Marvin Martinez & Bandsaw.ai: Practical AI for Real ROI

Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
SPEAKER_01 (00:00):
Hello and welcome to Bald Ambition.
I'm your bald host, MookieSpitz, and the one with ambition
today is Mr.
Marvin Martinez.
He is the founder of Band SongAI.
Welcome to the podcast.

SPEAKER_00 (00:19):
Thank you for having me.
It's a pleasure to be here andsharing with your audience.
It's a pleasure having you.

SPEAKER_01 (00:25):
AI, AI, AI.
Everybody's talking AI.
It sounds almost like a cliche,but it's not.
It's a true revolution.
And startups like yours arepopping up everywhere we look.
You seem to have a uniqueapproach to things, a cool value

(00:45):
proposition.
You want to tell us a little bitabout yourself, your startup,
how you got it going, and uhwhat you have going on.

SPEAKER_00 (00:56):
Absolutely.
I have been in the operationsworld for over 13 years in my
previous jobs.
I was the guy in charge ofmaking sure that everything in
the company was running fine.
Everything from hiring the newpeople, training them, quality
programs, production things, HR.

(01:17):
So I have the whole holisticlogistic view of an entire
business.
When the AI started to take ahype, I realized that a lot of
things that I did in thebusiness world could be done
with AI, right?
So I could significantly savetime.
And I see a lot of businessowners, people I know that they

(01:39):
struggle with doing thingsmanually.
So I decided to find uh to youknow start the company Banso AI
because I had the best of bothworlds.
Number one, understanding thebusiness operations and how to
map processes, because I was incharge of doing SOPs, uh, you
know, going into the drawingboard, uh setting up the steps

(02:02):
and everything, and thenlearning a lot about AI to put
AI in that process.
And right now, what I do in mycompany, AdvancedOAI, is helping
other business owners fromservice companies to save time
and money uh understanding whereAI takes place or can fit in

(02:24):
their day-to-day operations.
And we do that only in 10 daysbecause we focus on
understanding your processfirst, help you to map it, and
up to that point add AI to it.

SPEAKER_01 (02:38):
Great.
And in terms of your expertise,what size operation were you
responsible for?
So your core expertise in small,medium-sized businesses.
What what was your your your 13years in operations?

SPEAKER_00 (02:52):
What kind of I started with small companies uh
doing uh anywhere between 10 to50 employees, but one of my uh
most senior roles, I was incharge of up to 700 people in a
call center environment.
So that was a very, very huge uhlearning experience.

SPEAKER_01 (03:14):
Okay, so call center, you got a lot of
information, there's a lot ofdata, you got personnel, you
gotta handle uh you knoweverything from the
conversational reality, customerservice, all that, and then they
have a lot of tech plugins tomake it all happen.
And now I noticed that you relyon tech plugins as a way of

(03:36):
integrating the capabilities ofother applications, and then
you're adding what seems to mean AI layer to make that support
better and more efficient.
Can you tell us a little bitabout that, like relying on uh
Slack and many of the otherprograms that are essentially
softwares as a service to boostthe benefit you're providing

(04:00):
your clients?

SPEAKER_00 (04:02):
This is key.
Uh, there's a lot of people thatwhen they think about AI, they
think about trying the newesttools.
And uh with the way that AI hasmade it easier for people to
create new things and software,you will see a lot of tools out
there.
So most of the people I speakwith, uh business owners, they
believe that to implement AI,you have to bring all these new

(04:26):
things code, anti-gravity,Gemini, whatever, right?
Uh, but that's not the case.
They can continue to use uhtheir regular software, Slack or
uh Google Chat or their CRMs,you know, they can use
Salesforce, Hopspot, and uh it'seasy to integrate all those

(04:48):
tools uh together using toolssuch as NARN, Mate.com, Sapier.
On my side, my recommendation isdoing NARN to do all the
orchestration and connectionbecause that's where I noticed
that most people are losingmoney.
They have information in thisCRM, they have information in
this other system, they have theSOP in a third place, and all of

(05:12):
that is not connected.
So, of course, they lose money.
So, my job is to help themunderstand how everything is
connected and understand whereAI can make a difference there.
So, if you're thinking that toimplement AI workflows or all
that type of things that you uhlisten to in social media, that

(05:32):
you have to create this wholenew tech stack, that is not the
case.
You can do it with the sametools that you rely on every
day.

SPEAKER_01 (05:41):
So it sounds like you're solving some of the
interoperability problemsbetween incompatible or
redundant systems, you'recentralizing information and
workflow, and it also soundslike you're putting an
application on top of multipleapplications.
So you're creating an AI userinterface which could optimize

(06:03):
the efficiencies of theorganization for various touch
points throughout the operation.
Is that right?

SPEAKER_00 (06:09):
That sounds about right, correct.

SPEAKER_01 (06:12):
How do you operationalize this?
You mentioned you snap yourfingers, you do this in 10 days.
Uh that's a bold statement.
So I'm assuming workflow mappingis job number one to understand
the business, the touch points,the challenges at each touch
point, and the overall dataflow, understanding the

(06:35):
infrastructure that they have,the legacy applications they're
already using, and then you needto plug it all in and provide
your own UI.
Is that kind of how this shakesout?

SPEAKER_00 (06:48):
That is exactly how it works, and that is part of my
framework.
And the first step, when I jumpinto a company, my first
question that I ask is show mewhere your process is.
I speak uh generally with theowner or the founder of the
company because they have thevision of where they want to see
themselves.

(07:09):
And then I discuss with themhow, you know, like how much of
AI do they need.
And then I have a secondconversation with the person
that makes the operationsbecause they know the process.
Uh, most of the time, businessowners they know what they do,
they know their offer, but theydon't really know a hundred
percent the process, right?
So when I when I get in, I speakwith the operator, and then we

(07:32):
start uh, you know, like writingdown a checklist of the tools
they use and the currentprocess.
I use uh like a drawing mappingtool, uh usually mirror uh
board, and then just put it outthere for them to see it.
When they can see it, you know,that's a that's an aha moment,

(07:53):
like that's a wow moment saying,hey, we didn't know it was that
simple, right?
Because most of the processeslives in their heads.
So that's step number one.
Then step number two is how muchtime are you spending doing this
right now?
Because one key thing inimplementing AI is understanding

(08:13):
your before artifact state andafter the AI implementation so
that you can see if there'sactually ROI.
You will see a lot of companiesjust investing uh money in AI
tools, but they don't know ifthey are getting money back from
that usage.
So after we map the process,after we understand step one,

(08:34):
step two, step three, step four,we decide where AI makes sense,
right?
And and we start from there uhusing their legacy tools, uh
using uh the tools that theyhave in place.
And I can give you examples.
Uh, like one of the mostsuccessful use cases or
companies that I worked with uhwas these transportation and

(08:55):
logistics companies.
Uh they had 46 drivers, 46vehicles, and they did about
eight to 10 trips per vehicleevery day.
So that was a huge amount ofdata.
They had one dedicated personthat spent three hours just
putting that information fromtheir CRM into a Google sheet.

(09:16):
And when you think of this, youcan say to yourself, wow, is
that even possible?
But they were spending threehours every day.
That was 15 hours per week, 60hours per month, right?
If you do that and you do a rateof 25 to$30 an hour for that
person, you do the math.
That's how much money they wereinvesting.

(09:38):
And when we did the automationfor them, we mapped the process,
we put it together.
AI was only 10% of the entireprocess.
Only 10%.
The rest was just connecting thetools, and it was like an
automation with six steps.
Very easy.
After we implemented that, wecame from three hours in that

(09:59):
manual process to only 15minutes per day.
So we gave them back two hours,45 minutes per day.
That's a huge, huge ROI.
And the guys were really happy.

SPEAKER_01 (10:12):
And that's just one touch point and one task.
And an organization has many,many of these nodes that you
could progressively improve on.
So I'm assuming you take acrawl, walk, run approach, you
build trust.
Look, I saved you X hours and Ydollars in this one role, and

(10:33):
then you could scale up fromthat.
How does that before and afterstate look like?
So before you've got a bunch ofpeople using a bunch of
different software, could bespreadsheets, could be Google
Docs, could be Slack tocommunicate.
It goes on and on and on, right?
All the different applications,all the different companies,

(10:56):
from Adobe to Google towhomever, right?
So when you plug in bandsaw AI,uh, let's just hypothetically
think that you can transform anorganization from this
fragmented infrastructure to amore integrated AI optimized

(11:16):
organization.
What does that look like interms of their experience with
your application?
Do you have a dashboard?
Do you have a central UI?
How does this shake out so wecould begin with the end in
mind?

SPEAKER_00 (11:34):
Typically, that really depends on the scope of
work and what the personactually needs, right?
Uh, like for this particularcompany, for instance, the only
thing they were interested in issaving money.
But for other people, they wantto see how many customers they
uh, you know, they reached outusing AI, how much time uh did

(11:57):
they save by doing an AI-drivenaction.
And let me give you anotherexample.
I built a customer service AIbot uh in a website for this
stapping company, right?
So they they knew they weregetting visitors, but visitors
they would just look at the pageand they were not getting any
leads from that, at least notthat they were aware of, right?

(12:20):
So uh we created the AI chatbot,we created a database with all
the knowledge of the company,and that chatbot was in charge
of two things.
Number one, ensuring that theperson received a response for
any questions they had, andnumber two, pre-qualified them.
So the conversation in the chat,right, uh, would qualify the

(12:43):
people that was visiting thewebsite.
If it was a qualified person,then it would send the
information to a salesperson foryou know additional follow-up.
If not, they just put it in adatabase uh for the tracking
purposes.
But here's the interesting partthe person told me, or the owner
of the company told me, hey, howdo I know how many people we are

(13:04):
talking to every day?
And how do we know that thequality of the answers that we
are providing are to uh or meetour expectations?
So at that given point, wedecided on what he needed.
And even though we didn't builda dashboard for him, we built an
automated report every day at 8a.m.

(13:26):
in the morning where he got howmany people the bot talked to
the previous day, and and theanswers with the most common
questions people had so theycould work on that and make it
better, right?
So this was not a dashboard, uh,this it was a report, but it
really depends on the person andand what do they really want to

(13:48):
track, right?
But we can build we can build uhGmail reports or email reports,
we can just uh track ours, wecan create a dashboard, like a
like a front-end uh dashboardfor people to look at the
numbers as the automation goes.
We can track how many executionswe do per day.
Um, we advanced AI really wantto make it about you because we

(14:12):
care about you, right?
So depending on your needs, uhwe can do either or uh based on
your requirements.

SPEAKER_01 (14:20):
So it sounds like you're offering a custom
personalized AI-drivenoperational solution to small to
medium-sized businesses.
And to your point, it's notcompletely transformative.
You don't rip the IT guts outand put in a large language
model.
You respond to specific needsthat your clients have.

(14:42):
So, one example you cited wasjust operationally heightened
efficiencies for one workerhandling deliveries, and in
another instance, you're able toembed a chatbot from a customer
service perspective.
And then you mentioned puttingtogether a dashboard that fire

(15:02):
hoses in data from multiplesources to give them a
visualization about theinformation coming in and maybe
analyzing it.
So, what kind of tools do youuse in your in your
armamentarium, in your in yourquiver?
So, do you have a proprietaryLLM or large language model that

(15:25):
you could tie into?
Do you use agentic technologyfor some of the operational
tasks?
Uh, can we look under the hood alittle bit of Bandsaw AI to see
what kind of capabilities youcould bring to clients?

SPEAKER_00 (15:40):
Yes, I like to keep it simple.
Uh, one of the key things thatwe do differently is we want to
make it simple for you.
We reduce fraction to theminimum.
We're the gonna we're not gonnaask you to buy six, seven
different types of tools andsubscriptions, not really.
We use two main tools.
Number one, NADN, which is theorchestrator, uh, and it's the

(16:04):
way that we connect every othertool.
Uh, we need to access youremail, we use NADN to connect
your Gmail account or Outlookaccount or or whatever, right?
Um, we do you want to connectyour Slack?
We do it through N8N.
Do you want to do Google Sheets?
We do it through N8N.
Uh, you want to connectCalendly, your calendar, we do
it through NADN.

(16:25):
So NADN is uh one of our maintools.
That's one.
And we also use lovable.dev forsimple MVPs or front-end
applications so that people canactually see the automation in
action.
And they can change or providefeedback really quick.
Because we make it simple toyou, we deliver in 10 days, and

(16:48):
we believe because this thiswould be a question that a lot
of people may have.
Where do I even start?
Like, what should I automate?
Right?
I know that we can map theprocess, we can speak with the
operator, but but what processshould I automate first?
Right?
So that's another thing that wedo is helping you understand
that you have to start simple.

(17:08):
That's why we take 10 days.
We recommend the simplest or theuh low-hanging fruit where you
can get wins really fast, whereyou can see ROI, and then just
moving moving on from there.
So to answer your question, N8Nlow level.dev, just to keep it
simple, the LLM system cloud oruh chat GPT or whatever, like we

(17:33):
can decide that later dependingon the complexity of the task.
Uh, but you know, we use mainlytwo tools, and we start with the
simplest task so that you cansee ROI really quick.

SPEAKER_01 (17:44):
So, just in summary, just like you're pulling in
Slack and Google information,these are proprietary
third-party applications of somesophistication.
They lack, however, the deeplearning and what we would
consider AI.
Although Google now isincreasingly plugged in and
they're all kind of adding thesefunctionalities.

(18:07):
But at their core, they're moretraditional applications and
they're scattered all over theorganization.
So you use NAN, you use Lava,which is an intermediary
application, to integrate theseapplications in a way that could
be functionally useful for yourclient.

(18:29):
So you're not building anysoftware here.
It's almost like you have Legopieces and you talk to the
client, which is, hey, you havea business and let's map your
workflow.
Step one, let's identify touchpoints, then let's identify pain
points where you havebottlenecks and inefficiencies,

(18:51):
and then you go back and thenyou take all the Lego pieces,
and then you you take themapart, you put them back
together, and then you have afew connector pieces, right?
Which is like the NAN and Lava,and you're able to bridge them
and do your magic, and then youhave a reassembled Lego set of

(19:13):
IT, which is optimized for yourclient.
So your real expertise iscertainly on the technical side,
but it sounds to me like it'svery much on the operational
customer service side, which isa lot of clients don't even know
what their workflow is, it'sjust a big mess.
They have a whole bunch ofapplications all over the place,

(19:35):
and they never even reallythought about connecting them or
combining them or optimizingthem.
And you are the expert who comesin and puts it all, takes it
apart, and then you put it backtogether again.
Is that more or less correct?

SPEAKER_00 (19:49):
That sounds a hundred percent correct.
And again, uh, this is whatmakes us different.
We we want you to use AI wiselyin a way that makes sense to you
and that pays off yourinvestment of time and money.
Uh, we're not here for the hype.
We know how to, we we know thetechnical piece.

(20:10):
And you know, interestingly, thetechnical piece is the less
complex.
So people may think, hey, I haveto study a hundred days before I
can start implementing AI.
Well, what I tell people islet's work on getting your
workflow ready first, and thenthe AI piece is gonna be really

(20:32):
simple to follow.
I can give you another example.
I talked to this locksmith uhcompany in the US and the guy
was telling me, you know, everytime I am doing a job in the
field, if people call me, Icannot answer their phone call.
They go to my voicemail, andmaybe I forget at the end or I

(20:55):
don't reply back soon enough,and then those people are going
away.
Right?
So obviously what you want to dois to you know like avoid that
missed call.
I ask him, okay, how much moneydo you lose per customer?
He said$150 per service call.

unknown (21:14):
Okay.

SPEAKER_00 (21:14):
Then I asked him how many calls do you lose or do you
miss per day per week?
He said maybe one or two.
Right.
So think about this how muchmoney is this guy really losing?
So it's$150 per service call feeonce a week.
So that's$600 per month.
When you extrapolate that,that's about$7,200 per year in

(21:39):
something soft simple, right?
So I asked the guy, what wouldyou like to happen instead?
I didn't talk about AI.
So he said, I would like uhmaybe someone that can answer
the phone call, collect theinformation, and send me the
information.
That's all I need.
The guy didn't want a fancyrobot, the guy didn't Want, you

(22:00):
know, like the latest AItechnology.
The only thing that I want, hesaid, is take care of the
customer's phone call, collecttheir information, and then have
that information with me so thatI can call them back.
I don't want AI to call themback.
I want to do it myself.
Well, guess what?
We built an AI voice receptionsystem that would trigger only

(22:24):
when he misses the call.
This is not gonna trigger afterevery call.
So when he misses the phonecall, the call gets routed to
the AI receptionist.
The AI receptionist has a littlescript.
It's not fancy.
The call is about one minutelong.
It's not even crazy.
And then uh it's it says, youknow, um, the uh owner is not

(22:48):
available right now.
We're gonna collect yourinformation, give me your name,
so so so, what you're callingfor.
And then the guy, the owner,gets a text message with the uh
process information.
And that's it.
The customer gets the idea thatthey are being taken care of
right away.
So now even if he misses onephone call, he doesn't lose that

(23:10):
customer because he gets themback as soon as he is available.
But that's a real use case of AIapplication, nothing fancy,
nothing complex.
We used um we use one simple uhAI application which is called
uh retail for building the AIvoice agent, a couple of LLM

(23:30):
calls like Chat GPT, and andthat was it, and then just
connecting his phone number.
So you see how that easy thatis, and it's saving him at least
seven seventy two hundreddollars per year, which is a lot
of money when you think aboutit.

SPEAKER_01 (23:45):
And actually, it's probably more than that because
he probably underestimated howmany calls he misses.
And in a sense, it's a glorifiedanswering service or answering
machine, but it adds the AIpunch of personalization,
dynamic content, heightenedefficiency, and tweak uh
tweakability.
And what I love about it is it'slike there's just an everyday

(24:09):
problem that you need to solve.
It has immediate ROI benefit,and you're not bringing in a
ClaudeBot and you're notcharging them a million dollars
to plug in, you know, the latestversion of Opus, you have a very
pragmatic, immediate solution toa ostensibly a very simple
problem.

(24:29):
And what you brought up earlier,I think, is a terrific point
that even as a society, we'reblinded by the BS, that this AI
is so super duper complicatedand expensive.
You need gigabytes of power in adata center before you could
really do anything.
And I think what you'reillustrating is a very pragmatic

(24:51):
common sense approach to solving98% of the problem, not only in
smaller businesses like whatyou're describing, but even the
larger ones.
There's this there's thisMcKinsey approach that we need
to hire a consultant for amillion dollars, they need to do
an analysis for six months, andthen we need to rip open their

(25:13):
infrastructure and spend anotherhundred million dollars in order
to process an email.
And I really like your everydaypractical approach.
I made a joke on another uhbusiness call where I don't even
care about the trillion dollarAI industry, I just want a smart
bot so we don't spend 15 minutesof a 30-minute call trying to

(25:36):
figure out when we can have thenext call.
It's the simple human thingslike that missing a call, making
schedules, organizing that verysimple AI solutions can be
activated, especially if youknow the customer.

SPEAKER_00 (25:53):
Yes.
And in fact, you can even go uhlike more basic than that.
When I speak to people, I askthem, uh, you know, what type of
AI uh technology do youcurrently have in your business?
You know, the most commonanswer, they either say Chat
GPT, Gemini, or even Claude,like though like the most

(26:19):
popular ones.
And I ask them, how do you usethat currently?
Well, my employees have abusiness account, they log in
and they use AI, like foreveryday tasks, right?
So most people don't know thatthey can create custom GPTs or
artifacts in Cloud or or gems inGoogle Gemini where you can

(26:41):
provide projects in uh open AIexactly, right?
So you can actually build acustomized tool for your
business that has your voice,that has your brand, that has
your information.
So every time someone wants touse Chat GPT, it is not generic

(27:01):
anymore, right?
Uh, because most people would goand say, hey, chat GPT, write an
email for me or rewrite this forme.
And that's not the rightapproach.
So even you don't really have togo into you know automation
workflows, you can do somethingas simple as creating a custom
GPT, a Gemini uh gem artifact, aproject, as you mentioned, and

(27:24):
then just make it work work foryour people.
Just think about what are peoplerepeatedly doing every day that
is keeping them, like that isactually, you know, uh getting
their time.
So that's that's what a businessowner should be uh thinking of.

SPEAKER_01 (27:41):
I work in consultancy too, and I do some
AI stuff.
And what's amazing to me is thatclients are often not aware of
some of the most basic aspectsof AI, like differentiating the
application level from themodel.
They have trouble understandingthe differences between the

(28:02):
models, and they really don'tunderstand what GPT is, they
don't understand generative AIeven as a concept.
So everyone is is getting allexcited and and blowing things
out of proportion in terms ofanxiety and complexity based on
just not being familiar with AI101.

(28:26):
So it also sounds like you'reeducating clients with some
basic practical applications ofthe applications, like build
your own little nested kind ofthing in the models, which are
more reflective of your needs asan organization, rather than
just setting your employeesloose in the generic version.

SPEAKER_00 (28:48):
Uh, yes, and actually, one of the offers that
we have, like, like think aboutit.
Are you gonna hire one personfull-time to look at how AI can
impact your business?
Is that really worth it?
So, one of the offers that wehave at Banso AI is that we
serve as your fractional AI guy.
So you can hire us for a blockof hours, maybe 20 a month, 10

(29:12):
per month, like it reallydepends.
We are open because at the endof the day, my end goal and my
dream is help business ownersnot to be ripped off and then
just just use AI to generatemoney to save time, to reduce
error.
That's my end goal.
So we offer that service whereyou can hire us a couple of
hours a month, um, and then weare going to serve as your

(29:36):
consultant and let you knowwhere AI makes sense, right?
And where you're gonna get yourmoney back.
Uh, so that's uh that's a veryinteresting approach,
absolutely.
And uh the other piece I wantedto say as well is that you don't
really need to have a fancy, afancy setup again, right?
I know that we talked about thisat the beginning, but I just

(29:57):
want to reinforce it.
It doesn't have to becomplicated.
Think about the simpler,repetitive thing in your
business, and that's how you'regonna make a difference.

SPEAKER_01 (30:07):
And also laying a foundation crawl, walk, run, it
builds trust with you.
And it also onboards theemployees because they're
freaking out.
They're scared that AI is gonnatake their job.
They're scared that if theyimplement it, it's gonna be too
complicated for them tounderstand and they're gonna

(30:30):
lose their job.
And they and anytime youintroduce anything new and
disruptive, people are scaredand there's high anxiety.
And there's a normal reaction tobuy a million-dollar chainsaw to
just go blow everything up whenuh, you know, uh a thousand
dollar scalpel with a little bitof delicacy and expertise can

(30:54):
solve just basic little issuesone step at a time.
I I really love your practicalcommon sense zero BS approach.
And I think it applies not onlyto small to medium-sized
businesses, but it's an endemicproblem and a huge opportunity
even at the enterprise level.

(31:16):
This same issue keeps coming upfor huge organizations, they
have antiquated or traditionalIT departments who are
struggling to catch up.
They have, you know, likevultures circling above or
hyenas at the watering hole ofall these AI consultants and
vendors who want to sell themmulti-million dollar solutions

(31:40):
when no one really knows whatthe hell is going on, and 99% of
their problems are what youdescribe.

SPEAKER_00 (31:47):
You know, uh my personal recommendation to
business owners is when you lookat 100% of your process, 80 to
90 percent can be automated.
Pretty much, right?
80 to 90 percent.
But there's like always the needto having a human in the loop.

(32:09):
That's how we call it.
Like a human making sure that AIdoesn't mess up because AI can
hallucinate, right?
That's how they call it, like itit can make up uh you know
statements or data, and youdon't want that.
That's why when we work withbusiness owners, we try to map
the process in such a way thatthe steps are you know very well

(32:35):
defined.
Step one, two, three, four.
If else, then you have tworoutes.
But we we don't let AI decidetoo much.
At least I don't recommend thatto owners.
If you have ever uh heard aboutAI agents, uh, you will also
hear a lot of hype.
Hey, I have an AI agent thatreplaced my entire marketing

(32:57):
team.
Uh, you will see like you willhear a lot of that in social
media, but at the end of theday, you want to keep it simple
because when you give too muchpower to the AI, AI can make
mistakes.
That's actually a statement atthe bottom of every LLM system.
AI can make mistakes.
How much money can you lose foran AI mistake, right?

(33:18):
So don't try to think of havingAI do the entire job for you.
It can do 80 to 90 percent, therest, a human can do it.
And the other important thing isif you are implementing AI in
your business, involve yourpeople, involve the ones that
are in the front line becauseyou don't want them to be
afraid.
You were absolutely correct withthat statement.

(33:40):
Um, I don't think AI willreplace 100% of the jobs or
humans.
Uh, we're not there yet.
I don't know if we're gonna getthere at any given point.
But uh my point is involve yourpeople, tell them what your
goals are, and that's gonna makethe implementation easier.
Even early on.
Uh when you do an automation, aworkflow, you have to repeat the

(34:03):
process and iterate.
Everything will not be perfectat the first run.
Believe me.
One of my when I was studying,you know, I spent like two weeks
creating an automation for acompany, right?
And everything worked well in mytesting, but the moment they ran
the automation for the firsttime live, it broke.

(34:26):
So that was a hard situation tobe in.
But what I'm trying to say isthat it will not work perfectly
uh in the first run.
It will run, maybe it's gonnahave a mistake or two.
But the important thing isiterate fast and early.
That's why we only take 10 daysbecause we want to do an MVP, we
want to make sure you use it.

(34:47):
You go to your team, ask themfor feedback, you come with that
feedback, and we fix it, right?
So that's what I believe theright workflow is.

SPEAKER_01 (34:55):
And the right tasks for the agents or for the AI
that you plug in have a lot todo with the remedial, repetitive
work that many human people aredoing.
So rather than losing your job,you can look at it as my job
just got better and more fun.
If I am spending four to sixhours a day filling in fields of

(35:19):
a spreadsheet, that's not reallydoing justice to my college
degree or my training or mycapacity to add value to this
organization.
And if your AI does that forthem more efficiently with human
supervision, to your point, thenall of a sudden their job
becomes more interesting,creative, and strategic.

(35:42):
And those extra four hours couldbe put into better or more
frequent customer service.
It could be put into thinking ofnew ideas for the company to do
a better product.
It could involve betterrelationships culturally in the
firm.
So you have a more loyalemployee base.

(36:03):
So you're actually improvingjobs by removing the boring
repetitive work and freeingpeople up to realize their
potential in other ways.

SPEAKER_00 (36:14):
Yes, and that's exactly what we stand for,
right?
Just get the repetitive task outof the way and use your time for
the strategic thinking.
Uh, and that's what what what wewant to give to business owners,
especially if they are soloowners.
Like, for instance, myself, Irun my company solo for now.

(36:35):
So, what does that mean?
That my research of new clientsand and you know, scraping data
and also uh, you know, creatingmessages, that's AI driven
because I know my system, I knowwhat goes first, what's the
second step, because I did itmyself, and I use AI to amplify

(36:57):
that and to reduce my effort.
There's still some effort that Iput in to contact the people,
talk to them, it's likeexplaining uh everything AI can
do.
But at the end of the day, youcan actually save a lot of time
if you have you know yourprocess built and documented.
So, my advice for your audienceif you are a business owner who

(37:18):
wants to grow and that want tosave time, AI could be you know
the best way or the bestdecision that you can make at
this point, as long as you knowhow your company and processes
work.
Don't be afraid, you will see alot of hype, right?
Take the good, live the hypeoutside.

SPEAKER_01 (37:38):
Um and don't fall prey to the to the scams.
And if you look at it, it itcould be a very sophisticated
scam, but it's still a scam ifsomeone is is demanding millions
to solve a problem that youcould ostensibly do for a few
thousand, and it amounts to thesame result in terms of ROI.

(38:03):
And to your point, also, humansare always in the loop, not only
as a checking mechanism, butthat human-human interaction is
never gonna go away.
The lifeblood of a business ishuman expertise, is human
culture, and most products andservices ultimately provide
value to other humans.

(38:23):
Yes, and the the human neverleaves that loop because that
that's what we're about.
So, ideally, AI is sophisticatedpersonalized automation of
remedial, repetitive tasks thatmachines are good at doing.
Let the machines do that, freeus up as humans to do more human

(38:44):
work, and you're doing it fromthe bottom up in a very
practical, common sense kind ofway, where you start by really
understanding their business andnot trying to sell them AI that
they don't understand and don'tneed.

SPEAKER_00 (38:59):
Exactly.
You know, I really like thatstatement.
Getting AI that you don't need.
I I, you know, uh, this is thisis a fun experience in this AI
world.
I was speaking with a friend ofmine who likes automation,
training tools.
Uh, he was showing me this coolthing he did with Clot Code.

(39:22):
He said, Hey, you know, I havenine AI agents working for me,
and then he showed me adashboard of little people, you
know, like like 8-bit guys, likewalking in a room, and and he
said, Hey, there's my workers,look at how good they are doing,
right?
And don't get me wrong, it wascool to see that.

(39:42):
And I told him, Hey, that'scool.
Uh now what problem is thatsolving for you?
And he didn't know what toanswer, right?
Because it's cool to have tohave those things going on, but
uh don't don't say we use AIjust for the sake of saying we
use AI in my company.
That's that's not the rightmentality.

SPEAKER_01 (40:03):
And and use it and saying that and doing that when
you don't even understand thebasics of AI, like AI 101.
Like, what the hell is it doing?
Why is it doing it?
Why is everybody talking aboutit?
People have this tendency ofeither dismissing things
outright that they don'tunderstand, or when they sense
there's lots of excitement, theyembrace it with equal

(40:25):
enthusiasm, but even less realrationale because they still
don't understand what the hellit is.
So that's that's just humannature, and being pragmatic
about it is probably a goodidea.
So let me let me express somepotential concerns as a small to
medium business, and and let'slet's see you addressing them as

(40:48):
you would a client.
So, concern number one is dataprivacy.
You're gonna be looking under myhood, you're gonna see all my
goods.
I'm gonna give you access to allof my data.
So, how can I be assured that mydata doesn't get out there, that
you're following securityprotocols?

(41:08):
And uh, and if we are plugginginto all of these applications,
how do I keep my data secure?

SPEAKER_00 (41:17):
One of the best ways of doing that is um having or
creating your own server.
So uh typically we ask you to uhlike everything that we do at
Banso AI remains with you.
So every subscription, everypurchase that you do, it's gonna
go under your name.
And if you want really uh likeprivacy and you don't want your

(41:38):
information going out to thirdparties or anything, then uh we
can buy a server in any of theserver like the host platforms
like Hostinger and stuff likethat, and we can just install
local LLMs like uh Lama, forinstance.

SPEAKER_01 (41:56):
That used used to be open source, but exactly.
Nvidia sells Nemo too, so youcould bring in bring in Nemo,
you could bring in Lama, you canbring in a proprietary server
LLM, but it's closed.
Exactly.
No plugins to anywhere else,exactly.

(42:18):
So your data is secure, how theyinteract with the data is
secure, and and you canguarantee that through the
proprietary LLM that you bringin because they're designed to
be local.

SPEAKER_00 (42:31):
Exactly, and that's the best approach.
Uh again, we like typically forsmall companies, uh security or
privacy is not necessarily a bigthing, but if it is for you, for
your company, then we have thatapproach.

SPEAKER_01 (42:47):
And that's part of your business model is you do
not create the software, you donot create the AI, you tweak it
and you plug it in based on whatyou know about the business.

SPEAKER_00 (43:02):
Exactly, especially where it makes sense.

SPEAKER_01 (43:05):
So a way to look at you is you are an AI
implementations operationsconsultant.
You got it with the technicalknow-how to plug it all in.
Exactly.
Exactly, and with that, you doget the data security that's
part of these, part of thesepackages.

(43:26):
That is correct.
Okay, and then what aboutscalability and your bandwidth?
I have a company with 10 people,maybe uh I have a company with a
hundred, fifteen hundred.
How do you scale and what kindof proven expertise do you bring
in?

SPEAKER_00 (43:48):
For scaling, uh, typically we want to make sure
that things run in the mostbasic way.
That's why our approach isautomating the small tasks uh
that takes you the most time,right?
In a way that we can do thatsame task a hundred times per
day.
Like think about this.
If we create a chatbot that cananswer customer uh questions,

(44:14):
right, uh we want to make theautomation simple enough so that
it can run through it a hundredtimes.
Right.
So that when you get morecustomers, then we make sure
that the answers stayconsistent.
Obviously as well, when you aretalking about scaling, we also
offer uh subscription packagesso that we are on the lookout

(44:38):
and continue to upgrade yoursystem as needed.
Like so we build a system oncewe ensure it works correctly.
We continue to monitor it.

SPEAKER_01 (44:49):
And if you want to scale then that's going to be a
different discussion in in in aseparate time frame right okay
so your plug and play you havethat scalability you even have a
subscription model to giveflexibility and then what about
proven case studies?
Obviously you you you probablycan't or don't want to share

(45:13):
them here public domain but ifsomeone is going to contract
with you do you have provencases to share of analogous
businesses in the vertical ofabout the right size to show
that um that you guys are legitthat you guys are effective in
delivering these solutions?

SPEAKER_00 (45:34):
Yes we have a couple of clients uh that can serve as
referrals they can be contactedin case uh and we have worked
with staffing companies homeservices companies such as
locksmith cleaning companies uhlogistic and transportation and
general customer servicecompanies all of our use cases
are in our website at bandsaw.aiif if they ever want to go and

(45:59):
take a look at them we'll put uhin the in uh in the description
links so they can contact youtake a look at your case studies
that's always useful yeah butbut again right we can point
them to the people uh that wehave as references as well so
that they can have the peace ofmind and uh you know typically

(46:21):
typically when I speak withpeople and uh we get into
business this is this is what Ido if I don't deliver the
automation that you know givesyou the ROI in 10 days which is
my offer we'll continue to do itfor free until it works right so
we have a a guarantee as wellthat you're gonna get going uh

(46:46):
either way right so that'sanother thing um and uh you only
pay when everything is workingright you got two weeks to get
it done I'm assuming it's 10business days and if it's not
done in 10 business days youjust keep cranking to ensure

(47:07):
that it gets done right and uhand it's pay after you're
satisfied which is a which is anice business model and it
sounds like you don't messaround and I like again the
pragmatic approach becauseyou're plugging in software and
applications that are alreadyproprietary already exists and
you're doing the discovery onthe upfront.

SPEAKER_01 (47:29):
So you're figuring out exactly what they need and
then presumably you're lookingfor the lowest hanging fruit in
terms of improving ROI.
What's the simplest task that wecan automate without with the
least amount of disruption andhassle to bring you the the
biggest benefit financially interms of hours saved and and

(47:51):
revenue?

SPEAKER_00 (47:52):
That's an interesting question because I
get that asked every day uh soone of the simplest automations
that we can do for you isreactivating your database of
dead leads.
So throughout the time you builda list of people that have
purchased your business or yourproduct or were intending to

(48:17):
purchase it.
So companies usually have adatabase of customers let's say
a thousand two thousandcustomers so uh one of the
automations that we do uh isgetting that list of customers
and reach out to them uh with apromotional package so that they
can you know go back to you andstart service once more so you

(48:40):
do CRM for AI exactly that'swhat you do chatbots for
customer service you do CRM andSalesforce optimization using
AI.

SPEAKER_01 (48:53):
Exactly you do a host of these things and
depending on what their need isthat's how you plug and play.
That is correct and and again wejust want to make sure that you
get the ROI up front that's whyI was telling you if you have a
list a list of database we canhave AI reach out to them with a
promotional offer that you mayhave and if you convert five

(49:15):
from that thousand only fivewhich is less than five percent
you know uh you already paid offfor the system so that's an easy
one from a thousand that's 0.5exactly so you only five people
and then your your serviceproduct is five hundred dollars
so you like you recover 2500 upfront I love the the pragmatic

(49:39):
pragmatic approach and then whatdoes success look like for you
so laddering up your business uhlarger organizations doing
enterprise stuff you seem you'reyoung you're ambitious you
started your own company basedon just doing operations and
seeing how broken everything wasand then riding the AI wave in a

(50:04):
way I have to say that that whatyou're saying is very refreshing
because I've had other AI folkswhere we're really talking in
the abstract for an hour andit's very challenging for me to
bring them down to earth can youcan you share with me a
practical application can youshare with me client benefit and

(50:26):
the conversation always likeit's filled with helium it just
goes back up again intoabstraction I like that you're
very nuts and bolts and down toearth chatbot for customer
service CRM improvement for deadleads I'm gonna be a glorified
answering machine if you missyour calls and see this is this

(50:48):
is basic it's not conventionallysexy but it immediately provides
benefit to clients yeah exactlyand my end goal is to again
right I was a business ownermyself managing small businesses
and I know how broken anoperation can be so if I can

(51:09):
make you uh you know ensure thatyour operation gets better and
you spend less time doing itthat's my end goal right I speak
with uh many people uh aboutthis I don't like the BS or the
hype that you see in socialmedia if you actually go to my
LinkedIn I don't talk aboutfancy stuff that you can do in

(51:30):
one minute I talk about real usecases of things that I've worked
on and let me give you the lastexample uh talking about yeah
talking about CRMs connectionsand stuff um I was uh working
also with this uh staffingcompany same client I've done
multiple projects for them sowhen they hired someone new they

(51:52):
would take that information andthey would put it in three
different systems their CRM aGoogle Sheet for uh for not
tracking purposes but foraccounting purposes and then
another CRM so they would spendlike 15 minutes with the first
person right and the biggestdeal about this is that they

(52:14):
could make mistakes by copypasting the wrong information.

SPEAKER_00 (52:18):
So I created for them because we were able to map
the process this is a 10 stepprocess workflow where we
created uh like a form thatcustomizes for them where the
recruiters could enter theinformation of the applicant.
Once they hit submit it does thefollowing number one it creates

(52:39):
uh the profile for the personwhere it copy pastes the
information in all the relevantCRMs and systems at once so they
don't have to go ahead and copypaste everything.
So that's two.
Number three it sends aninternal email communication
that a new person has been addedso everyone is aware.
And number four it starts asequence of welcome emails to

(53:03):
the person that's hired.
Right.
So all that was being donemanually taking between 15 to 20
minutes and now only one persondoes it and once they hit submit
everything happens magicallybased on their own rules.
So we're saving 15 to 20 minutesin one task that was repetitive

(53:23):
and not only that but theyreduced their error rate to
pretty much zero.

SPEAKER_01 (53:29):
So you are Marvin the macro you are the you're
you're plugging in kind of thisagentic stuff but in a in a very
basic way so there's a sequenceof tasks which used to be
manual.
You heighten the probability oferrors because there's cutting
and pasting and moving dataaround manually it's boring

(53:52):
remedial work best suited for amachine and then you set it up
so you understand where theinformation needs to go and how
and then you automate it.
And you're not bringing in aquantum computer or a nuclear
reactor it's pretty basic ifthen if then if then and then

(54:13):
woo off it goes and then theworker who used to be you know
going blind looking at thesespreadsheets they basically just
enter it once one time right andthen woo off it goes and then
they hit the button and that'sand that's a huge benefit you
know and that's a great way tosink your teeth into

(54:35):
operationalizing AI in a waythat's pragmatic and zero BS.

SPEAKER_00 (54:41):
And do you know how much of AI did we use there in
that whole process we used zero.
Yeah actually zero is less thanthat much exactly we don't have
any AI step because the logicwas well defined and the only
then if exactly exactly it'scompletely deterministic.

SPEAKER_01 (55:03):
Exactly you don't need a neural network doing deep
learning to do just conditionalpre-programmed commands.

SPEAKER_00 (55:14):
Exactly and that's and that's why I recommend no
agentic stuff unless it'sstrictly necessary for your use
case.

SPEAKER_01 (55:22):
And that's terrific because again you're starting at
the bottom and then you'rebuilding based on need most
consultancies start at the topwith abstract solutions to
problems that often don't evenexist yet and then it just adds
to the swirl and the BS andheightens anxiety.

(55:45):
So I think we need more peoplelike you Marvin in AI and even
if you've you've got you knowthe the latest version of Claude
and you're checking systems forsecurity protocol you know at
the kernel level your commonsense approach is very very
refreshing because you keep youreye on the prize which is at the

(56:08):
end of the day these are justtools.

SPEAKER_00 (56:11):
And if the tools don't have a pragmatic
application and they can't beimplemented rapidly with minimal
drama and they don't immediatelyprovide a boost in ROI then
we're wasting everyone's timeand money exactly exactly ROI
that's that's what I'm focusedon because I that that's what I

(56:33):
like to give back to people ROIfor their investment.

SPEAKER_01 (56:37):
Great well it's been a real pleasure talking to you I
am going to put your links anddescription in the notes for the
podcast and like comment shareeveryone bald ambition with
Marvin Martinez who is thefounder and it sounds like
you're the uh you're in a sensethe operations and consultant

(57:00):
grand poopah of Bandsaw AI andI'm tacitly assuming too that
you do a lot of these clientconsults yourself yes uh most of
the work is done by uh myselfbecause I want to ensure that
you get the highest quality fornow yes and I love your setup
and again the common senseapproach the basic issues that

(57:23):
could be solved simply if theycan be solved simply you don't
oversell or overpromise but itsounds like you do deliver which
is uh very very r refreshing socheck him out how does it work
if they click on the link theygo to your website contact

(57:44):
marvin uh do you organize uh alittle consulting session do you
have a free consulting sessionhow does your own business
development work yeah so youhave two options you can either
email me at info advance oraithat's the email you will see on
the website um or you can book acall directly with me um the

(58:08):
first call is 15 minutes umbecause the first call which is
free I just want to know if Ican help you right uh because if
I can't I will let you know upfront as well or if you if you
have something very messy thatyou're not ready to automate
then you're gonna hear that fromme as well.

SPEAKER_00 (58:28):
I like to be very transparent with people in those
15 minutes that we talk is justto understand what process do
you have in mind and if we canautomate it.
And if we decide that yes we area good fit for each other then
we have the first initial callwhich is the the discovery call

(58:48):
and in that discovery call I'mgonna map the process for you.
I'm gonna speak with you ask youa couple of questions and map
the process with uh for youright after that you get my
proposal and you are going toget uh the process map
regardless if you hire me tohelp you or not so that's that's

(59:09):
gonna be like an asset thatyou're gonna get back anyways
right so you do work you doworkflow and discovery on spec
and then if they like you thenyou plug it in and then you sign
a contract for implementation.

SPEAKER_01 (59:24):
Exactly and you have 10 day guarantee exactly
excellent excellent well thankyou so much for your time I wish
you good luck but it's not luckbecause I think you have a
terrific attitude and I I wishyou could do a TED talk because
you need to go up there andthere's so much smoke being

(59:48):
blown right now in AI and it'llbe wonderful for you to go up
there and just present thisbrass tacks bottom up solution
because it'll uh it'll helpreduce some of the hysteria
that's out there which iscounterproductive thank you

(01:00:09):
again like comment shareeveryone and go check out
Marvin's website and if you gota small to medium sized business
and you know you need tooptimize give marvin a contact
and he'll talk to you for 15minutes and you'll figure it out
thanks for listening andwatching

All Episodes

Episode Transcript

Popular Podcasts

Hey Jonas!

Crime Junkie

Betrayal Weekly

.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;}Marvin Martinez & Bandsaw.ai: Practical AI for Real ROI

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;}Hey Jonas!

Crime Junkie

Betrayal Weekly

All Episodes

Marvin Martinez & Bandsaw.ai: Practical AI for Real ROI

Hey Jonas!