Orbecta v3.5 — The day a contract stopped being a document and became a workbench
Today I rebuilt the place where you actually work on a contract.
Until now, Contracts gave you a Briefing. Seven lenses, an executive summary, recommendations you could use at the table. Good output. But it was something you read, not something you worked in. You got the analysis, closed the tab, and the work happened somewhere else — in your head, in a separate to-do list, in a calendar nobody connected back to the contract.
That gap bothered me.
So today I built the Clause Workbench.
What it is
Open a contract and you get two things side by side. On the left, the full contract text with a heatmap that marks every clause by risk. On the right, every detected clause as a card — with its negotiation point, a note field, a calendar button, and a status.
The text is read locally in your browser and stored nowhere. It lives in your session and disappears when you close it. A contract is the most sensitive thing a procurement team owns; the analysis can be saved, the raw text doesn't need to be.
You can select any passage and ask Orbecta to analyze just that clause — an eighth lens that runs on demand, on the text you pick yourself. It pulls in benchmark data and similar clauses from real contracts, and tells you what the clause means commercially, how it compares to the market, and what to put on the table. Not legal advice. Procurement judgment.
And every clause is now something you act on, not just something you read. Plan it on a date and it lands in the same agenda as the rest of your day. Write a note and it stays with that clause — and travels with it into the agenda. Mark it open, in progress, or done. Done clauses fold away into a single green line so your overview stays clean.
The contract finally became a place where the work happens, not a place you leave to go do the work.
Where the day actually went
I expected to spend it on features. I spent most of it on judgment, again.
The Workbench felt slow. The easy assumption was that reading a 20-page PDF was the bottleneck. So before optimizing anything, I measured it. Extraction took 0.6 seconds. It was never the problem. The problem was that the screen rebuilt itself completely on every click, every scroll, every text selection. Seventeen full redraws where a handful would do.
If I hadn't measured, I would have spent the day making the fast part faster and the slow part would still be slow.
Then a smaller thing, the kind that decides whether people trust a tool. You'd save a note and the field stayed open, as if nothing happened. Technically it was saved. But it didn't feel saved. So I gave it a real saved state — the note as text, a pencil to edit. The data didn't change. The trust did.
And one bug that taught me something. Selecting text and asking for an analysis failed with a flat error. The cause: I had two settings switched on that the model doesn't allow together. It had never worked with that combination — it only looked finished. A reminder that "deployed" and "working" are not the same word.
The thing I keep relearning
I asked the same question today I asked last week.
Does this help the user do the work, or am I building it because it's interesting to build?
The agenda button, the notes, the status — those passed. They turn analysis into action. That's the whole point of Orbecta: closing the gap between knowing and doing. This is that gap, inside a single contract.
But I also caught myself wanting to sync the note back from the agenda into the contract, two-way, perfectly mirrored. Elegant. And useless. A note travels one way: from the work into the plan. Nobody edits a contract clause from inside their calendar. I built the one direction that's real and left the clever one alone.
A month ago I would have built both.
What I learned as a founder today
Speed problems are usually somewhere other than where they feel.
Measure before you fix. "Saved" has to feel saved, not just be saved. And "deployed" is not "working."
The technology decides what's possible.
Judgment decides what's worth shipping.
Ken, founder of Orbecta
procurement AI
contract intelligence
Orbecta