Introduction: AI as a Thinking Assistant

We are going to talk about AI as your thinking partner and it's a weird positioning because when you think of a partner like it's certainly it's not supposed to be your virtual girlfriend or boyfriend or whatever that is and you know you start using AI you maybe you start asking it questions and maybe you go a bit deeper and at first things are going well and I don't even know

better analogy than this like it's great things are like you're on the beach or you are going to wherever you're going but then pretty shortly after things kind of going down fast you're running out of tokens connection you don't have internet or some bad things happen and uh it's terrible uh so that's not good that's not good and um when when i was preparing

Why “Partner” Is the Wrong Metaphor

hearing for this talk, when talking about or looking at it as AI as your partner, I don't think that is the right way to position it. I think it's not quite the right word.

I think the right word is an assistant, because this partner is a bit of a too much. I think it's a little bit too much to put in technology and elevate it to such a high level.

Story Time: Background and Context

So today, it's going to be the story time.

Hi, I'm Sergey. I work in marketing operations. I worked with Salesforce, HubSpot, Parda.

I worked on a creative side doing video production, doing content marketing, demand generation, brand work, and I spent the last 12 years in B2B.

For the last three years, I ran a video production company focusing on B2B and working in brand videos, animated videos, videos, and testimonials.

The Trigger: Losing a Major Client

And essentially what happened was that last year I lost an enterprise client. Might have happened to you. Not a good thing.

You are losing a serious chunk of your money. And so we need to increase sales. What are we going to do?

Using AI for Deep Thinking Sessions to Increase Sales

And I started using AI as something that to ask questions, but not simple questions, but multi -hour sessions to understand how to increase sales.

Building a Launch-Prediction Approach

And what I've done, I basically, I've created an algorithm to predict when companies would be doing product launches from my ideal customer profile, then I could hit them and I could actually sell the product videos that I was doing.

And all of it was good and it all made sense, except I've asked a question, I had an idea also around product videos, how to optimize the process and make them faster.

When you're doing that, when you're working on product videos, a lot of the stuff that you do is out of your hands because it's dependent on the B2B company

and their processes. There are multiple stakeholders, there are internal things that they do, and you have to play by the rules.

I had an idea how to do it faster, and AI tells me, it's the most important thing you said. I'm like, you know what? I think you're absolutely correct.

Why the Nobel Prize wasn't placed under my feet already? I got to be the best. I got to be so smart. And, you know, in this case, in this case, this was true.

Whatever, like AI said, like, you know, the most important thing that I said was how to make the production process faster and how to do it, you know, how to not spend three weeks or not spend four weeks and do it much quicker.

Reality Check: What Can’t Be Optimized

And, you know, what was very obvious is that you can't do that. You can't do that. It's a brick wall.

You can't because it's dependent on companies. And that means the business model, that means the delivery model is not changeable. This is it, there's no way to go from here.

Zooming Out: From Video Production to New Problem Spaces

So that led me to look at what other options exist, what other options exist outside of what I was doing specifically with the product videos, with animation videos in the industry I was in.

And essentially what I've done is I zoomed out and I look at all the problems in the video industry. With my experience, I could immediately tell I'm not interested in solving any of them. Zero.

I'm not interested. So that only took a few hours. And then after some time, I was like, well, you know what?

Why am I looking at the video industry? I can go way further back. Let me just zoom out. And what I've done is I looked at a wider set of problems,

Finding a New Fit: Competitive and Market Intelligence

and then I did an overlap with my experience, with what I've done before, with what I'm good at, with my connections network, and then there's certain overlap. And what I ended up was with with a system that does competitive intelligence and market intelligence.

I can talk about it, how it works in the Q &A, but essentially it looks at who your competitors are, how they're winning, how much money they're spending, and how you could apply it to your own company. That's like a simple explanation.

The Overbuilding Trap AI Encourages

But what I've done throughout that process, I spent more than a month on building the system and that took a while.

And of course I've overbuilt. Are you kidding? Of course I've overbuilt. because this is what it is.

AI encourages you to overbuild like crazy. You're spending tokens, you're building, you're building, you're building. And most people, I'm sure myself, are tending to overthink.

And even though I did do feedback sessions, I could have done it faster, more and faster. And I was getting some, but it's not, you know, it takes a while.

And, you know, what happened was I was speaking with a colleague of mine who was in marketing and she's running a smaller company, but essentially this was types of companies that I was targeting.

They're in life sciences or medical research.

And she said something around, it was a colleague of a colleague was working in AI and she was making like $2 ,000 an hour. And she sent me the website. I went, great.

Following the Money: Business Model Research

And then I looked it up and that girl ended up doing training. She's doing training for B2C. She's selling YouTube courses, TikTok model, Facebook group, all like standard model. But the point was, what really stuck with me was the concept of following the money.

Now, it sounds so ridiculously obvious, of course, like you're supposed to follow the money. But when you're in it, when you're building a system, when you already build it, when you're doing the validation, you're speaking with clients, it's not as obvious.

And so what it led me to do this thing is to go back and do more research. I looked at like 150 companies around the globe that are doing AI services. And I broke them down by different business models. I broke them down by go -to -market, by how much they're charging, who they're working with. And there's a very clear segmentation, very clear patterns that you could find on that.

This is obviously based on public information. A huge chunk of companies are fully invisible because they're just not online. Because this is the industry that is very, very much based on relationship, is very early, and a lot of it is they're not even posting because 90 % is referrals. So then you have quite a bit of limitation there.

Choosing a Direction: AI Services and Workflow Automation

And so what it led me to do is to find the ultimate solution. solution. Essentially, what I ended up finding was going into AI services and changing how people work with AI workflows, AI automation, custom solutions, all of that. So that's where I ended up going direction -wise after doing all of that research.

Knowing When to Stop Thinking and Start Acting

And it's supposed to be pretty simple, but it's when you're going through these thinking sessions, when you're going through all of this process, it's hard to know when to stop.

There's bias always. At what point do you go too much and you're just thinking and you're going into passive action?

That moment of active thing, you're doing something that is connected with the market to where you're just purely researching, purely having a conversation. It's a very tricky thing.

It's just very hard to do.

A Practical Technique: End Every Session With a 24–48 Hour Plan

What I tend to do now and what I recommend is if you're having a session with AI, if if you're thinking through a certain particular problem is to ask it, hey, like by the end of this thing, give me a practical plan, implementation plan

that is connected with the real world. What exactly do I need to do in 24, 48 hours?

It's like, not because, you know, we all build docs and cloud code or any other AI encourages you to build docs or documents or any of those things. But like, how is it actually connected with the real world?

How many people did you speak with? Because the amount of stuff that you could learn is insane. saying.

I was speaking with one person at a coworking and he had like a visual code, crazy number of files. And I spoke with him. I was like, I mean, this guy got to be like a very, very

intense developer. And then I'm speaking with him for 20 minutes. I'm realizing all he wants to do is just to help small businesses with websites and WhatsApp.

Why do you need like 250 files? You don't need that.

You don't need anything. You just need a landing page. You go to a few businesses, you sign a few clients, and you get going.

Then you build all other files. You don't need to build it. You really don't.

And so there's a very, very big thing with AI that encourages you to spend all this power. I love watching AI changing the code. It's beautiful.

I could watch it at night, in the morning, anytime. I'll come to you and we'll sit together and watch it. But

does it really help? And that's a tricky one. That is really tricky.

Validation First: Avoiding “Build and Ship” Blind Spots

And you know, this is such a great script. This is a great script from the movie. Like, Mr.

President, my philosophy is simple. Shoot first, ask questions later. This is so great.

You know, before AI, people were not doing validation, they were shipping or they were doing very little validation. Now there's so many more people doing that because why would you spend time on doing something that's harder to do and

it doesn't really scale very well. It requires to speak with people. You can just like shoot it and you can ship it and see what happens.

And that is really not exactly the way forward. It's a very tricky thing.

Like there's this concept like, oh, we should just build this MVP and send it. And it makes sense in certain ways. But a lot of times you end up overbuilding and you end up like not really going into like getting closer to the promised land.

It just doesn't happen. Like it looks like it is, but it's not.

And this is really, really, really important.

And so I love this after building quite a few projects. And this is my favorite step. It's the step before all the

other steps. Can we actually do this? Is it actually possible?

What are all the possible downsides? What is the fastest way to speak to somebody? Because I've built, like many of you,

I'm sure, 20 prompts that build into a shell script, that build into some other shell script, And it looks great, except do we even need to do that? So I really like this validation step to really understand some of the basics before jumping into the project and spending the weekend or spending a week or something like that.

AI Is Better Than Ever—But It’s Not Enough Without Humans

And the part that I want to come back to is, I mean, AI as a thinking assistant is extraordinary. It absolutely is. It could give you incredible suggestions.

I used it in 2024. It was not as intelligent. Like, you send real context, and I use, like, voice, and, you know, I send, like, 20 minutes worth of notes.

It's just not useful. It doesn't quite give you really what you need to solve real -world problems. In 2025, and certainly this year, it is.

Absolutely, it can give you incredible advice. But, you know, I don't think it is enough. I don't think it is enough.

Because, really, if you're doing services, if you're building products, most of the time, not always, you're doing it for people and you need to be connected to the humans and this is such an important part of going doing the unscalable stuff coming to rooms like these and speaking with other people and not just purely banging out

the code and then you know the software or whatever you're building I think it's very very important and with that there's a there's QR code and if you if a few a few things there that you happy to chat have to answer any questions.

Where This Landed: A Demo Project and a New Services Business

To come back to the story where I ended up basically I ended up doing a demo project for conference software that is based on the competitive intelligence methodology for this company that does like life sciences. That's where we are.

And then I've, like I said, I've launched the AI services, AI services business, still doing the video of production, but doing this new direction at the same time. So that's where we ended up with.

Q&A Topics: Memory, Context, and Session Carryover

And happy to answer more practical questions around when you are doing the thinking session, what are the good things to do, what are the things maybe you shouldn't do, how to save context, how to work with memory, how to do the carryover between sessions, all of those more specific steps. I didn't really spend too much time on that.

Conclusion

Thank you, guys.