Six Months of User Research. Done in Minutes.

01 /The Concept

Lakmoos claims the era of instant user research has arrived – and backs the claim with an approach that reframes the question entirely.

Instead of studying real users, companies study AI personas that represent those users. Each persona lives on the Lakmoos platform and acts as a composite stand-in for a specific audience segment. Teams can give each persona a name and a visual identity to make conversations feel more natural.

Persona depth is built at two levels. The foundation is assembled by Lakmoos itself, drawing on external data and surveys to represent a wide range of audience segments. On top of that, each client adds their own layer – proprietary survey results and customer data that sharpen each persona's accuracy for their specific market.

Traditional research requires recruiting hundreds or thousands of respondents to achieve statistical significance. With Lakmoos, a team selects the relevant audience segment using up to a hundred filtering criteria and gets back a single AI persona ready to answer questions immediately.

The platform's data sourcing goes deep. Lakmoos holds aggregated, anonymized information about financial behavior across demographic groups – so asking a persona "how much do you typically keep in savings?" returns a statistically grounded answer. The same applies to questions most surveys can't ask directly: religious beliefs, risk tolerance, purchasing anxieties. Because the persona isn't a real person, there's no social desirability bias in the response.

Those emotional dimensions matter. The platform doesn't just return a factual answer – it maps the emotional landscape around it. A persona saying "I'm not planning to take on a mortgage" might reflect contentment with current housing, or it might reflect quiet frustration at not being in a position to plan at all. The distinction changes everything about how a product team should respond. Lakmoos surfaces both the answer and the sentiment driving it.

Lakmoos claims that its AI personas' responses align statistically with responses from real members of the same audience segments – a finding validated by the research firm Ipsos.

For context: setting up a traditional user research study takes about six months from objective-setting through vendor selection, question design, fieldwork, and analysis. In that same window, a team using Lakmoos can run multiple rounds of research and launch faster than competitors.

Clients pay a subscription fee for platform access. Lakmoos takes roughly a month to build the first version of personas calibrated to a client's target audience; updates and improvements then happen automatically as the underlying dataset grows.

A second client type is research and marketing agencies, who can run Lakmoos under a white-label arrangement for their own clients.

Lakmoos was founded in Prague in the spring of last year and has now closed its first €300K in funding.

02 /Why It Matters

A [recent review](/review/horoshaja-jekonomija-horosho-prodajotsja) covered Rally, which raised $8.85M for a CRM built to make user research faster and cheaper. Rally's funding signals genuine market demand for more efficient research. Lakmoos is solving the same problem – but through a completely different lens.

Rather than streamlining traditional research processes, Lakmoos replaces the process itself with something that only became viable through advances in AI.

The approach isn't as radical as it first looks. Earlier this year, [a review](/review/za-takoe-obuchenie-kompanii-tochno-zaplatjat) covered Hyperbound, which builds AI personas representing different buyer archetypes – a hard-charging VP of Sales, a methodical Marketing Director – so that B2B companies can train their sales teams on realistic simulations. Hyperbound came out of Y Combinator.

Lakmoos takes one step from that: instead of personas for sales practice, it creates personas for market research. B2C companies get synthetic stand-ins for their target customers rather than actual customers.

Zooming out, a broader trend is visible: AI personas representing specific types of individuals. Lakmoos and Hyperbound both build composite archetypes. Dopple, [covered previously](/review/individualnost-cepljaet-i-prinosit) last fall, goes the other direction – a marketplace of AI personas modeled on specific individuals: public figures, historical characters, fictional heroes. Users engage with them for advice, support, or companionship. The appeal is precisely that each persona has a defined personality and real limitations rather than being a generic oracle. Dopple raised $1.88M.

03 /Opportunities

The trend taking shape is the development of AI chatbots, assistants, and personas with distinct identities. That's the broad direction.

Applications vary widely.

Personal AI, [covered](/review/vzorvat-rynok-obrazovanija) early last year, built a platform for creating personal digital twins – originally pitched as a messaging proxy that responds on your behalf. The more obviously valuable application may be in online education, where instructors could deploy their own twin to handle the repetitive, time-consuming questions from students. That startup raised $13.7M.

Chirper, [covered](/review/revoljucija-podkralas-nezametno) last summer, built a Twitter-like social network where only AI personas interact – no human users at all. Someone creates a persona with a brief backstory, and the platform generates a full profile that posts and comments autonomously. It reads like a toy without a clear use case, but it raised $750K, and the underlying concept of synthetic social interaction has obvious unexplored applications.

Digital employees are also multiplying. Artisan AI ([related review](/review/trebuetsja-cifrovoj-sotrudnik-s-opytom-raboty)) is a YC company in this space. 11x raised $2M for digital workers in sales, recruiting, and support roles.

Giving those digital employees distinct personalities is an underexplored angle. Different people respond to different communication styles – and when hiring human salespeople or support reps, personality fit is explicitly evaluated. There's no reason the same wouldn't apply when "hiring" an AI employee.

The design challenge is specificity: in which domains does a strong individual identity actually help a persona accomplish something a generic assistant cannot? The answer determines what data you need to build it, and which use cases are worth pursuing first.