For B2B SaaS teams about to spend a quarter on an unvalidated bet
Your team builds less. The metric moves more. The board sees why.
Start here · one week
Discovery Reality Check
€900
100% money back if it didn't get clearer.
01 · The situation
That's where I come in — inside the team, as part of the team, not as a deck delivered from a distance.
02 · Nice to meet you
Product discovery & AI product · Warsaw
I run product discovery and AI product operations for a living. Until September 2026 I did it at RedTrack, a MarTech attribution platform, as Product Discovery Lead and AI Champion — where I turned discovery from a phase into a company function and put 25+ AI agents into daily production. I now run this practice independently, from Warsaw.
Before that: CPO at Roketo through the 2022 Web3 downturn, product lead on an AI wellness MVP shipped in two months, and eleven companies across four industries.
Before product, I trained as a lawyer — which is where the habit of not believing a claim without evidence comes from.
I work with two clients at a time, hands-on, inside the team.
I help B2B SaaS product teams stop building on opinions — by installing discovery as a function and AI agents that run in production, not in demos.
I don't hand over a research report. I install the function that keeps producing them after I leave — and the agents that run it every day.
03 · Imagine
04 · How I help
For teams where research happens, and the roadmap still runs on opinion.
For the quarter-sized bet currently sitting on the table.
For "we need AI" with no line yet between the useful and the theatre.
Not sure where you fit? Start with the one-week check — that is what it is for.
05 · Proof
Two of these are running. One I killed after a month. The third one is the one worth reading.
RedTrack · support team of 8 · 140 tickets a day
Support was answering the same questions by hand and escalating anything that needed a look at the data. I put an agent in front of the queue that answers directly, and gives the human the investigation already done when it has to escalate.
−1.5 FTE in the first month, −2.6 in the second · first response 50 min → 1 min · investigation 2 hours–2 days → under a minute
RedTrack · discovery installed as a company function
First the function: a request standard at the front of every hypothesis, a live backlog, risk memos before big bets. Then a flow of 12 agents that runs the desk half of a deep study end to end.
Desk research 40–60 h → 4 h per study · 85% of hypotheses killed before engineering started · customer interviews stayed human, with two tracks of them partly automated
RedTrack · shipped, measured, killed in a month
A proactive agent that watched the user's campaigns and pushed personalised alerts, with a companion persona on top. There was no behavioural precedent in the category, so I took the mechanics from an adjacent market with the same underlying Job: day traders on the stock and FX exchanges, who also sit over a live number and act on an alert.
In a month it never reached enough users for a statistically significant read — so there was no verdict to have, and it stopped.
What the dead agent cost, and what it bought
The expensive mistake was mine and it came earlier than the launch: the first success metric was the number of interactions with the agent. That measures whether an agent is entertaining, not whether it is useful — and it would have kept looking healthy right up to the point where nobody changed anything because of it.
I replaced it with two: the number of real changes the agent caused a user to make in the product, and the time from the alert to that change. Those are the two numbers I now set before an agent ships anywhere.
A month bought that lesson. Finding it out after a full launch costs a quarter — and by then the metric is in a board deck and much harder to take back.
From founders and executives I've worked for. This practice is new, so these are references rather than client testimonials — the first ones will be published as they arrive.
«The way he frames problems now — that's strategic-level thinking.»
«He is the kind of person who turns complex and risky startup hypotheses into functioning businesses.»
«When budget cuts hit Roketo, Vlad secured grants, retained the team, and led our pivot as CPO. A highly reliable product leader.»
06 · Working together
Handing part of your product judgment to someone you don't know is a big step. So the first step is small, fixed-price and refundable.
One week
€900
The right entry point when the roadmap is full, the team is shipping, and nobody can point to the evidence behind the top three bets.
In five working days
You walk away with
100% money back if after the debrief you don't agree it got clearer.
Two weeks
Where AI pays inside your product and your operations, which three agents to build first, what each costs per user per month, and — explicitly — what not to build. A prioritised build list with unit economics attached, not a maturity model.
€2,500
One day a week · minimum three months
I own the discovery function and the AI product line: what gets built, what gets killed, how quality is measured, and how your team learns to work this way. Two clients at a time, maximum.
from €5,000 / month
07 · Q & A
You'd find this out in five minutes anyway, so here it is first: I left RedTrack in September 2026, I run this practice from Warsaw, and I am also open to the right full-time role. Both are true at the same time, and I would rather you heard it from me.
For the fixed-scope products it changes nothing — the one-week Reality Check and the two-week AI Payback Map are bought and finished inside a month.
For the fractional retainer it is a real question, and the answer is built into how the offer works. Three-month minimum. A written handover of everything installed — the standard, the backlog, the flow, the agents — is a deliverable, not a favour at the end. And I finish an engagement I have started: nobody gets left mid-quarter.
The whole premise is that the function keeps running after I leave. If it stops the day I stop showing up, I did the job wrong — and that is equally true whether I leave for a job or because the three months ran out.
One day a week, fixed in the contract — an agreed weekday block. Two clients at a time, maximum. The limit exists so both get a real day rather than the leftovers of one.
If you need someone five days a week, you need a hire, not me — and I'll say so on the first call.
No. Two clients at a time, in non-competing niches, and I say no to the second one if there's overlap.
That's also why the entry product is a week and not a quarter — you find out cheaply whether this works.
No. It's me. When a project needs a designer, a data engineer or a researcher, I bring in people I've worked with — and I say so before, not after.
B2B SaaS product teams, 20–150 people, Series A/B, with a product that already sells and at least one product manager. The typical trigger is a quarter-sized bet on the table with no evidence behind it, or "we need AI" with no idea where it pays.
Not for: pre-product startups and ideas with no revenue, outsourcing shops, and anyone who wants a research report for an investor deck. That is a different job.
An agency hands you a report and leaves. I install the function that produces the reports after I'm gone — the standard, the backlog, the flow, and the agents that run it.
If the only thing you have three months after I leave is a PDF, I did the job wrong.
Usually because the pilot was measured by "does it answer" instead of "what does it cost, where does it break, and who notices when it's wrong".
I put logging, evals, confidence thresholds and a cost-per-user model in from day one. That's the difference between a demo and something your team is willing to depend on.
Advanced Jobs To Be Done and Next Move Theory — the Zamesin canon, not the generic Christensen version. 300+ customer interviews so far.
In practice it means we segment by Core Jobs and success criteria, we rank risky assumptions before we build, and we treat every feature as a value hypothesis with a cheapest-possible test.
Based in Warsaw, CET. I work CET ± 3. English and Russian; I've also delivered a conference talk in Polish.
08 · Ready to start?
There is no one-size-fits-all here. If it turns out you don't need me, I'll say that on the call.
Book a free 25-minute callTools I work in