What would have to be true
If you sit on an investment committee — or write the memos that go to one — you know the weakness of every risks section: it was written by the person who wants the deal done. Big funds staff an adversarial read; small funds don't have the bodies, and the deal closes in days. Timepoint fills exactly that gap: the downside section, generated and argued by simulation, in time for the vote.
Timepoint AI is a Santa Monica company that runs decisions as grounded simulations: the specific, named actors around your decision, played forward through branching timelines. Findings come back ranked, with the reasoning in words and the weights labeled uncalibrated. Outputs are AI simulations, labeled as such. How the work runs →
The champion's problem
The issue isn't honesty — it's structure. A champion's downside list contains the risks that occurred to the champion. What an IC actually needs is the risk that occurred to no one, plus a disciplined answer to the only question that matters at the margin: what would have to be true for this to work — and which of those load-bearing assumptions does every rosy scenario quietly share?
So the deal runs as simulation: the market forward, the incumbent's product meeting from their side of the table, and the portal run — starting from the write-down everyone says can't happen and reasoning backward to its earliest common cause.
A seed fund (fictional) is deciding on a vertical-SaaS company selling scheduling-and-billing software into dental clinics, at a price that assumes a land-grab. One live question: does this survive the incumbent practice-management suite noticing?
| Rank | Branch | What the runs showed | Stated uncertainty |
|---|---|---|---|
| 1 | Incumbent bundles a clone, free | The modal branch. The incumbent's simulated roadmap meeting reaches for the bundle reflexively — but the clone ships late and undercooked in most runs, and the startup survives if it has moved upmarket to multi-location groups by then. | Stable on the response, divergent on the timing — 12 to 30 months across runs. |
| 2 | Incumbent ignores; land-grab holds | The memo's base case. Real, but rarer than the memo priced — it required the incumbent's (simulated) org to stay distracted by its own migration, which several runs didn't grant. | The optimistic branch is the fragile one; the runs say the price assumes it. |
| 3 | Channel partner acquires early | The branch nobody had discussed: the distribution partner the startup depends on is also its most motivated acquirer — capping the upside the valuation was paying for. | Low weight across runs, high consequence; belongs in the memo as a term-sheet consideration, not a footnote. |
Backward from every good outcome: the founder's head of sales — the one person who has sold into dental groups before — is still there in month eighteen. Forward-looking diligence had a retention line item for the founder. The simulation priced the wrong person's departure.
What lands in your memo
A one-page appendix: branches ranked with the reasoning shown, the incumbent's most likely move and its timing spread, the load-bearing assumption to verify before wiring — and the sentence an LP most wants to see in an emerging manager's memo: here is what would have to be true, and here is how we'll know early if it isn't. Engagements are scoped and priced in one call; if the question is sensitive, it can run quiet engagement — on public information only.
See it in practice
- Six sample deal-team simulations — term-sheet negotiations, takeover defenses, LBO rate scenarios — example runs of the same engine
- A full engagement, worked end to end — a fictional venture firm allocating one partnership slot across three portfolio companies
- Why the engine refuses to bluff — the realism checks and the enforcement gate behind every deliverable
Field Notes on Deal Risk — an 8-page guide to stress-testing a decision across its branches: the five practices, a worked example, and the checklist. No gate, no sales call.