From Spreadsheets to AI Agents with Zulfiya Forsythe

Mark: 00:43

Hello everyone, welcome back. Today we're joined by Zulfiya Forsythe and she's the CEO and founder of the Omadli Group. Where they help companies optimize their operations. They gain valuable insights from data and AI, and she helps them buy back their time using AI. She started a career in accounting, where she got very stressed out with the endless spreadsheet, taught herself VBA and SQL, and now she's running the Omadli Group. She's going to share her journey and her story here with us today. Zulfiya, welcome and welcome. Please give us a bit of an introduction.

Zulfiya: 01:14

Thank you so much for having me here, Mark. I'm really happy to be here.

Yeah, it's been quite a journey, but I think one thing that always, and just going back even before I moved to the United States, almost 19 years ago, it's crazy time flies. I always kind of ask myself this question for myself. How can I make this better? What else has out there that's better than this? Or how can I help? I think buy back my time or automate processes or improve myself in any way.

So I think when I made the career transition from accounting to IT, I was so fortunate. I'm really, I'm very fortunate about this. And I'll tell you why. Because at that time, I was an accountant. And I wasn't even trying to apply for some type of analytical jobs. I've got so many no's from recruiters. They were like, "Nope, you have an accounting job. This is your box that you fit in. You are not even a financial analyst." So for me to go into system analyst was like a huge jump, basically. And that jump was given to me by the CFO at a company that saw potential in me and gave me this opportunity.

So I learned from scratch. I'm super grateful for that and just all went from question, "How can I make this better?".

Mark: 02:26

How did you get to, I mean, you said that you taught yourself VBA and SQL as an accountant, right? As somebody that's into spreadsheets and I've worked with those people quite a bit. What makes you think I really want to learn VBA and SQL?

Zulfiya: 02:41

Yeah, I just did not want to do repetitive things over and over again. And I, at that time, I had no idea that there are other tools out there. I just started getting curious. And actually, it was a funny story.

You know, sometimes life shows you signs, but you kind of ignore them. You think like, well, you know, this is just. And the sign was, so I had this bank-working solution that I know would take me probably, you know, I had like five of them. That would take me like maybe, you know, somewhere three to five hours to get done. But I would sit there, spend the entire weekend trying to automate. Because I'll try to, like, I just cannot do another month of the same thing. What's another way that I can automate it? And I would, because I don't know anything about automation at that time, I don't know about anything. I basically, at that time, there's no AI that can speak to and ask to, right? Now you just have a internet and you have YouTubes and you have the forums.

So I would just sit there and I would spend like 15 hours trying to automate something that, It took me like three hours, but I knew that because, hey, I did vibing. Reconciliations, but it would be additional more that would be coming once I get those done. And I just did not want to do the same thing over and over again. So yeah, this is, I think, what just in seeing little results of how, wow, like, you can actually record this and it works. It just kept my, kind of in sort of syntactic.

Mark: 04:03

Okay, I understand why you ended up in technology and dropped accounting. That makes a lot of sense.

So and you said you got an opportunity then finally, by someone to work for as an analyst, what did that look like? And how did you start in that company? What was your childhood?

Zulfiya: 04:20

So actually, I was an accountant at the senior care management company, and we used Yardi. Yardi is a property management software.

So I was a kind of like accountant that used that software. So I remember when I thought of myself as like almost a janitor of the books because I would do things every month. Same things, I would pull the reports. And I remember, like, I was like, all right, everything's coming from that system. Why can't it just come in the way I want to? Why do I have to download these spreadsheets, like marry them together, figure out these formulas, you know, why can't it just all come the way I want? So I spend the weekend mapping things out. And I remember that time I had a friend who had a business intelligence background, like Zulfiya, you need to be on the other side. You need to go to BI. And I'm like, no, but I'm a CPA. I do certified public accountants.

You know, I was really kind of like... But what I was doing there, I was doing data mapping. Just kind of on a plain level, telling, okay, which accounts come from where, how they group. So it literally was data mapping. And then after that, My CFO saw that work and started automating like expense distribution via Visual Basic Studio. And then I started hanging out with IT folks in the same company. And I see like what they do. They're like, all right, write the code, press F5 button. And now it's like, my God, I'm going to be able to do this. All the information comes in. I was like, what is that? I need to get my hands on this. Like, what am I? So everything in the system, you have it. And yeah, I got the opportunity to move across the department. And that's where my journey started. That means weekend evening in the buses watching about like, okay, structured query language, foreign key, primary key, joins left and right. I'm like, what is that? But let's go learn it. And I actually remember my coworker said something to me. So, okay, You're really good at this. You have a talent. I was like, really? Really? And I kind of got all like, wow, okay, I'm onto something. So when I moved into the IT department, so I was kind of like spending my time into that database of Yardi. Like, basically, I would close the tickets as a system analyst, would put those reports, fix the bugs of those reports. And I just loved it. Sit there and do like joins and left and then try to see how the data is coming. But now with AI, everybody could see. Here's Amar. Here's what I've seen. That's incredible. I spend so much time trying to learn how to query databases. To get the most accurate information. For example, we were trying to ask, okay, And the RDSA, it's a very large property management software. It's very well known in its space. It's like, so if somebody is asking, all right, show me how many units are occupied in this type of a property, for example, in this residential property in the city, in a property, let's say Madison. So where I would go, I would have to go write really good query to make sure I actually pull that property, pull the right amount of units and put like the occupancy will calculate correctly. So it's been substantial time to make sure it doesn't, you know, data doesn't get hallucinated for me, basically. But now it's incredible. Now everybody just uses human language to query things. You don't have to write select statements, do left join, right joins, and learn about where clauses and havings. And like there's a whole syntax that you have to learn. Now it's just you use a human language and you can retrieve the information.

Mark: 07:52

And even if you wanted to build an SQL query, you can still do that with AI. You can just feed it the data model and say, just extract the data from me. Give me the SQL statement that I need. And that's it. That's all you need to do.

Zulfiya: 08:02

Yeah, and you can ask from AI to help you even teach how to write query language. Okay. "Well, what's the, can you evaluate this max for now? "Like give me, if you're trying to see, okay, "what's the maximum rent that was charged "in this property for this unit?" Before, I would have to like, you know how to write that query. Now you could just, pop it to AI and AI will write for you and it won't even critique it. Invaluable. It does hallucinate.

So I still think we'll need to do expert knowledge, but.

Mark: 08:32

Yes. I hope - I agree.

Zulfiya: 08:33

So much.

Mark: 08:35

And now you're on the other side of the fence, right? You moved to an IT department, you've started your own company. Does your background in accounting, does it still help you now? Or do you think with AI, That's easy anyway.

Zulfiya: 08:49

I think, you know, what's interesting was thinking the other way. Okay, with AI, now knowledge is on everybody's fingertips, right? Knowledge is no longer a supreme. Knowledge is there. But you know what's supreme now? Judgment. That's what I thought. Judge is now, because how do you know what's right? What's wrong? How do you know if it's correct information without blindly testing it? And this is what domain expertise at every level, if you're a marketing person and using AI. Now, let's say, it was interesting. Now, when you have automation and now that automation is supervised by IT department, but it's a marketing automation.

Like you have no, as a person in IT, you don't understand marketing, might not know all the aspects of the marketing. So the person in the loop should be the marketing person evaluating the responses. Now, the IT and the same thing. Vice versa, for example.

So I think what will be I is judgment is so important. And where accounting helps me is understanding business processes, understanding the financial bottom line effect. On things, like for example, we have different models out there, but how do you make sure that it's cost optimized? How do you make sure that it's not using up like tokens and basically it's cost effective and affecting what is ROI on this example automation? Does it make sense or not?

Mark: 10:08

Yeah, it's so cool. Makes sense. No, and I agree. Judgment is the hardest thing I've seen.

he middle of COVID, September: 2021

Zulfiya: 10:41

Absolutely. Thank you so much.

iness after, but I started in: 2020

So I went to Upwork. And what do I know? I know Yardi well.

So how can I now utilize my knowledge of this system and help other people with, you know, setting up a Yardi maybe, automating their processes? I remember my first client, I was like, my God, I'm going to be a Yardi.

,: 2020

So we'll never forget this because when I got this first client, I was just like kind of going like a, you know, sending thousands basically of proposals. My god somebody hired me now i don't have a behind the team that i can go and ask questions from right i have no ai as well so if i screw something up basically well now somebody's business that i'm like basically you know not helping and somebody's trusting their they've time, wouldn't we? All right.

So I took the step and I went on all and it was great. I was like, okay, I actually could help. And I love the idea that I don't have to go through. For example, when you're trying to apply for a job, you go through the entire interview process that takes weeks.

And then you meet with this team, that team here, but the owner, okay, I'm using Yardi. Can you help me? I have a mess. Can you start tomorrow? Shuck! So right there, I'm at the task, putting things together, jumping on and getting stuff done. And now you see the immediate return on your time and immediate impact as well.

Mark: 12:27

Okay. And I've got to ask this question, right? Because it's something that I saw when I was looking into you. You're everywhere.

So we're in a community together and I see you post there and chat with people all the time. I know you have this weekly, I think, open AI hour. I've seen what you do on social media. You're everywhere. And still you run a business. It seems like you're, and you just talked about the thousands of proposals that you put out. And I'm actually tempted to believe that is actually true, that number. What's... What makes you tick and how do you get How do you show up everywhere all the.

Zulfiya: 13:00

Time? Okay.

So a thousand proposals was always to me, but I've sent a lot of proposals. I think I have a team. And what I'm learning is, I'm still learning. It's still an A-B testing, Mark, because I've read the book, Buy Back Your Time, right? De Martel, you got to buy back your time and have systems. And this is what I've been doing. Putting on for the last month, working on the systems, having a team that I cannot show up everywhere because I have two little kids and I have a business to run, ops to deliver, and as well as sales to do.

my own. And I think after In: 2024

So you understand Mark, all these years I was doing it all by myself and as I got executive assistant now, so we're both on Denmark TALIS and I read the book, "Buy Back Your Time", but I think I was pregnant so a lot of things will haze too. I was consuming all of the books, all of the content, and I was implementing, but not as implementing at the speed that I should have.

So I would say for everyone who's starting the business, if you got one successful client, Hire EA, hire executive assistant that can even part-time can come help you now, can help you to put the case study, can help you to put your word out about the things that you do, the profile of person that X you serve offering Y. So help you to copywriting content ideas.

So help you to hire and put together the standard operating procedures with others. So your time goes into more, I would say, okay, refining those standard operating procedures. Get the word out there and training people so you can scale faster and scale in more, I would say, a realistic way where you not stretch yourself thin.

So I started with even podcasting. I started in February of this year.

So: 2025

Mark: 16:27

Yeah, I think it's such a cool lesson and such a valuable lesson instead of just scaling, seeing where you can buy back your time. And I agree.

I mean, we've met with / through Dan Martell and it's an incredible book that he wrote. Indeed.

So now you're running, you started as a freelancer, you're running the Omadli group. Where's that name actually from? What does it mean?

Zulfiya: 16:46

Yes, I'm originally from Uzbekistan. It means lucky.

Mark: 16:50

It means lucky.

Zulfiya: 16:51

Yes. And it's interesting.

So it's my husband is third generation Irish and you don't have Latin sign. So he turned up for me.

Mark: 17:00

Okay, and wow, that's so cool. Okay. And you launched in pandemic times, middle of Corona, which is, I'd say, Courageous, but very difficult finding clients, although Upwork is very much online still, but a lot of businesses were suffering. How did that shape your business in any way.

Zulfiya: 17:19

So I actually was looking for full-time jobs. So I did two jobs.

at the same time. But then in: 2023

So it was already, and I'm not a quitter mark. I'm not a quitter, so it's almost like the universe was giving me like a nudge, kind of very heavy nudge. All right, you cannot keep working until two o'clock in the morning. You need to basically focus on one thing and grow that thing.

Mark: 18:01

Okay, that makes sense now, but that's such a cool story. And then you're now, well, you've built a business. It looks like a very successful business. You have like a client story or your most successful story. Of what you did with your Omadli group.

Zulfiya: 18:16

Absolutely. Well, I absolutely loved working with my clients. And I think, so we helped clients in e-commerce space and Yardi, a property management company, for example.

So with Yardi, it's clients who are just like, all right, can you do this? And that's what I love. Just reaching out, finding you on LinkedIn, be like, all right, I have this thing. We're a business.

You know, we're trying to build the business intelligence in our company. Is there something you can help us to automate our spreadsheet? Sure. And it just turns from one spreadsheet, and this is what I love. It's always just one small task. As soon as they see the benefit of automation, now we have built entire business intelligence suite for them. We've built multiple reports supporting the multiple departments, property management department, asset management department, then we talk about financial department as well.

So now they have these reports that Nobody's putting together anymore. Before, when we came to them, they had this 14-tab report that they had to support on a weekly basis to show what their performance was like, large performance. And now they can grow, scale their business as well because they have their numbers. They can see what's happening with their, for example, utility report, the grabs, and what's happening with their occupancy report, what's happening with their delinquencies as well, who's paying, who's not.

So that was another one was with help of the e-commerce client as well. They had someone left the job and all the reports was kind of hanging, you know, and we came in and I remember looking at one of the reports. It was 34 joints, like a kind of like when you open a SQL point, 34 joints. It was a mess, basically.

So we came in, we cleaned it all up, delivered one report and then two plus years. And, you know, the project was going on.

So it's great. And then I think hearing the feedback from all clients was just like, all right. One of the stuff that they said, we look at auto developers and they want everything structured, meaning they want, okay, can you give us the exact scope? Can you please like put everything for us in detail so we know where to go? For you, it's like, all right, we have a mess. Can you come help us? Sure. And we kind of figure out, put the mess together. We help them because You know, it takes, I understand them because. We want to make our clients' life easier. We don't want them to make extra help. We're experts, we come in, we'll put structure for you, we'll clean up your mess. We put some documentation for you as well.

So now you have a clear processes, you see what report is coming, how it's going. And we basically then went on supporting the CFO, went on supporting their marketing department, and then just the operation department. And yeah, that was one of the things that just like, whatever they give us, we can handle, and we can basically give them clarity. As AI came in now, when I was starting with business intelligence and now artificial intelligence. It just made sense for us to move in that sense, in that direction now that we have data, right? We work with data at the end of the day. The way I see the AI is, you At that time, we still have the silo system.

So there's a lot of systems that companies use, lots of them, right? They use for marketing, then they use for financial, then they use for ops, then project management, then there's a time management to get it tied. There's a client relations system as well.

So there's a lot of systems that they use, and there's a lot of, let's say, gaps in between of them. The way we look at this, okay, we assess when the company comes to us. And one thing is great, Mark, that with AI, we say, okay, I want to automate my business. But when you look at the business, you're like, you don't even need AI here. You just need business intelligence in here. Business intelligence is AKA data analytics, basically. You just need automation where you're reporting. You don't need AI yet.

So that's good too, because you don't need to spend money on LLMs. And with large image models, it's generative AI.

You know, it needs more time, greater duration, testing, where with data analytics, What you do is, all right, well, you retrieve the information, you have the formula, you have more predictable results. So we assess what the need is of the company.

So do they need automation? Do they need, artificial intelligence or is this a distance intelligence task?

So then You don't need to Slack AI on everything basically.

Mark: 22:35

And still you've developed a couple of your own AI agents. I was actually looking at one of your videos and I thought that was a really cool one, because I think it speaks to a lot of us in tech, where you said that you spoke to somebody who was saying they used AI agents and basically they were just using ChatGPT. You turned your screen around and it was like full IDE with Python codes everywhere. And that's the reality of building Still quite a bit of coding to get a full agent up and running. You've built a couple of your own agents at Omadli. Can you tell me what those are for and why you built them?

Zulfiya: 23:10

So with looking at everything, I've learned that I really wanted to niche in the industry to kind of become expert on that industry. That way, we can really help a lot of companies address their pain points. A couple of agents that now we productize to make it so much easier is our Vicky voice agent to help construction service companies with picking up the phones, because as you know, as you might know, some of the, for example, HVAC, like people might call with, but we are still. I would say there's still heat season when I come in, but we're at the end of it, but we're coming into plumbing season where the frost comes in, the coal comes in, burst pipes, people start calling. And a lot of times what we notice is for small business owners is air. We all know that if the contractor is not picking up, it's moving on to another contractor.

So what we do is we have Vicky voice agent that helps to answer all the calls, helps to route them to the urgent calls, help them to book appointments. And another very cool thing is as we speak to a lot of business owners, they might have a need for second language.

So now we're adding Spanish, for example, or we add in Russian. So a lot of business owners come in, they're immigrants and they that English is sometimes a challenge. They're really good at what they do, but they might have the clients that are the non-English speakers.

So we're adding that and really cool actually, So what we have, the wiki answers the call in English, but it gives it translation to them. In notes.

So then you can see the translated in their language. Well, what's the, what the person was asking? Because when we hear the story, sometimes as a business owners, they will pick up the call and they don't understand the English. They would have to call their buddy.

Like, hey, someone is calling. Can you please translate it to me? Help me to kind of, how can I, or they would have to call someone and their family members. They can speak to kind of assist that person. But now Vicky can take all of that. Do the translation and send them, book that in the CRM. Be like, this is a translation in your language, what they were calling and asking you about.

Mark: 25:18

Wow. That's really cool that it does the translation, but for the business owners to translate that English conversation into Spanish or Russian for them. I had no idea. Boom. Okay. And I think that's incredibly helpful. You said you're specializing. Is that mostly to construction and property management or what's your specialization now?

Zulfiya: 25:37

Yes. So we're working right now on construction.

So we're helping construction companies, whether it's plumbing companies or So I would say plumbing, trades, fire alarm system as well to onboard BK, but automate other processes that might have data analytics. Same thing, you need reporting. And now we know very well Yardi, so I'm kind of narrowing down and helping a Yardi property, other Yardi businesses. Property management ways as well to automate the.

Mark: 26:09

Processes. And you're working with all these people as plumbers, for example, there may be not the most technical people out there. How do you explain... AI or data analytics or automation or reporting to a plumber that's running a plumbing business?

Zulfiya: 26:24

That's a good question, actually. And then I would say it's still a challenge that I'm working on and I have to do door-to-door. I'm going to do door-to-door and I'm thinking, all right, this week actually I have to do some door-to-door in my local neighborhood. And the way I say is this, okay, this is technology. I'm trying to kind of remove the AI part of it, but really what are AI is this technology to help you to answer all your calls. To book more jobs and help you to make more money. And I have a demo. They can speak to AI, for example, on the call and assess how it performs. But it's definitely going to be a challenge, Mark, because a lot of people that are owners, I need to take them on the other side. I need to show them, hey, there's a better way because a lot of us, they know this works. Even if I'm doing spreadsheets, I know it works. I know there's a repeatable process around it. This is a new thing to me in terms of automation. What helps is showing other win cases as well, showing how I helped even if they you already like, look, they went from Excel spending 192 hours to zero hours now. Zero hours.

I mean, You can't beat that. If you want to report to see what your numbers are, how many jobs you closed, Like for example, now you have to ask your assistant, put it together, or someone in your team, put it together. They have to scramble, but then you have this live data at your fingertips that you can see each time. And I want to talk about one more, Pharma. We made Pharma on AI agent is a knowledge-based agent. And that knowledge-based agent is actually... It's great because they can take all of their historical submittals and now they can query because the traditional way they would have to go in the system, find the profile of the customer that they are trying to, let's say, order equipment, like water heater.

And then they have to dig through information. Where now, if they could just query it with a human language, just in English, all right, well, I have this potential customer. Let's say for this type of a building, we need this many water heaters.

Well, can you tell us? Who did we order from before and what capacity, what sizes? And now the information kind of comes to them instead of them digging through their system, finding it out and like, you know, spending all that time and need potential rework as well. To be avoided.

Mark: 28:50

So very much show and tell. Show it to them. Show what it can do.

Zulfiya: 28:56

Yes. Working on my content, so a couple of pieces that I'm recording right now is how Vicky is literally I'm talking to Vicky and it's going to my Go High level and entering the podcast. Urgent. Basically training Argentinian.

So I'm a customer. I'm saying I need urgently to fix my AC and Vicky is on go-hi-la-la and creating tickets. And putting words in buckets.

Mark: 29:25

Putting that in an online social media video so everybody can see it.

Zulfiya: 29:28

So cool. - Yes.

Mark: 29:30

- Okay, I wanna talk a bit about the other side. I mean, a lot of people that are listening to this are CTOs, they are in technology, they are CIOs. You speak to a lot of, small business owners, but maybe the same applies to other larger businesses as well? What do you still see as the biggest misconceptions or misunderstandings about AI and the people that you talk to?

Zulfiya: 29:51

Very good one. Like in that video, I use ChatGPT every day. I think that's a misconception. Or for example, I was listening to an interesting article. A lot of people think, okay, I'm AI, competitive just because I rolled out chat GPT and I rolled out copilot. But.

You know, that's not really AI competitive there. The training, assessment training needs to be done. You need to be looking at other processes that you have that are still lagging. And I'm telling you, there's so many people that I talked to, Mark, that are so successful and so large that This is a wake up call now because they've been running their business as like 10 years plus back. They still have a lot of coin key systems that they use. Their QA, QC process is very like even how they submit the invoices, like a large companies, they still lag on their quality process, which takes them cash flow effect on that.

So basically, which cost them the cash flow. So just because there's a, this process is so manual, So I would say, yeah, just looking and saying, okay, well, I use JetGPT every day and, there could be The question needs to be asked at the process that we'll look at. What can be made better here? What can be done better here? If it's repetitive like two, three times every month or every week, What's another way of doing this? Because if we're not employing our mindset, our competitors are just going to eat us for lunch.

Mark: 31:26

Yeah, I think that makes sense. And I think that was my next question, but I think you sort of answered it. A lot of CIOs, CTOs, they work in these larger businesses, these larger organizations, and they see all these. Flunky systems around them, like you said about the invoice check, how should they go about or how should they think about convincing their stakeholders in finance or somewhere else? You were an accountant. What would you expect from somebody like that to persuade you and to help you?

Zulfiya: 31:54

I think the number one thing is what is it causing on a cash flow? What is the financial impact of this? And All right.

So with the QA, QC process, if they have quality assurance process now delaying the submittal of the invoices and how much money this is costing them, and that could be calculated. If you take the invoice times the days of delays and... It's a huge cash.

Mark: 32:19

Impact. And then measuring success of an impact after a project, after you run an AI project, is quite simple using the same cash flow impact. Has that changed? Has there been a difference?

Zulfiya: 32:30

Yes, exactly. Exactly.

Mark: 32:33

Cool. I think the biggest challenge that we see, and while you deal with a lot of construction companies, but in larger companies is data, right? It's very nice to have clean data to have a perfectly clean database that is never the case or at least not in the business that i've seen How do you go about addressing that in the companies that you work with? And what would your advice be to CIOs or CTOs and larger businesses?

Zulfiya: 32:57

This is actually an example of QA/QC, such a good example that we can bring up because the data is created by us humans, right? So we insert the data in some way and shape and form. And when we work with the construction companies, for example, what we see is, well, we work with technicians that say on the field, they've done their job and now they're entering their information that turns into the invoice. But the information that they might be entering is grammatically incorrect, is incomplete, is inconsistent with, let's say, there's guidelines and rules, right?

So now what happens, the financial side of the billing department cannot submit that invoice because first of all, they have to review it manually. Then they have to put together invoice based on it manually, let's say, right, to kind of clean it up because they can't send that to the client.

Well, Right there. Is data inconsistencies, you know, garbage data, garbage in, garbage out. This is where AI could help. This way AI could basically assess the information that came from the technician because they're busy. Their job is not to enter the admin work, basically. Their job is like, all right, I'm on the field. I assess what's happening. But my job is to kind of like fix it or propose some type of a solution to fix it, basically. My job is not to write eloquent things. Statements here So that's where AI could come in. Basically evaluate, all right, well, this is our protocol. For example, these are the you know, the rules that we need to submit our information. Is it matching it or not? Because it's really, AI is really great at pattern matching. Now it has seen so many of them and it could say, all right, well, the product, for example, the piece about the instructions are not complete. Can you please provide additional information?

So it can go back to it and actually clean up as well, clean up the writing. We do that very well, right? Copy and paste all this stuff into ChatGPT and rewrite it. That does really well. Summarize it. Put the nice heading to it. And it Then bring it. Back to the, for invoicing.

So don't, and have a human, for example, at invoicing now that they're experts and know what information should go in there, quickly evaluate. And then as your thought process now is so much smoother. And I think as we talk about AI cannot run on chaos and AI cannot run on systems that don't have APIs because there's no back end access. Right.

So like there's nothing you could do. And actually we are dealing with that as well.

Like we want to help a couple of companies to automate, but their system does not allow. And there's still a lot of companies that use systems that there's no backdoor. For you to push and pull information.

So you're kind of stuck. And this is where you need to make a decision, okay, do you move to another system? What's the process of migrating to another system? Because, or does the system open the API at some point or not, but yeah. And data cleansing is very important because AI generates content based on the data that you give it to.

Mark: 36:01

Exactly. And that's so for integration, you depend on the APIs that are in those systems. And if they're not there, it's... Kind of hard do you ever use like the integrated api because almost every tool supplier these days thinks we're going to give you an ai agent or an ai interface that you can sort of work with? Do you use those tools as well? The integrated AI is in those tools?

Zulfiya: 36:21

Good point. So for example, about Vicky Voice AI agent, right?

So there's a jobber in construction that has the AI, go high level has the AI. But here's what we see with our clients. Our clients are They're kind of a little worried about turning that on without proper supervision, without proper, I'll say fine-tuning, right? About proper giving direction. Because again, that system is picking up the calls.

So you want to make sure you audit the calls. Making sure that there's safeguard rails around it that's not hallucinating, giving pricing that you probably never hold it to, you know, or give the services that you didn't ask or actually give like system prompt, for example.

So, yeah. What we see is, I almost like at the beginning I was kind of like my god like every com every software is going to come from day over the eye I was kind of like what am I going to help with right so I was a little worried but now the way I look at this a little more like wait a minute it's the same way as like with analytics you already came with their own analytics nobody uses them because it's not great because it's out of box and it does not speak to that client and it's The systems are so large, they cannot make it for you, knowing all your, let's say, pain points, all the processes that you run your company on.

So what we do custom for them. Or we fine tune them, kind of like, you know, adjust the ones first.

So I think the same thing with what I see with the AI as well. It's like every system comes with its own AI, but it comes out of package, so now you can integrate it. For example, Salesforce comes with their own force agents, but they still need to be integrated. You can't just like turn them on and now like go wild, basically go around wild.

Mark: 38:04

I love that analogy with integrated analytics and I think that resonates with me. It makes a lot of sense, that analogy. I can see how that works.

Zulfiya: 38:13

For example, right, Mark, if you run a NetSuite, same thing. There's a NetSuite whole consulting company, but NetSuite has the analytics.

So why does it need consultants? Same thing. You need integrators.

So the way I look at this, I'm a construction integrator, basically. That's where I'm That's right.

Mark: 38:28

At. Helping them get the best use out of their tools and adding additional tools if necessary.

Zulfiya: 38:33

But I'm not competing with Google's and, you know, and basically open AIs. I'm just using the tools that are out there that are given and then fine tuning these models now to have that domain knowledge and expertise.

Mark: 38:46

Okay. And that's my next question. I think that's very close to what you just said. If you're a leader in tech, right, you're a CTO, CIO, CTO. What are the skills that you need to build yourself, but in your team to stay competitive in that world?

I mean, they all come from... And needs. The NetSuite, the ICPs of the world, or these older systems, it's a complete new world. What is it that they need to learn or know or understand to stay competitive themselves as people?

Zulfiya: 39:14

I think obviously ChatGPT is number one, Google Drive Studio. I'll say go for the AI package. Understanding, like, Google basically, I'm just saying all these AI tools that are out there, let's say ChatGPT, Google Drive Studio, Notebook, LLM. And I know this is going to be like too much, but I think it's good to know what tools out there just even have a little bit of idea of co-pilots, right? What they are capable of. And know a little bit of wipe coding.

Well, for you now, for CTOs that have a technical background, wipe coding is almost like, well, both their python skills together and they get more use out of it they get more i would say that somebody has zero technical experience, still will struggle a lot with that. Because you can see, all right, well, how can I clean up my data? For example, how even for right now for us, when we're doing BI side, well, how can we use AI here to help with the caption? We write better queries. What's the process around data cleaning, for example, with tools out there from Microsoft, if our clients are Microsoft or my clients are on Amazon, for example, what are tools that we can use to kind of streamline those processes and keep it secure as.

Mark: 40:27

Well? Okay. I think that makes a lot of sense, but then still all the technology goes so fast.

I mean, even from just Google, you have a new model every month, every two months, and there's so many different providers. How do you stay up to date? How do you actually... Keep updated on everything that's going on out there, because it's ridiculous.

Zulfiya: 40:45

It is ridiculous. I would say what I'm trying to do as well is, okay, not to get overwhelmed with everything, is first of all, there's a benchmarking example on the LLMs.

So let's say there's a GPT-4, right, was now swapped with GPT-5. There's a nano version. Of GPT-5 as well.

So there's a website for benchmarking. You can see, okay, well, if this is out now, GPT-4 is not available. Before you move to GPT-5, what other model you can swap to? That are kind of performing on the latency, performing on accuracy performance, and kind of just, you know, make that decision that way. I would say, but for CTOs, Okay, we're looking at what environment does your company work in? Is it Microsoft environment or it's Amazon? Environment and now looking at the tools that are available within that environment and kind of going more a little bit deeper on that and see, okay, well, Is this useful or is this not useful? Is this just like extra noise, for example? And saying, all right, if you have something that's working, does it need to be swapped or not? For example, we have Power BI Microsoft. It's working great for our clients. It's performing, it's pulling everything well, that doesn't probably need to be swapped for anything else just because there's no reason for it. It's already giving ROI, the reports are consistent, but if you're trying to add now additional AI layer to it, well, what's available there can i turn on copilot now give me summaries on that report maybe what or do i want a deeper knowledge and deeper knowledge do i need to know maybe bring farmland for example so instead of charging so farming the way with farming is it's a knowledge base so it will be trained on the internal docs as well as for example you already database and you can query it. But here's the difference. Instead of you buying the ChatGPT licenses for each of the users, now where the fireman comes in, you basically it's a custom GPT that basically has an open AI agent behind it as an engine. And now, instead of you buying licenses for each of the users, you don't have to do that.

And then you can just restrict the, let's say, role of security. You can add that, receive accesses, and kind of it's a much more scalable solution.

files.: 1500

So Now you would have to buy 50 licenses for it. Each person to have access to that. But if you have solutions like ours now, Basically, you don't have to buy licenses and it's secure. It's on Azure hosted, for example, It's encrypted. There's a data backup in three different places.

So what I'm saying is much more scalable solutions now available with AI.

Mark: 43:54

Cool. But again, that requires the expertise of what's available out there.

Zulfiya: 43:59

That's right. That's right.

Mark: 44:01

Okay. The last couple of questions that I want to ask are more about leadership and mindset. You started from being an accountant and grown to running your own business. What's like one of the biggest mistakes or failures that you made that really, It was an incredible learning experience for you along the way.

Zulfiya: 44:18

My gosh. Just like the phone that I said, I think one of the lessons that I'm trying to learn in life is proper asset allocation. Time, money, energy. Proper asset allocation. And I think I'm like, I'm banging my head many times, one point. To learn that. First thing that I didn't get EA earlier. Second thing, I didn't get the outreach system in place. Meaning if you're trying to run a business, you have to have a, when you do yourself, then you pass it on to somebody and have an outreach. Maybe you have your EA, maybe someone else like another virtual assistant can help you with the outreach system.

So you're not doing all because it's impossible. I would say when I was hiring people, I was hiring based on likes.

Like, I like you. Okay, great. And you have a pulse. And you're telling me you can do this job? You're telling me? I think that was a lot because sometimes things don't work out. And I did not have that. Clear KPIs. I did not have a clear, let's say, job responsibilities. I did not know And I'm still learning that. Just FYI, I'm still learning those skills, but it's cost, what it costs on the other side, it costs delayed projects. It costs, basically, you know, Yeah, the lay project is huge because it's your reputation and every business that comes into you, it's like, it's a gold, basically. For me, it's the trust that is just worth so much to me.

Somebody trusted me with their business. I want to help them. I know I'll help them. But if the execution is failing, it's important. It's I think, as a founder, it's all my fault.

At the end of the day, it's all my fault. It's all my fault. And I think that burns a lot.

Like those that ingrow, like, Kind of like, I'd say those... This calluses, you know, that sometimes tell you like you want to throw a towel, like you don't know what the heck are you doing. You don't know what you're doing.

Like, you're not hiring the right people for the right jobs or, you know, things are not working out. Process gets delayed.

And then you don't have systems because you just... Your overstretched bin. And I think systems is number, proper asset allocation requires systems.

Mark: 46:37

Final question then, where do you see yourself, the Omadli group in five years from today? I'm five years is a very long time, but still.

Zulfiya: 46:45

Yes. Five years from today, I want to scale business in construction space. I want to be the expert. We want to be the expert in construction space as well, delivering great value, helping the construction businesses buy back their time because I know right now it's day time. Because with the AI, basically, there's so much demand. Data infrastructures that requires Buildings need to be built for all that code to be supported. And I know there's a lot of need for training compliance from that end, as well as processes that needs to be systemized and automated so they can scale and actually take those jobs as well on. And the way I see this is. We'll see if the agents are still going to be there, but we will. More pirate. We will pivot with the way of delivering the value for our clients. Okay. If there would be another technology private piece company, come next. It's just normal evolution. First there was what, XLS that came through, right? Then there was a database.

Well, Excel was kind of database. And now you have the SaaS and then you have basic old analytics and now we have AI.

So it's just, we are going to be evolving with technology with the goal of providing still-applied AccuTask.

Mark: 48:05

Thank you so much. It's been incredible having you here, Sofia. Thank you very me.

Zulfiya: 48:08

Much. Thank you for having.