I was made redundant from a PM job in offshore wind. Built an AI contract tool instead. 25 users in, here's what I've learned.
Good catch — fixed throughout. Here's the full post:
I spent years as an IP and commercial lawyer before I ever touched a line of code. Trademark licensing at Dr. Reddy's Laboratories, SaaS contracts at Datum Cybertech - reading other people's contracts for a living. Then I moved into product and growth, leading B2B SaaS launches at Vekta Group in offshore wind. That's where I ended up on the other side of the table: reviewing vendor and partnership contracts as a non-lawyer, feeling the exact gap I used to fill for other people.
Earlier this year I was made redundant from that role. I had a choice. Look for the next PM job, or find out if I could build the thing I'd spent years wishing existed. I picked the second one. No technical co-founder, no dev background - I'm the legal engineer, and I've been learning to build alongside that.
The problem I'm building for: most founders sign contracts with nobody checking them, because a proper legal review costs more than the deal's worth and takes longer than the deal can wait. The real choice most people face isn't "lawyer or AI." It's "AI, or nothing." That's a much lower bar than most legal tech pitches assume, and it's the one Pactora is actually built against.
What it does: upload a contract, get a plain-English risk verdict across 12 clause areas - IP ownership, liability, indemnity, data protection, termination, and more - in about 30 seconds. But the part I care about most isn't the flagging. Every risk comes with a negotiation ladder: your opening ask, a fallback, and a narrowing position, each with a usable script. Most tools stop at "this looks risky." I wanted it to stop at "here's what to say back," because that's the part a lawyer actually gives you.
Where it's at: live in beta, 25 users, pre-revenue. Pricing is mapped out (£49 to £249/month across tiers) but payment isn't set up yet - the focus right now is making sure the core detection holds up under real use before I open up paid plans. Ran a head-to-head test against ChatGPT and Claude on a real £54k SaaS contract recently - Pactora caught a cap-indemnity mismatch worth up to £400k that both general-purpose tools missed, because they were reading clauses in isolation instead of checking how they interact. That test told me more about whether this is actually differentiated than any amount of my own conviction did.
What's been hard: trust, mostly. This reads people's actual contracts, so "your documents are never stored or used for training" isn't a line in a privacy policy nobody reads - it's on the homepage, and it needs to be, because legal tools have a much higher trust bar than most SaaS. Getting people to try it on a real contract rather than a throwaway one has been the actual unlock.
Happy to answer anything - what convinced me to keep going after the redundancy, how the clause detection actually works, what building this as a non-engineer has looked like day to day, or just poke holes in the idea. That's what I'm here for.
