How to do GTM for an AI-powered product in 2026

This is the second interview of the series, where we explore Birch’s transformation into an AI-powered platform. Find the first one, which focuses on building Birch AI from idea to beta, here.

Can you name a software company that isn’t taking its AI product to market this year?

A few months ago, Liliya, a product marketing manager at Semrush, shared their journey and some thoughtful insights with us. But every company is making its own bets on positioning, value propositions, and promises about what their AI solution will deliver, ranging from taking someone’s job to solving a real business problem. In this interview, we’ll talk about what shaped Birch AI’s go-to-market strategy.

Dasha is in charge of product marketing at Bïrch. The first question was simple: what has actually changed in this work over the past six months? Turns out, it’s not the part you can see.

Positioning is a document

When we talk about a go-to-market strategy, people often think about two types of work.

There’s the foundation: ICP, positioning, narrative, storytelling. This work is often invisible. Most people on the team and everyone else couldn't care less about it, because it’s how you achieved the result. Nobody cares how much time you spent developing the positioning. Positioning is a document.

And there’s a checklist of actions: send an email, publish a banner, run a webinar. These artifacts are visible, they’re easy to evaluate. The checklist part hasn't changed, it's more or less the same. The channels through which we interact with the audience haven't really changed that much. Yes, some things got added, for example, the importance of being indexed by AI. Before, when we talked about go-to-market, we always said: you don't have to do SEO right away, because it will get indexed in three or six months. AI works faster. You publish a LinkedIn post, and it can show up the next day. And disappear three days later.

But the other part, the invisible one, the one about the foundation, has changed a lot. All this groundwork: research, audience, messaging, storytelling, that whole story got turned upside down. And it keeps changing very fast.

A sweet spot that lasts three days

Before, this work took an incredibly long time. Take a SWOT analysis or any similar framework; you’re always looking for where your sweet spot is, and in a way that keeps that sweet spot just as sweet in the near future. Now it's become much harder. You can find that sweet spot, and three days later, nobody believes in it anymore. Or the advantage has evaporated, and you’re a fool for picking it. Or some big player released their own AI that's cooler than you, and they took it. And that, in fact, was one of the core moments of doing go-to-market.

How do you feel in this instability? What do you lean on when you need to draw the line and say: this is what we’re going with?

I think I have the right childhood traumas, so I feel fine.

I feel comfortable with change. And I still use frameworks. When you’re launching something, you have very little data at the start and need to act under high uncertainty — without a framework, you won’t be able to do it confidently enough. Sure, you can go with your gut feeling, but I'm not that kind of person. I’m a systems person, so to me, instability means you have to revisit more often.

There are a few life hacks here. Revisit more often and use AI to do the research and bring it to you. And the second one — let go of your decisions more easily. I can be very strongly convinced of something, genuinely emotionally attached. But if I see it’s no longer relevant, you have to admit: okay, that’s it, now it’s this way. These two things help: process faster and be less attached to what you came up with before.

What it looked like in practice: the launch of Birch AI, from hypotheses to the beta.

Two hypotheses

We followed a more or less standard framework, starting with use-case hypotheses. We had several ideas, none of which are super original, because they mostly leaned on the market.

That’s an interesting moment, by the way: if you think of it simply as a product launch, then most likely all these use cases will repeat other players. Everything in our domain is very automated, very secure, scalable, and so on, everyone says the same things. But if you think of the launch within the ecosystem — aimed at the users you already have, with the product you already have — a particular angle opens up. On one hand, it strongly narrows the value the market will see. On the other hand, it makes it genuinely workable. If we tell people who don’t use Birch about this value, for them, it will be much smaller. But for Birch users, this value lands perfectly, it’s a bullseye.

So, the hypothesis, which isn’t that original: small businesses often lack working hands; they genuinely need an analyst who will provide ad hoc answers to analytical questions. A super clear use case. I’m sure we all do this: I have an urgent question, I need an urgent answer.

The second use case is directly connected to Birch. It’s the analysis of your automation. People have already set up their automation, and they inevitably have missed opportunities and some mistakes. This is genuinely hard to catch — we’re all human, and everything changes fast. Here, you need a helper who will look from the outside and point things out. This use case is much narrower, and it’s hard to find a company here that will compete head-to-head with us because it’s our automation. There are no companies that do the same automation we do.

And this use case ended up working better. We launched the beta as a closed group of users; we monitored activation and asked for feedback, then gradually opened it to more and more people. As a result, we have substance — what exactly Birch AI helped people catch in their automations. Based on that substance, by the way, we’re now writing LinkedIn posts.

So the GTM strategy also became customer research.

Yes, fair to say. Also, I believe use cases are becoming even more important in a launch now. Because that’s the only way we can genuinely show what people achieved by using the tool. You can write absolutely anything. And in terms of functionality, now you can also build practically anything. But not all of it will actually be useful.

Trust as a funnel

The next phase that opened up is recommending automated rules. Birch AI first found them and then recommended what to do. And here, one moment was important to us: there is a certain distrust towards AI, especially when we’re talking about money. Few people will outright trust AI with their budgets.

How it works: it suggests rule drafts, but doesn’t switch anything on itself. A person has to click to send the rule live.

So, if you look at it top-down, the go-to-market went through building trust. First, we highlight a mistake, and we don’t even say you need to act on it. If people agree with the opportunity that Birch AI highlighted, they can act on it. If they see the result, then it's easier for them to accept what Birch AI suggests. That's the flow.

From here, a straight path to the question of positioning: what it’s built on, and what it deliberately doesn't contain.

Granularity as the base, "no promises" as the constraint

One of the main elements of everything we say and how we say it — we don’t make empty promises. Because predicting the future is impossible, and doing other people's business for them is impossible, too.

But those are rather constraints, we always apply them. The base is in granularity and in the focus on how the reality changes for the person who uses the product. And the constraint is to never overpromise. Like: okay, Claude, write me a SaaS that will bring me ten thousand euros a month. Don’t ask any follow-up questions.

I guess another thing here is that many AI products are only appearing now, they have no history with customers before this. We do.

Are you deliberately betting on granularity and use cases? Why?

Yes, I am. Because I’m a bore and I don’t believe in hype. I’m not a super dreamy person, I’m more of a data person. When I see how people interact with the product, what exactly they do,  that’s data I can lean on.

It seems to me that good B2B SaaS differs from bad B2B SaaS in being more responsible. Clients must be able to rely on us. B2B SaaS takes on a certain guarantee. Not necessarily written into the contract, but that’s how the world works: there’s a social agreement that when I pay for a product, there’s a guarantee. And how can you promise something nobody has achieved yet? We all see it on LinkedIn. That’s not my thing. That’s the difference of B2B SaaS: there has to be substance.

Dasha keeps a reference shelf: examples of how AI products go to market. I asked what patterns she sees, besides the obvious "AI will take your job."

The reference shelf: hype, GitHub for marketers, and buzzwords with a shelf life

Hype actually creates so much toxicity. There’s this company — Artisan — that had the scandalous billboard "Stop hiring humans." Of course, they later published that it was a provocation. But apart from their explanation, the narrative itself still exists: that you should hire not people, but AI. And as a human, you now have to do the work of ten or even a hundred people, otherwise that’s it, you're not qualified. It seems to me there’s a big push right now toward being more technical. Someone on LinkedIn is already saying that if you’re a marketer and you don’t have a GitHub full of your own pull requests, you’re falling behind.

Related to that is how products go to market: many launch through education. There are now far more tutorials, demo days, and daily onboarding sessions that teach you how to build their product into your workflow. That’s cool, that’s useful. What’s not cool is when, under the wrapper of education, it’s just marketing and sales.

There’s also a trend I don't like: that any positioning has to contain the words AI-native, AI-powerful. I even once made a timeline of these buzzwords. They really change very fast. And if you use a slightly old word, you're not cool anymore.

And one more trend: I haven’t seen it in the data, but I rather felt it through individual companies. How it used to be: there was an agency that did everything for you. There was a B2B SaaS where you clicked sign up on the landing page and started using it. And now some companies are doing a semi-hybrid model: they use their own technology, but the approach to the client shifts toward the agency side, where a huge amount gets set up for you. They position themselves as B2B SaaS, but there’s such deep custom configuration per client that it would be hard to show a demo applicable to all clients. It’s interesting whether this will develop further. I think it should, because it's a good way to compete with what people want to build on their own. When people choose between B2B SaaS and their own AI-something, the arguments they usually bring up are customization and price. Price, as we know, is a stupid argument. So customization remains. And if customization is possible in B2B SaaS, it immediately becomes much easier to explain.

Secret question

I asked everyone in this interview series the same question: what does AI do genuinely well in your personal life? Dasha answered it in reverse — why she is removing AI from her personal life entirely.

There are a few things. First, I'm very sensitive to text. I read a lot, and how something is written matters to me. When I started noticing that in my ordinary, non-work life, I see a lot of generated text, it really started to depress me. I don't want to see it on my social media feed, my personal space.

The second reason: I don't want to rely on AI so much in the moments when it's important for me to make my own mistakes and my own decisions. I want to consciously say: this is my decision, I understand it may be wrong.

And the last one is a bit strange. It seems to me AI often takes away our reason to talk to another person. Before, if something hurt, I always texted my mom. As we know, moms know everything about that subject. And then, at some point, I started asking the chat about it, and one of the reasons to text my mom disappeared.

Here's a more recent example. I have a friend, she works in aviation. Sometimes I see an interesting random story from aviation. I can ask the chat why it happened, or I can ask my friend. The chat will most likely give me a better answer, but the reason to talk will disappear. It seems to me that reasons to talk are very important, so I'm removing it from personal interactions as much as possible. I don't even remember the last time I used the chat for anything personal.

I used to ask ChatGPT to help me write difficult messages, for example, to support a friend who got laid off. Over time, I realized I don't want to outsource that process: difficult communication teaches me to engage with another person not only from what's good, but from what's hard. Including making a mistake, saying something dumb, and finding out only once it's already been said.

Yes, exactly. And you caught that thing about "making a mistake." We often use AI, thinking it will either take the mistakes on its own, like it wasn't you who made the mistake, it was AI. Or that it will definitely give you the right answer.

I believe that’s not the case at all.