AI Adoption for SMBs: Automate Workflows & Drive Results | Marvin Martinez | S1E13

Utsav: Welcome to Alt Consulting where we do AI adoption conversations. My name is Utsav. I'm founder of Strattof. We help companies drive AI adoption and innovation led growth.

Most companies think they need AI. What they actually need is fewer manual steps, fewer handoffs and fewer humans moving data between systems all day. That's where the real opportunity is. Not AI as a chatbot, AI as an operational leverage and small and mid-sized businesses might benefit the most because they can redesign workflows faster than large enterprises could ever do.

So today there's no theory, ⁓ no hype, ⁓ just practical on how actually gets implemented inside a business. I am joined today by Marvin Martinis. Marvin has spent over a decade in operations across hiring, training, quality, IT and end-to-end process management. Today, as founder and CEO of Bansaw AI, he works directly with business owners to identify operational friction, implement targeted automations and build systems that run predictable every day.

Marvin, great to have you here. Marvin J Martinez: Thank you so much for having me and it is a pleasure to speak to your audience and to yourself. Thank you so much for the invite. Utsav: So let's start with your focus.

Why have you chosen to focus on small and mid-sized businesses? Marvin J Martinez: it's because that's where I have seen the most return of investment ⁓ when they apply AI. The processes that every service and mid-sized company ⁓ are for the most part predictable. ⁓ they pretty much use kind of similar tools for CRMs, for reach out.

So it's easy to build an ecosystem where everything is connected and most importantly where they can actually see actions being done automatically. And then ⁓ business owners, founders, operators, they can focus their efforts on doing something else that is more productive and that needs more human reasoning, like being more strategic, think about how to reach out to more people, think about how to improve their current services, processes, products. So that's why I think this is my target.

Utsav: Great, and since you've worked across a range of areas, let's start with a say a back office example. So if you could share an example where you recently worked with a client automating a part of their back office, could you share that? Marvin J Martinez: Yes, I can share one example. ⁓ I was recently working with a client that works in the transportation and logistics services.

This company or this owner had 46 vehicles. It had 46 drivers and every driver was doing about five to 10 trips every day throughout the day. So every driver was reporting their trips. to a single operator, which was the one that was compiling all that information from every driver, and they were inputting that information into a CRM system called Trello.

So they had one car per day, and then they had one list per driver, and they were just inputting manually that information. That was not the worst part. Then what they had to do was to get all that information from Trello, which was their CRM, and just export it, or not export it, but copy paste it from Trello into a Google Sheet. And then what they had to do was to take this Google Sheet, kind of rearrange it, and send it over to the accounting team.

So the accounting team ⁓ could just generate the invoice for all the clients that they were servicing. If you notice, this is like a four steps process, probably more steps, right? But they were doing this every single day, Monday through Monday, ⁓ manually. This process was taking them three full hours per day.

So if you think about that in terms of money, if this person was making $30 an hour, it was just $90 that this company was spending just for a back office, manual, repetitive work every day. When I was speaking with this company, the first thing that we did was to map the process. I spoke with the founder because the founder had the vision of implementing AI automation and connecting the tools that are disconnected. But I also spoke with the person that was in charge of the process.

We came into a drawing board. We asked, where does the process start? What is the next step? What happens if this happens or what's the other?

thing that can happen. And we started just to draw in a paper, in a drawing board, the entire process. When we had that already mapped, we started to use one of our tools, which is N8n for orchestration. And we started to make all the connections.

So the process of getting the information from the CRM system Trello into Google Sheet and then into accounting for processing, ⁓ it's now fully automated. And what they were doing in three hours every day, now they are doing it ⁓ in 15 minutes. So what they got, what these guys recovered was two hours and 45 minutes ⁓ per day, pretty much. So that's, that's a lot of time that these guys saved.

And what is it that they did with those two hours and 45 minutes that they were not spending? So this operation person who was really knowledgeable. Started to work along with the founder and the owner of the company and they started to think about a strategy is to attract new clients They started to work in their marketing efforts creating ⁓ ads for Facebook Instagram and now they are getting more clients because they Could you know get together two or three times for a for a couple of weeks until they figure one step out That would have not been possible if they didn't have this this enough time back And one last thing that I want to mention is that ⁓ in the entire process, AI would only take part in just 10 % of the process.

So what that means is if these would take 10 steps, nine out of the 10 steps were repeatable steps that only require connecting the tools. And only one step required the implementation of AI. What am I trying to say? That if you want to be successful, in this AI world, one of the first things that you have to do is to ensure that your process is clear, simple to follow, and that it's very well defined.

Utsav: ⁓ it's a very simple example that you took ⁓ of a mid-sized not something which people will ⁓ think, you know, as first process they would want to automate or imagine AI helping in that process. But a couple of things that you said quite interesting just to capture them. First is you need to have a very clear understanding of what process you're running. ⁓ Second, you need to sort of reimagine what that process could be when you start applying AI.

Third, I think is that it's not that each and every step is AI, API is being called. Most of it might be just a process automation which has been happening for decades. So that's something which industry knows really well. There are parts of it which are becoming intelligent and that's what I think is the area for companies to focus on.

So I think a lovely example in back office and right now if you search about AI use cases, there are so many use cases on shared services. Let's take an example, if you can walk us through an example, which is ⁓ client facing, know, maybe lead generation or marketing workflow, which you automated, ⁓ which touches the sales team, that would be interesting just to get the contrast between different kinds of processes. Marvin J Martinez: Yeah, absolutely. Is it okay if I share my screen so that I can show you the backend, how everything looks?

All right. So I'm going to do a quick screen share. This is a process that I use for myself. I am a solo founder.

And of course, if I talk about AI, then obviously I have AI implemented in my company. But before talking about AI implementation, again, the first thing I did was to sit down. Utsav: ⁓ yeah sure. Sure sure.

Marvin J Martinez: Try to think what am I doing every day that is taking away my time if you are a founder if you have a company You have to start asking that question How do you identify that first step that you need to automate or use AI for? Just think about that one thing that you or your team does Every single day and that takes you five ten minutes because you may say it's only five minutes It's only ten minutes, but it compounds over time when you do it every day five days a week for weeks a month.

So I use one tool. ⁓ This right here that I'm showing is N8n. It's the tool that I use to orchestrate and connect all of my tools that I use. I have an external tool out to this which helps me to identify how many people have connected with me in my LinkedIn.

So what I do is people come to my profile or I go to people's profile, we make a connection and as soon as that connection is established in LinkedIn, the information comes here. That's my first step, that's my trigger, that's what initiate this automation. I have a little script, very simple, that extracts that information that came through and arranges it, it cleans it up. to be added to Google Sheets.

My CRM of people that I talk to every day, it's in Google Sheets. Your software doesn't have to be complicated. You don't need to have a fancy, expensive subscription to make it work. And lastly, I sent myself a message in Google Chat because I use the Google Workspace environment saying that there was one person added.

So that's like a four steps simple automation that has my CRM, which is right here, updated with all my leads information. And I also get like one message like this every time someone is at it. So if you think about it, hey, you know, this doesn't have AI included. Like it doesn't have to be in the first place when you start automating.

Again, the most important thing is, is this really saving me time? Is this really ⁓ giving me something back in return of what I am spending? Now, because I now have the profile information, including the name of the person, their email address, even their company name, then I have a second step added right here. So if you notice, this is an action that I have scheduled every day.

So every day I go to my Google Sheet. and I filter all the leads that are new to the Google ship that I showed previously. Then I just select a few number of people because if it's too much information then the automation may break depending on how much you load the automation. And then what I do is just to have one single AI instance which is this which does a full review of the company.

What does it mean? It looks at the company, what they do, their website, ⁓ how ⁓ many employees they have, if they are Series A, B, C, D, and so on and so forth. And that information is enriched and it's added to the Google Sheet again. So in just one day without me moving any finger, I have a full list, a full CRM with...

people and enriched the information about their company, what they do and where they stand. So that way, when I have all that information handy, I can go back to the person and say, Hey, Utsav, I saw that you have 10, 20 new employees in your companies. Congratulations on your growth. then I can start a conversation because the end of the day, ⁓ AI amplify your process and what you have, but it will never replace that human interaction.

For the audience, how many times have you received sequence of messages in your LinkedIn, in your email inbox of people saying, hey, and you can actually notice that it's a generic email that's coming through with just a couple of face holders, right? So that's when it is important that your process has 80, 90 % automation, and then the other 10 % should be a human in the loop that is intervening. understanding the process ensuring that the output is correct. Right.

And one last advice. I was speaking with a founder just yesterday and I was showing him a tool that I did to start understanding how much money and time founders can make can actually save with AI. And he said, you know, I know that AI can save me time and money. The problem is that I am not sure if AI will give me the results that I produce, right, or the quality of results that I produce.

And that's true. That's why my recommendation is always to have deterministic process. If you notice the processes I showed, it's very linear. Step one, two, three, and four.

There's no room for the AI to take decisions because again, decisions that needs to be taken by human to my ⁓ experience because AI can hallucinate. or maybe AI cannot read your instructions well, or even more, maybe the instructions you provided were not clear enough or detailed enough for the AI to make a decision like you would. So ⁓ start small, don't overcomplicate things, don't make it complex, just make sure that it works. ⁓ And AI is important, yes it is, but it's only 10, 20 % of the process if your process is well defined.

And this usually happens with small business owners. Utsav: Yeah, no, think ⁓ thanks for sharing this and thanks for sharing something live which you use because it becomes more practical than an abstract theoretical concept. I think ⁓ one thing which ⁓ I was thinking while you were sharing the ⁓ example was seems to be a great solution for solopreneurs seems to be a great solution for someone who is trying to reach out to their own clients and doing their own analysis with the data.

But then you sort of in a way replace part of features of LinkedIn sales navigator, which does something similar charges much more maybe than using your own automated tool. And you also have flexibility to do features in your own way. Now the challenge which comes in is is when you start looking at this kind of an automation ⁓ for a large sales team of a large organization, everyone has their own unique way to reach out to people. Everyone like you might prefer a particular way to reach out to people, some triggers which you might prioritize as compared to others.

how have you sort of, have you faced that situation where in an organization, there are a couple of people, everyone has their own perspective and they are all right in their own way. you can't actually lay down a standard process and you would need to factor in customization which they want. How do you handle that kind of a situation? Marvin J Martinez: That's a very unique situation and I've gone through that.

There are certain salespeople that like to do things one way or the other. And you can actually branch out the automation in multiple outputs. So for instance, you have your main process. Step one, I look at the lead.

Step two, I get their email. Step three, I put that information over into a CRM. So at that point in time, what do I do next? then I have to assign it to a salesperson, right?

Depending on if the person is, or if this lead is maybe an owner, a CEO, maybe it's just an operations manager. So at that point in time, you can actually have AI say, hey, based on this criteria, if this company is this big, and if you are speaking with this executive level, then you assign this person to this sales. Associate right and then if this if this prospect you know has Less tenure or whatever criteria right then it goes to a different branch So the way that I fixed it for one of my clients in the past is just by branching it out depending on on the style of each salesperson But here the question is how do you like how do you decide who gets what?

That's when your process needs to be clear enough to say, hey, this is what I'm doing. ⁓ And again, business owners, they know their process, they know their products, but typically they have it written down either in their heads or maybe they share it with someone, but they don't have full documentation on how to do it. Utsav: I think earlier we were speaking about that example of the training organization which was trying to get all the information about potential opportunities for their clients to get placed.

I found it to be a lovely example of how you use it for a mid-sized company to ensure they capture all the opportunities they have for their clients in a target market. be great if you could sort of walk us through that because it's a very, again, just one more thing which I just want to let people know listening to this is purposefully we are taking very simple easy processes because YouTube is flooded with complicated examples of you 15 different steps that you need to take and multiple applications which need to be integrated.

We want to make it purposefully very simple and this example is of a training company if I'm not mistaken which is into placement of people. So that's something which I think most of us can relate to in terms of you know if you have applied for a job ever or hired someone in your business, you would understand the pain. So Marvin, it will be great if you could sort of walk us through that. Marvin J Martinez: I was speaking with this that ⁓ hired ⁓ talent in Latin and would match that talent with companies the United States.

They had four or five ⁓ recruiters. ⁓ What were is ⁓ understanding the job requirements and hiring people. When they were hiring people, ⁓ they were doing this process. Number one, they were capturing the people's information, the new hires information, and they were sending that email manually to different stakeholders.

The problem with that is that the emails were different from recruiter to recruiter. One people would do it this way, some other people would put this other information. ⁓ It was not the same information that was coming through every time. Sometimes, they were missing critical pieces of information.

Then after that, they would have to put that information into two or three different CRM systems, which serves different purposes. And then after that, they had to send an internal email ⁓ to accounting and all the people that created the credentials, just to let them know that they had one new person coming in. And then after that, they would have to send a sequence of emails to the new person that was hired, ⁓ letting them know about the company, what they did, what to expect next.

And all of this was being done manually. Now, we are not talking about ⁓ losing money or losing time, because if you think how much time can you spend just putting some information into an email, maybe not. But one of the challenges these people were having is that they were making entry. mistakes.

They were putting the wrong information for one person, they were putting an ID where an ID was not suitable, and stuff like that. So what we did was to automate the process. Again, the first step that we took is to sit down, look at how the process would look like, look at who initiated the process, point A, point B, point C, very linear, put it in a paper. go to the drawing board.

That's the first step every time. So if you hire someone that's going to help you with AI implementation, just make sure to make it easier and faster to have your process mapped out. After that, we were thinking, hey, how do we ensure that everyone enters the same information? So we used a Vive coding tool, which is called lovable.

dev. Basically what it does is you explain in plain English. what you would like to accomplish and this tool will ⁓ make all the coding and everything and it's going to create the application for you. And we created some sort of a, think about it like a landing page.

We created one single page where the recruiters through an authentication process because they have to enter their name, their credentials to just get in for security of course and to see who was doing what and they were just inputting. the new hires information in that ⁓ form. That was the front end. That was the tool that everyone had access to.

After they hit submit, that information goes automatically to the two or three CRMs they were using. And at the same time, that information was going in an email, same email every day, same structure, nothing changed to the multiple stakeholders. So everyone was getting the same information and the same structured details from every new hire. And lastly, that information was put ⁓ into an ⁓ email sequence for the new hires.

So we were triggering the emails day one as soon as they were hired, day three, and then day five. With the same template, everyone was getting the same communication on time without anybody missing anything. Up to this point, what this company saved was the fact that they are not making mistakes. So in that process, there are no more errors or mistakes any longer.

Utsav: And that, think, is something which people have seen in CRM systems. ⁓ So many industries have similar problem. I still receive certain emails from companies. have been a client for a long time, misspelled my email and my name.

And sometimes the email also goes to some other party because somebody made a data entry issue while jotting down or sort of entering in the system the email ID from the form which I had filled. So I can totally relate to it. ⁓ One last question, quick thoughts that you have for anyone which is a mid-sized company, CEO or the leader, trying to drive AI adoption, what guidance would you give them? One thing which has come in this conversation is be very clear and be very ⁓ particular about which process you are picking up, document that process, but beyond that, what other things would you suggest to look into?

Marvin J Martinez: In AI adoption for mid-size ⁓ businesses, try to involve everyone early in the process. One of the things that you will hear in social media is that AI will replace humans. I don't think it will, 100%. It will replace repetitive processes, right?

But what I'm seeing is that employees are afraid. that maybe because a company is implementing AI that they will lose their jobs. For mid-sized businesses, ⁓ what happens on the contrary is that the employees get more free time after implementation. But the business owners, they don't necessarily involve their operators and frontliners at the beginning of the process.

And that's where the change management makes it hard for the business owner. My recommendation is speak to your team before any AI implementation. Tell them, hey, this is an issue we have. We want to save you time, and we want your input to ensure that what we're doing makes sense.

So when you have the first draft of the automation, when you have your first MVP workable system, speak to your team and say, hey, this is what we have. I want you to try it. I want you to tell me, does it work? Does it not work?

Is it helpful? Does it matter? Is it relevant to you? And then that's going to make it easier at the end to make the implementation.

Utsav: Yeah. I think you made my case. This is what I have been telling that change management or AI adoption, the behavior side of the equation is something that you need to start as early as possible. You need to involve the right set of stakeholders for reimagining the process.

Even before that, figuring it out where you're headed, what's the new organization going to be? And then when you can start involving more people with technical background to automate a particular process or ⁓ know, redesign an existing process and see how AI can help you achieve benefits. So thank you. think that's a great place to wrap.

What this conversation makes clear is that AI adoption is not about tools. It's about how work actually gets done. And once you change that, the impact actually shows up quite quickly, especially in mid-sized organizations. Marvin, this is very practical.

Thanks for joining. And for everyone listening, if you're thinking about AI, don't start with the tool. Don't start making an agent. ⁓ before you are very clear with the workflow.

That's where everything changes. Thank you for joining me again and I see you all in the next episode. Marvin J Martinez: Thank you.