It is May 2026 and somewhere inside a Fortune 500 finance department, an AI agent is about to make a decision worth seven figures. The agent does not need a smarter brain. It needs the actual terms of the renewal clause - the one buried in a 2017 master services agreement scanned in twice, named MSA_v3_FINAL_signed_FINAL2.pdf, and forgotten. Pramata is what reads that document. Quietly, for two decades, that has been the entire game.
Who they are now
A back-office company in the age of front-office AI.
Pramata sells contract intelligence to large enterprises. Comcast Business, Hewlett Packard Enterprise, NCR, McKesson, CenturyLink, Pitney Bowes - the customer logos read like a directory of companies whose legal departments measure their archives by the linear foot. The pitch is unfashionably specific. It is not "AI for everything." It is the patient, expensive work of turning a few decades of signed paperwork into structured fields an algorithm can actually trust.
Stop storing contracts. Start using them.
- Pramata's tagline, which has been correct since 2005 and only recently fashionableThe problem they saw
The banker's box of doom.
Praful Saklani tells the origin story without ornament. In 2001, he was selling a previous company. The buyer's attorneys, on the diligence call, asked to see the contracts. His team produced a banker's box. Inside the box, a representative cross-section of what every large company actually has: photocopies of photocopies, amendments stapled to wrong agreements, a Post-it that read "renews automatically?" with a question mark. He shipped the deal. The image stayed.
The contract is the truth of the relationship. Sales calls it pipeline, finance calls it revenue, customer success calls it health score. The contract is what the company is legally bound to do, and to whom, and until when. And in most enterprises that truth is - delicately - inaccessible. Pramata is the company that decided to make it accessible. The first version was less artificial intelligence and more artificial patience: people, software, and a stubborn belief that the data inside contracts was worth digging out.
The contract data nobody reads is the moat nobody builds.
- The slogan they would print if they were vainerThe founders' bet
Ten years bootstrapping, which is the boring word for conviction.
Praful Saklani and Christian Misvaer founded Pramata in 2005. They raised a Series A in December 2015. That is, by the calendar, a decade of operating a real software company without writing a single pitch deck for a venture firm. Most observers would call this slow. The customers they signed during that decade - the ones who renewed, and renewed, and brought their next contract problem back - would politely disagree.
Saklani's prior life is the giveaway. Before Pramata, he founded Yatra Corporation, which applied AI to travel management. Before that, Invotech Systems, a consulting firm. Before that, executive work at WaterHealth International, a social venture filtering drinking water in developing countries. The man has spent his career applying inelegant new technology to genuinely tedious old problems. Contracts were next.
The product
A platform that does the unglamorous part first.
Pramata's platform ingests legacy contract repositories - scanned PDFs, attachments, amendments, the lot - and produces a clean, structured digital archive. It extracts the parts that matter (parties, dates, obligations, payment terms, renewal triggers, special clauses) and exposes them through dashboards, search, and APIs. The platform integrates natively into Salesforce, so the renewal team sees the actual terms next to the actual account.
In November 2024, Pramata shipped AI Design Studio, which lets enterprises spin up custom generative-AI applications on top of that clean contract data - bespoke playbooks, clause libraries, deal-room creators. In March 2026, they followed with AI TrueCheck, which is the more interesting product because it is the less marketable one. AI TrueCheck tells legal teams when the AI is probably wrong. Three checks - machine confidence, statistical sampling, and a human-in-the-loop review - all in service of admitting that contract extraction is not yet a solved problem. Most vendors hide the doubt. Pramata charges for it.
What you can actually do with it
Find every contract that auto-renews in the next 90 days. Pull every MFN clause across a $400M book. Tell sales which customer commitments are blocking a price change. Hand an AI agent the contract context it needs before it answers a question. Catch the renewal you forgot. Do the diligence in two weeks instead of two months.
Milestones, abbreviated
The Pramata stack, by gravity
Where the work happens, by approximate weight of effort inside an enterprise rollout.
Illustrative weights from public discussion of CLM rollouts. Translation: the boring 92% is the moat.
The proof
Logos that legal departments recognize.
Pramata's customer list is the kind that does not get bought on a free trial. Comcast Business runs commercial contracts on it. NCR, Hewlett Packard Enterprise, McKesson, CenturyLink, Pitney Bowes, Vertafore, Allergan, FICO, ICE - all of them, public references at one point or another. These are companies whose contract books run to tens of thousands of executed agreements, whose audit committees ask uncomfortable questions, and whose general counsels do not enjoy surprises. They are, as a rule, the hardest customers to win and the slowest to leave.
Most CLM startups sell software. Pramata sells the clean data underneath.
- A useful way to read the companyThe mission
Contract data, treated like grown-up data.
Customer data has CRMs. Financial data has ERPs. Engineering data has observability platforms. Contract data, for most of the modern enterprise, has - a folder. Pramata's mission is to fix that asymmetry. Treat contract data like a first-class corporate asset, with the same hygiene standards, the same query layer, the same access controls. Pipe it into the AI agents and autonomous processes that increasingly run the business. Let the decision-maker see what was actually signed.
It is a mission that has aged unusually well. In 2005 it sounded niche. In 2015 it sounded promising. In 2026, when every AI agent demo is one hallucinated contract term away from a lawsuit, it sounds like infrastructure.
Why it matters tomorrow
Agents will need ground truth. Pramata has been mining it.
The next eighteen months in enterprise software will be a noisy argument about agents - what they can do, where they get their data, who is liable when they get it wrong. Most of the noise will be about the model. The quieter, slower story is about the substrate: where do agents read the company's real, signed commitments? Pramata's answer is, more or less, the only answer that has been shipping to Fortune 500 customers for two decades. Whether the market values that head start is the open question. The customers already do.
It is May 2026 and somewhere inside a Fortune 500 finance department, an AI agent is about to make a decision worth seven figures.
- And this time, the contract it needs is already structured, validated, and one API call away.The banker's box is still out there, in a storage room in New Jersey, untouched. Praful Saklani's company has spent twenty years making sure the next deal does not depend on it. That is the work. The headlines about generative AI will come and go. The unglamorous infrastructure of knowing what was signed - that gets to stay.