170 Enterprise Leads, 80 Investors – and They Never Ran a Single Ad

01 /The Concept

A [recent review](/review/dlja-rosta-pribyli-ne-hvataet-obratnoj-svjazi) covered Ontora back in April, when the company was still going through Y Combinator. What's happened since is worth paying attention to.

The startup has received over 170 inbound inquiries from potential customers – without running a single sales campaign or ad – and has already started working with five companies. Alongside that, more than 80 venture investors reached out proactively, with some depositing $750K before the company even announced it was raising.

Why are companies knocking on Ontora's door unprompted? Because the same scenario keeps playing out.

Companies know they need to do "AI transformation" – plug AI into their business processes. So they buy enterprise Claude licenses, distribute access to employees, and watch those employees use Claude to draft emails they then send manually.

The problem isn't the AI. It's that companies don't know where or how to apply it. Sophisticated companies bring in consultants – who say that just mapping the existing processes to define the automation scope will take four to five months and a matching invoice.

Why so long? The real play here: the vast majority of companies have no formal documentation of their processes. And if they do, it's outdated before the ink dries.

You can't automate something you don't understand. And 80% of that operational knowledge lives in employees' heads, not in any document.

Consultants spend those four to five months interviewing staff, extracting knowledge, and assembling it into formal specs. Then add another twelve months to program and test the automation of those specs.

Ontora compresses that entire timeline from months and years down to days and weeks. The platform works across three layers:

- AI agents interview all employees to understand what they do in normal operations and how they handle edge cases.

- The agents compile this into a unified knowledge base. Where employee accounts conflict, agents run follow-up interviews or escalate to management to resolve the discrepancy.

- AI agents are then inserted into processes, drawing on the knowledge base to understand what they're supposed to do and how.

What keeps the system from going stale is continuous monitoring: the platform tracks what humans do, what AI agents do, and where humans correct what AI agents did. That feedback loop constantly updates the knowledge base to stay in sync with how the business actually operates.

Ontora's early customers span a wide range – from a US private equity fund using the platform to drive AI automation across its portfolio companies to a European media company with 800 employees serving clients like Volkswagen and Bosch.

02 /Why It Matters

Ontora is not alone in this space – not even close.

But every company in this category is catching serious momentum. A few recent data points.

Jedify raised $24 million for a platform that builds a "context graph" of a client's business – giving AI agents the operational understanding they need to function inside it.

Edra ([covered here](/review/buterbrod-s-bolshimi-dengami)) appeared in March with a $30 million raise for a platform that "learns how a business works, then automates it" with AI.

These startups use different techniques to capture process knowledge – from AI-conducted employee interviews to AI agents that observe employee actions directly. They also organize that knowledge into different structures: knowledge bases, semantic layers, context graphs. Each claims theirs is best.

The specific technology is, in some sense, beside the point. What matters is the unexpected insight that's surfaced from all these AI transformation projects: the hardest part isn't automating processes. It's figuring out what to automate and how – because understanding what a company actually does turns out to be harder than it looks.

One more thing worth noting: these platforms differ from consultants in a critical way. Consultants are brought in periodically – because the work is too slow and expensive to do continuously. Platforms like Ontora need to run permanently, tracking what's happening in real time and keeping process descriptions in sync with evolving reality. Because the moment the knowledge base goes stale, AI automation starts breaking – agents operating on outdated rules.

03 /Opportunities

AI adoption is either already underway or imminent for virtually every company. That creates a massive market for AI process automation platforms – making this a very well-timed direction to move.

That said, platforms built on the model of "describe your processes, we'll automate them" won't cut it. The overwhelming majority of companies simply cannot describe their processes accurately. Without the ability to discover and continuously update that information, those platforms will find no buyers.

Most startups in this space are going broad – building universal platforms that claim to work for any business.

That's worth questioning. Specialized platforms, purpose-built for a specific set of processes within one vertical, should deliver better results on both speed of deployment and quality of automation. Selling into a niche is also more tractable, and first-mover positioning in a vertical can be durable.

Building a specialized process analysis and automation platform for a specific niche seems like the more interesting path. Faster customer wins, better outcomes, and the chance to become the default solution in that niche.

Which niche would you take a platform like this into? And honestly – what's stopping you?