongly believe that is going to
be the future for workfast management too. Everything else is
in the back end, like this traditional user interface where

(36:45):
you go through menus and mobile apps, and yeah, it's
all going to be an exception, not the norm. The
norm is going to be, Hey, next week, I want
to for not to work on Thursday and done, and
now something will take that into struction and we try
to do something else with it. So that type of
way to engage with software is absolutely the future in

(37:08):
my humble opinion. Uh and and and that also is
great for everybody because the the the bar of learning
software and I mean everybody knows how to waste it's
a way of time exactly. So so that's one part.
The second point you made is agents talking to other agents, right,
So it's like and that is also we're already seeing that.

(37:30):
By the way, we are working with a lot of
our our our customers. You noted, we are working with
some great, great large companies. They are building their own AI.
They are building internal AI platforms for their own employees,
and they are coming to and saying, hey, we want
our AI to interoperate with Legions AI. We want our

(37:51):
if if I'm a if I'm a regional vice president
that that runs a chain of one hundred restaurants in
the Western region, I don't want this regional vice president
to go to your software necessarily. I want them to
stay with our software that we've built. But ask questions
about Legion and your agent should answer that question. So

(38:13):
let's figure out how we're going to how these agents
are going to interoperate so that these things become possible.
So that is absolutely happening. And there are there are
protocols that are emerging like MCP and A two A
and things like that, which are all facilitating these types
of discovery and collaboration between AI agents. And and I'm

(38:33):
a strong believer in that.

Speaker 4 (38:35):
Cneche. I have another question that's gonna uh normally would
come from our other co hosts, really, but it'll come
from me.

Speaker 1 (38:44):
Oh boy, here we go.

Speaker 4 (38:45):
So here's my question. So we're now taking in you know,
we've acknowledged that much of the work workforce does not
want to have the conversation to say I need this
day off and it's I will job hop if there's
any sort of friction. So is your technology improving workplace culture?

(39:08):
Because it really sounds like you're taking kind of all
the things that my eighteen year old self would have
had an issue with at the sushi bar that I
worked at. So are you seeing from your clients that
they're seeing Obviously it's affecting turnover, but while people are

(39:29):
in the workplace, is it improving internal culture? Like less stress?
Sounds like it's me.

Speaker 5 (39:37):
That's a great question. And let's look at it from
the perspective of the manager.

Speaker 1 (39:42):
Right.

Speaker 5 (39:43):
So the manager, let's say the manager of a restaurant
or a retail store. Let's say there are twenty twenty
five people. On an average, they get one point two
requests per employ per week. That is a stack, right,
So the request maybe not necessarily for a schedule change.
It could be for, you know, hey, can I can

(40:04):
I leave thirty minutes early today? Or or have you
looked at can you approve my time sheet? Because you
know whatever, they take yeah, time off all of these
employee requests. So managers are overwhelmed by all this incoming
from employees, and they even if they have best of intentions,

(40:25):
they revert into this behavior of saying just no, I
don't have time for that. Right. That is that is
a problem that automation fixes, and we've seen the improvement
between the manager and employee relationships because almost all employees
want The number three reason if you look at our data,

(40:46):
why why employees leave? They feel their managers don't like them.

Speaker 1 (40:49):
Hey, I can be the bad guy.

Speaker 5 (40:52):
Hey, I can be the bad guy. And also I
think that's a good guy.

Speaker 2 (40:56):
Right, I can eliminate it that problem.

Speaker 5 (40:58):
Yeah, exactly. So manager's job is a lot better now
they are just saying, Okay, put your question, legion, and
if they if it's possible that you'll get that shift,
you'll get that shift, right. I don't need to intervene it.
I don't need to say no or yes, or or
take time away from the things that I'm going. So overall,
the equation between employees and managers works out a lot

(41:19):
better in almost every single case. And managers learn once
managers learn to trust the system, which we go out
of our way to include a few things in our
user experience so that they are there's full transparency. But
once they learn to trust the system, after that it
becomes hey, this is still the software that's helping both

(41:42):
of us. It'll be more efficient and that really helps
with the culture.

Speaker 1 (41:49):
Rudy would be proud of you, Stacy.

Speaker 4 (41:51):
Thank you. I mean, that's a great back yesterday. But
I did include a question that was ready really asked.
So see. The other thing I'd love to know is
I play in the emerging brand space, so most brands
are less than fifteen units. I have one brand that's
one hundred and twelve at the end of the year,

(42:12):
it'll be one hundred and fifty units. But my question
is this sounds like a wonderful piece of technology that
can help brands a lot, But at what point are
you ready for it? And what point is it affordable
for a brand? You have to be of a certain size.

(42:33):
Can it independent operator use this?

Speaker 5 (42:36):
Yeah?

Speaker 2 (42:36):
Who's your ICP?

Speaker 5 (42:38):
Yeah? Our ICP is so generally we work with businesses
that have more than three thousand employees, right, because that's
where the need for efficiency and automation really starts making
a big difference. Before that, we have seen it's still important,
but you know, at smaller scale, customers may use spreadsheets

(43:02):
and just be paying more attention. But once you start scaling, uh,
the labor costs, you have to figure out how to
you know, have a playbook that that guarantees cost control
on labor and turnover and things like that. Legion become
that playbook.

Speaker 4 (43:17):
And do you do you also do you work with
both corporately owned and franchise systems?

Speaker 5 (43:23):
Yes?

Speaker 1 (43:24):
Cool?

Speaker 5 (43:24):
Cool?

Speaker 1 (43:25):
Well, all right, guys, we got we're going to cut
the short because we're running along. But the key thing
is we've learned a lot about this today so Sniche,
we may get you in here as our AI expert
because you you gave some good answers and we threw
some hard questions at you, I think as well, so.

Speaker 5 (43:43):
Great questions.

Speaker 1 (43:44):
Kudos to you for being able to be ready for
that kind of stuff because I'm sure you have to
pitch this a lot, you know, and everybody is always skeptical,
you know of how this is going to rk. So yeah, anyway,
for you guys are going to learn a little bit
more about uh what Legion is doing. Just jump over
to their website which is Legion dot co. You guys
can find them over there. And Soniche, thank you so

(44:07):
much for coming in today. We appreciate it.

Speaker 5 (44:09):
Thanks thanks for having me you.

Speaker 1 (44:10):
Bet, you bet excellent stuff. All right, Uh, masterminds, we
have seen another another amazing episode here. AI is going
to be a big part of the show going forward
because we have so many brands that are starting to
test use it and now we have you know, solutions
out there breaking out. What do you guys think about this?

Speaker 3 (44:31):
Well, first and foremost, you know a lot of times
when we throw AI around, UH, particularly with UH inside
a restaurant tech, it tends to be more of a
feature right to be then it then it is rather
a full fledged solution. And I think what we just
learned today was here we have a full fledged UH

(44:51):
solution that is, you know, AI found that is foundational
in the different facets of AI like Suniche described, whereas
I think a lot of other tech companies might look
at AI as being an enhancement. And I think, and Stacey,
I really want to get your I'd love to hear

(45:12):
your opinion, particularly from the brand perspective. You know, right
now is a really crazy time in our industry where
there seems to be a lot of wait and see.
You know, there's a lot of this uncertainty absolutely, you know,
so does does AI?

Speaker 5 (45:25):
Stacy?

Speaker 3 (45:28):
Are people just kind of like waiting to see like
how this is all going to like play out or well?

Speaker 4 (45:33):
I think, you know, it's hard for me to make
a big generalization like that because of the size of
the brands that I'm I'm playing with. Yeah, you know,
we're using it in like starts and spurts, so your
hypothecy like we're waiting and seeing like what it can
do for us. I think until it's more ubiquitous throughout

(45:57):
the industry. My brands can't afford a lot of the tools.

Speaker 1 (46:03):
It's going to get cheaper, you know, it's going to
get more more inexpensive to deploy some of this stuff,
you know, over the next year.

Speaker 4 (46:10):
But I think everybody's excitedly waiting and that this is
going to be you know, complete, a complete metamorphosis for
the way that we do things.

Speaker 1 (46:21):
Have you started using it in your own you know,
marketing teams?

Speaker 4 (46:27):
Yeah, I mean we're I think there's some marketers who
might be a little bit afraid for their not marketers,
kind of strategic marketers, but people who are actually producing things.
So you know, you're you're creatives. And I keep going
back to this this quote I heard like a year

(46:50):
and a half ago that like AI is not going
to replace people, but people who use AI are going
to replace people who can't use A. So we're trying
to get our teams, you know, we're trying to get
our teams up to speed on using it and using
it for efficiency purposes. I just used it in an

(47:16):
predictably in an in an HR capacity. Actually, we took
all of the attributes of the candidates and kind of
got's the best, Like this is what we want because
you know, it's learning what our businesses. So that has
been enormously helpful. You know, all of the brands are

(47:40):
using it for things like copywriting and and actually and inspiration.
So if I have a small brand and we don't
have a full time designer on and I have an
idea in my head, I could say I want a
an image of a guy eating corn chowder and he's

(48:08):
a gen z who loves I don't know.

Speaker 1 (48:13):
Yeah, but and the and but I think you're you're
hitting it. And that is is that the model is
developing quickly. It's kind of like in the early days
of social most of the brands didn't know really what
it was going to be and how they would end
up using it and driving sales with it, getting new
customers and employees. So I think it's very early stage,
but it is accelerating faster.

Speaker 3 (48:34):
It's you know, it's an early stage, but it's at
the same time, there are a lot of forward thinking
brands out there that are utilizing AI tech solutions to great,
great advantage. I mean, ultimately it's an ROI story. And
I think if you look at Legion and some of
the brands that they have and you look at you know,
you do a little bit of a deeper dive and

(48:55):
understand how they're saving UH brands money. And I'm sure
Soniche is going in with some kind of you know,
ROI modeling and payback you know when is you know,
I'm sure there's a very compelling case for this, particularly
when you know what thirty percent on average is what

(49:16):
you're paying for your cost of labor. It's you know,
it's a there's a monumental amount of money that goes
into well.

Speaker 1 (49:25):
The percentage was staggering. I mean when you looked at
the QSR and the fast casual data that he had
on his site. Uh, if you were to calculate that
in a you know, a fifty or one hundred story unit,
you guys, you're going to have that stacy with By
the way, which brand is going from one hundred to
one fifty?

Speaker 2 (49:42):
That's Ziggi's right, Is that Ziggies.

Speaker 5 (49:45):
With a bullet?

Speaker 1 (49:46):
Yeah, with a bullet.

Speaker 2 (49:48):
A second, Paul, because you know, let's just talk about
this for a second. If you have labor.

Speaker 3 (49:55):
That's what I said earlier about companies that there are
AI solutions and then there are AI features. I think
when you look at these full fledged AI solutions, they
have an eye towards future proofing.

Speaker 1 (50:10):
Yeah right, this is exactly.

Speaker 3 (50:12):
Focused on AI, and there's no there should be greater comfort,
I think with across restaurant brands when you're working with
a company that is AI.

Speaker 1 (50:21):
Based, Yeah, yeahs building as opposed to someone who's bolted
it on.

Speaker 2 (50:25):
Yeah, bolting on a I get it. They're an AI business.

Speaker 3 (50:30):
This is what they do, this is what they and
so they're going to grow along with it.

Speaker 2 (50:34):
So I have to imagine this more comfort.

Speaker 1 (50:36):
It's going to be crazy, man, going to be crazy. Listen,
you guys are going to love a lot of our
lineup the rest of the year. We are lining up
some big rock stars coming in. You're on, of course,
the Masterminds podcast, and if you are not catching some
of our other podcasts, you can do that just by
jumping over to savor fm. Catching us over there on
the website. If you are listening to the audio podcast,

(50:57):
you can find us on YouTube, which is just save fm.
Search us over there. Over a quarter of a million
subs now on YouTube, big numbers out there. Subscribe now
if you have not Masterminds, it's gonna be a good
What are you eating there?

Speaker 2 (51:14):
I haven't ate yet. I'm sorry.

Speaker 1 (51:17):
To the end of the podcast, it really happens every time,
every time. All right, guys, we'll catch you next week
right here on the Restaurant Masterminds podcast

All Episodes

Episode Transcript

Popular Podcasts

Stuff You Should Know

Crime Junkie

The Breakfast Club

.css-15opob5{left:0;position:absolute;top:0.8rem;} All Episodes

.css-14f5ked{margin:0;word-break:break-word;display:-webkit-box;-webkit-box-orient:vertical;box-orient:vertical;-webkit-line-clamp:2;overflow:hidden;}Building an AI Integrated Workforce

Episode Transcript

Popular Podcasts

.css-r6mb8g{margin:0;word-break:break-word;display:-webkit-box;-webkit-box-orient:vertical;box-orient:vertical;-webkit-line-clamp:1;overflow:hidden;}Stuff You Should Know

Crime Junkie

The Breakfast Club

All Episodes

Building an AI Integrated Workforce

Stuff You Should Know