Skan AI sees the work nobody else is watching.

/ 01 - The SceneTwelve browser tabs and a sticky note.

It is a Tuesday morning at an unnamed top-ten US bank. A claims analyst opens her laptop and begins her day exactly the way she did yesterday - and the day before that, and every Tuesday since 2019. Twelve browser tabs. Three internal apps that don't talk to each other. A sticky note pressed to the edge of her monitor with a checklist nobody put in the SOP.

None of this work appears in the bank's process diagrams. None of it is captured in the BPM tool the firm spent eight figures licensing. None of it shows up in the consulting deck the COO is reviewing two floors up. And yet, this is the work. The actual one. The one that pays.

Somewhere in Menlo Park, a piece of software called Skan is, in fact, watching.

/ 02 - The ProblemThe map was always wrong.

For two decades, enterprises tried to understand their own operations through three blunt instruments: interviews, workshops, and database logs. Interviews capture what employees remember saying about their work. Workshops capture what employees say about their work in a conference room. Database logs capture what the database happens to remember. None of these is the work.

Traditional process mining - the field Skan would later disrupt - sat on top of those database logs. It was useful, mostly, until you noticed that an enterprise process spans dozens of applications, half of them legacy, and nearly all of them stingy with their event data. The process mining tool would draw you a beautiful map. The map would be missing the parts that actually broke.

Robotic process automation arrived next, promising to bottle that work into bots. It worked, sometimes. It also failed, often, because the bots had been pointed at the wrong process - the one in the diagram, not the one on the sticky note.

A small irony

The companies most desperate to digitize their operations had spent the least time observing them. They had budgets for transformation. They did not have a microscope.

/ 03 - The BetTwo friends from Kanpur build the microscope.

Avinash Misra and Manish Garg met in elementary school in Kanpur, India. They went on to be classmates at IIT. They built a company together, sold it to Genpact, and then - in the kind of move that only people who genuinely like each other make - decided to build another one.

Their bet, in 2018, was unfashionable. While the rest of the market was pushing harder on database event logs, Misra and Garg argued that the only honest way to map enterprise work was to watch the screen. Pixels. Clicks. Application switches. Pauses. The boring stuff. The real stuff.

They called the resulting category "dynamic process intelligence" - a phrase that sounds like it was workshopped, because it was - and built a platform around it.

/ 04 - The ProductZero integrations. Total observation.

The Skan platform is, in concept, simple. It uses computer vision and AI to watch what employees do on their machines. It does not need an API into the underlying applications. It does not require IT to crack open a single legacy system. It treats every application - SAP, Salesforce, that 1996 mainframe nobody talks about, the spreadsheet on a shared drive - as a flat plane of pixels and behavior.

The output is a kind of x-ray of an organization. Every click mapped. Every handoff timed. Every variant - and there are always more variants than anyone in management believes - laid out side by side. The first time most COOs see a Skan dashboard for their own department, the reaction is the same. A long pause. Then a quiet "huh".

/ 05 - The MilestonesFrom sketch to category.

/ 06 - The NumbersReceipts.

Enterprise AI is a category in which everyone has a deck and almost nobody has results. Skan's pitch is unusually quantitative, in part because the product itself is built to count things.

/ 07 - The ProofBanks, insurers, hospitals - quietly.

Skan does not publish its full customer list, which is a habit most enterprise software companies acquire once their customers are large enough to have a procurement department. What is public is enough: more than twenty of the world's top financial services, insurance and healthcare companies use the platform. Gartner reviewers, an audience not famous for their patience, give it warm marks in the process mining category.

The investors map onto the customer base. Citi Ventures wrote a check; Citi the bank is, for obvious reasons, deeply interested in process intelligence. Dell Technologies Capital led the Series B; Dell sells into roughly the same enterprises Skan does. Cathay Innovation, Zetta, GSR, Liberty Global, Firebolt - a cap table that reads less like a fund list and more like a strategic chess move.

/ 08 - Why It MattersAgents need ground truth.

For most of the past five years, process intelligence has been a back-office category. Useful, but rarely the subject of a Monday morning all-hands. That changes when agentic AI shows up.

An AI agent deployed into an enterprise has the same problem the consultant had in 2014: it does not know what the work actually looks like. It has the org chart. It has the policy document. It does not have the sticky note. Without a Skan-style observation layer underneath, agents tend to do what bots did - automate the wrong thing, beautifully.

The Context Graph of Work is Skan's argument that observation is the substrate of AI deployment. It is not the loudest argument in the AI conversation. It might be the most useful one.

/ 09 - The Scene, RevisitedThe sticky note has been seen.

Back at the bank. It is now a Tuesday morning in 2026. The claims analyst opens her laptop. Twelve tabs. Three apps. The sticky note is still there.

Two floors up, the COO is reviewing a different deck. This one is not a consulting recommendation. It is a Skan dashboard. The variants are mapped. The bottleneck is named. The automation queue is ranked by measured time saved, not by which team shouted loudest at the workshop. The sticky note - or rather, the workaround it represents - is now part of the formal process.

Nobody calls it a revolution. It is the opposite of a revolution, really. It is just the work, seen clearly, for the first time.

Which, in enterprise software, is a more radical thing than it sounds.

/ 10 - The FootnotesFind Skan AI in the wild.