Mimica.

Somewhere in a Fortune 100 office right now, a claims handler is alt-tabbing between six windows, copying a policy number from one into another, doing it for the four-hundredth time today. Nobody wrote that procedure down. It lives only in her hands. Mimica is the company that decided this invisible knowledge was the most valuable - and most ignored - asset in the building.

Mimica builds process intelligence software. In plain terms: it watches how people work, then turns the watching into something useful. Its tools capture the clicks and keystrokes that make up a job and translate them into end-to-end process maps - a documented, data-backed picture of how work actually happens, as opposed to how a slide deck claims it does. The company runs out of London and Brooklyn, employs a team built around engineers and researchers, and in September 2025 raised a $26.2M Series B to chase a bigger prize: feeding all that ground truth into AI agents.

It's a quietly contrarian position. The rest of the industry spent two years arguing about model size. Mimica's wager is that the bottleneck was never raw intelligence - it was knowing what to do with it inside a real, messy, regulated business.

Nobody knows how their own company works

Here is an uncomfortable fact most executives would rather not test: ask ten people who do the same job to describe their process, and you'll get eleven answers. Work is full of undocumented shortcuts, exceptions, and "oh, I always do it this way" workarounds. The official process map - if one exists - is a polite fiction, usually drawn by a consultant who left two reorganizations ago.

This was a nuisance for decades. Then it became a crisis. Companies started trying to automate that work, and later to hand it to AI agents, and discovered they were automating a fiction. You cannot teach a machine a process you cannot see. The numbers bore this out: industry research the company likes to cite says 95% of generative AI pilots fail, and Gartner has forecast that more than 40% of agentic AI projects will be abandoned by 2027.

The traditional fix was to send in analysts and consultants who would interview staff, sit beside them with a notepad, and reconstruct the process by hand. This took months, cost a fortune, and produced a snapshot that was out of date the moment it was filed. Mimica's founders looked at that and saw a problem better suited to software than to clipboards.

Two engineers and a hunch

Tuhin Chakraborty and Raphael Holca-Lamarre met in 2018 in a London cohort of Entrepreneur First, the talent program that pairs technical people and dares them to start something. Chakraborty took the CEO seat; Holca-Lamarre, the CTO chair. Their hunch was specific: if you could observe work at the level of individual clicks and keystrokes, you could infer the underlying process automatically - no interviews, no guesswork, no consultant's clipboard.

The technically hard part is everything after the recording. A raw stream of mouse movements and keypresses is noise. Turning it into a clean, generalized process map - one that recognizes that "copy the policy number" is the same step whether it happens at 9am or 4pm, on a fast machine or a slow one - is a genuine machine learning problem. That inference engine is the company's core IP, and it's why the pitch isn't "we make screen recordings." It's "we make sense of them."

Investors warmed to it gradually, then all at once. Episode 1 Ventures backed the seed. Khosla Ventures led a $6M Series A in 2021. By 2025, Paladin Capital Group was leading a round more than four times that size, with Khosla, LGVP and Entrepreneur First all returning - the kind of insider re-up that tends to mean the early bet is paying off.

Miner, Mapper, Measure

Mimica's platform is, mercifully, named in a way you can remember. Three products, all starting with M, each handling one part of the journey from "we have no idea what's happening" to "here's exactly what to fix."

The selling point that makes operations leaders sit up is speed. Mimica claims it delivers actionable process maps in roughly two weeks, against the months a manual review would take, and without pulling internal analysts off their day jobs. The newer chapter is what those maps feed into: rather than just guiding old-school RPA bots, the maps now act as a training playbook for AI agents - the context that lets a model act reliably and compliantly inside a real enterprise.

There is a tidy irony here. A company built to automate repetitive human work is itself the automation of one very specific, very tedious human job: the business analyst with a notepad. Mimica essentially put that role out of a job by doing it better, faster, and without needing coffee.

Receipts, not adjectives

Process intelligence is the kind of category that attracts grand claims, so it's worth grounding things in numbers the company puts its name to. In the 18 months leading into the Series B, Mimica reported growing annual recurring revenue by more than 570%. It now serves over 30 large enterprises, including multiple Fortune 500 companies, across healthcare, logistics, financial services, insurance and manufacturing.

Named customers include Goodyear and ClearBank. The recognition column is equally specific: analysts have tagged Mimica a Leader and Star Performer in Task Mining (2023) and in Digital Interaction Intelligence (2024-2025). And on the partnership front, the maps don't live in a vacuum - they feed leading automation platforms like UiPath and Automation Anywhere, where the prioritized blueprints get turned into running bots.

Give people their time back

Strip away the enterprise vocabulary and Mimica's stated mission is almost old-fashioned: empower people to reclaim their most precious resource, time, by delegating repetitive digital work to AI. The founding idea was that humans should not spend their careers copying policy numbers between windows. The company's tagline for its larger ambition - "building a future free of inefficiency" - is the sort of thing that's easy to print on a wall and hard to actually deliver.

What makes the mission credible rather than decorative is the order of operations. Mimica doesn't start with the AI and hunt for a use case. It starts with the work - the real, documented, sometimes embarrassing reality of how a job gets done - and treats that as the precondition for everything else. Visibility first, automation second, agents third. In a market that mostly ran those steps backward, the discipline is the differentiator.

The map becomes the operator

The next few years will be a referendum on whether AI agents can do real work inside real companies. The optimists and the skeptics agree on one thing: the agents that fail will fail because they don't understand the context they're dropped into. That is precisely the gap Mimica spent seven years instrumenting. The map it draws of how work happens is starting to look less like documentation and more like a job description a machine can read.

There's risk in it. Recording employees, even with privacy controls, sits on sensitive ground, and the company will live or die by how seriously it treats that. The category is crowded - Celonis, UiPath, Microsoft and others all want this territory. None of that changes the underlying bet, which has only grown more relevant: in the agent era, whoever owns the ground truth of how work is done owns the most valuable input there is.

So back to that claims handler, alt-tabbing through her four-hundredth policy number of the day. In Mimica's version of tomorrow, the software watched her do it once, learned the pattern, and quietly handed it to an agent that does it right and gives her the afternoon back. The hands that held the only copy of the process finally get to do something else with their time. That was the whole point.

Demos & interviews