Telogence evolves how the firm reasons — from the first screen, to the final IC, through the operating KPIs, to exit. As AI makes answers free and instant, the edge moves to the questions you ask. Better decisions drive higher returns.
How a deal moves from first look to portfolio company, and how outcomes feed back to sharpen the way the firm reasons.
Fast triage, before opening the CIM. Three pages in, the team calls it — pass or pursue.
three skill files, one page each
What the company does, who the customers are, the unit economics.
Project Atlas example ↗Twenty KPIs, heat-mapped — outliers surface at a glance.
Project Atlas example ↗Industry-specific commentary — what only matters in this vertical.
Project Atlas example ↗Everyone reads the CIM — then the full team talks. Five skills orient the read.
five orientation skill files
Who runs this company — and what's their track record?
Atlas example ↗The lay of the land — web-based deep research.
Atlas example ↗Where has the firm seen a business like this in its own portfolio?
Atlas example ↗Where has the firm seen a deal like this before — and what happened?
Atlas example ↗The revenue cube, cut by customer, service line, and vertical.
Atlas example ↗Defined by the full team at the end of Baseline. Every turn of the loop re-enters here — questions sharpened, split, retired, added. The Question Tree for Project Atlas:
Nothing floats free: every answer is attached to a question and backed by evidence — and everything lands in the Reasoning Graph, always exposed via chat.
A question (Q) pulses → the lake is swept → evidence (E) links upward → an answer (A) forms as a falsifiable claim. Where none can be reached, the question splits — dig deeper. The joined structure is the Reasoning Graph.
The QB plans, juniors and subagents execute, the QB approves — the team reviews the day's changes.
Open ↗The Decision Brief to read; the Report as reference; the Graph and Lake as chat.
Open ↗The approach to the key questions.
Human and AI, side by side.
Every answer — including all subagent results.
The Reasoning Report, daily.
Optionally.
New answers, retired claims — reviewed by every team member, every day.
Today's Atlas changes ↗The Report is a rendering; the record is the Reasoning Graph — chat-queryable at all times.
The deal team queries the graph and the lake continuously; the IC and MC receive them in weekly cycles.
The one thing the IC is expected to read.
Atlas example ↗No expectation to read it — there when a thread needs pulling.
Atlas example ↗Every answer linked up to its question, down to its evidence.
Atlas example ↗The raw conversational context — no investment-team reasoning applied.
Atlas example ↗The weekly discussion updates the Question Tree and the Reasoning Graph — only then is the updated graph sent to the Model Counsel.
A score 1–10, a short opinion, a max willingness to pay — from every IC member, deal team member, and the Model Counsel.
Post-close, the KPIs are defined from the Reasoning Graph — each a live test of a claim the case rests on, flagged automatically on deviation. Project Atlas, four quarters in:
Performance vs. plan drives the call — continuously, the whole hold. The graph says exactly what depended on each assumption.
Whatever happens next — good or bad — the claim that motivated it gets the result. The cycle scores the reasoning itself.
Hiring on a six-month ramp? The record shows how that assumption performed every time the firm underwrote it — and can suggest strategies.
The deal ends where the reasoning is finally judged: the realized return against the case the firm underwrote — claim by claim. From the record: .
Every deal — invested, passed, lost, or exited — leaves its full Reasoning Graph in the record, and every stage draws on it. Deal after deal, it compounds into the firm's Reasoning Traces: how it reasons, with outcomes attached.
/portfolio-similarity and /deal-history-similarity search past graphs directly —
If in ten deals we underwrite X and X never happens, the system flags the firm to stop underwriting X. Tracked votes develop human judgment — and are how the MC may earn one.
You've seen the system — first screen to exit, and the record compounding at its center. The next step is seeing it on your deal.