ParetoBridge — Flow Blueprint
A founder and an investor each confide in a neutral AI "Fiduciary." It hears both sides privately, drafts one revised term sheet meant to help both, and the revision only happens if both independently accept. Declining costs nothing: the original deal stands.
Founder-private
Investor-private
Shared / visible to both
AI model call — click any tag for the exact prompt
Zone 1 · Founder secretsInterview transcript, summary bullets, founder-only rationale. Never shown to the investor.
Zone 2 · SharedThe uploaded term sheet, the proposal itself, the "good for both" argument, the final result.
Zone 3 · Investor secretsInterview transcript, summary bullets, investor-only rationale. Never shown to the founder.
Step 1 · Setup
Create the workspace, upload the deal
- One party picks their role (founder or investor), fills in company name, round type, and jurisdiction, and uploads the draft term sheet (.docx, .pdf, or .txt — text is extracted for the AI; the original file is kept byte-for-byte for later download).
- The system creates a shared workspace with two role-bound links; the counterparty is invited by email and joins through theirs.
- The uploaded document is both the shared starting point and the fallback: it is exactly what survives if the process produces nothing both sides like.
Private · founder only
Briefing
Before anything is asked, the founder sees the ground rules: the Fiduciary is neutral, everything shared is confidential, honesty gets better outcomes, and you cannot lose — disliking the proposal just means keeping the existing deal. An example proposal can be previewed.
Private · investor only
Briefing
The investor sees the same ground rules, framed for their side. The point of the briefing is to earn candor: the whole system only works if people say things here they wouldn't say across the table.
Private · founder only
Interview
A chat with the Fiduciary. One question per turn, drawn from a founder topic menu:
- Exit horizon and liquidity timing needs
- Confidence in near-term milestones; downside scenarios
- Control vs. economics tradeoffs
- Trust and governance comfort with this investor
- Planned pivots needing term flexibility
- Personal financial pressure; acceptable structured liquidity
AI call 1 · next interview question
Private · investor only
Interview
Mirror image, with an investor topic menu:
- Fund years remaining; target exit horizon
- Follow-on intentions across future rounds
- Conviction level and acceptable risk
- Strategic-acquirer relationships affecting exit optionality
- Portfolio conflicts and governance safeguards
- LP distribution pressure and liquidity timing
AI call 1 · next interview question
↓
Interview mechanics (both lanes)
How turns are counted and kept honest
- The interview needs 5–8 substantive answers; a progress bar tracks them. The AI sees the term sheet text and full transcript each turn, so questions build on prior answers and never repeat.
- Clarifications are always free. "What does that mean?", "can you give an example?", or asking the Fiduciary for its opinion never advances the counter — the Fiduciary explains, then restates the pending question.
- Low-quality messages get one grace pass. A typo, keyboard mash, or off-topic question ("what color is the sun?") is ignored once; a second one in a row counts as a used turn. A real answer resets the grace.
- Guardrails on the AI itself: it must stay on financing and governance topics (no product roadmap, hiring, or go-to-market chat), and its message must actually contain a question — otherwise a scripted backup question is used instead.
Private · founder only
Approve "what it learned"
The AI condenses the interview into 4–8 plain bullets addressed to the founder ("You would walk away from full-ratchet…"), keeping numbers and qualifications, adding no advice. The founder can add, edit, or delete any bullet.
AI call 2 · summarize interview
Private · investor only
Approve "what it learned"
Same on this side. Crucially, these edited bullets — not the raw transcript — are the only thing that flows forward into proposal drafting. Each side has final say over what the Fiduciary may use.
AI call 2 · summarize interview
Step 2 · Sync point
Wait for the partner (with an escape hatch)
- Whoever finishes first waits on a status screen with a simulated 24-hour clock counting toward a timeout.
- They can trigger a reminder email to the partner.
- If the partner never finishes, the waiting party may proceed solo, choosing between: use only my information (the proposal reflects one side and is flagged as such — more likely to be rejected), or simulate the counterparty (demo/testing only: a canned, clearly-labeled set of "typical founder/investor" priorities stands in for the missing side).
↓
Step 3 · The bridge
The Fiduciary drafts one proposal
This is the only place the two private zones ever meet — inside a single AI call that reads the term sheet plus both bullet lists and returns one package:
overall_argument — why this helps both sides (shared)
package_summary — the package in one breath (shared)
rationale_to_founder — founder-only; must not leak investor secrets
rationale_to_investor — investor-only; must not leak founder secrets
proposed_changes[] — per change: section · current text · proposed text ·
why it's an improvement · benefit to each side
improved_term_sheet — complete clean rewrite (shared)
redlined_term_sheet — same, with additions/deletions marked (shared)
The instructions favor a few specific, internally consistent changes over a wish list, forbid inventing facts, and record whether it ran in two-party or one-party mode. A second precision pass then locates each change in the exact paragraph of the original uploaded document and marks the insertions and deletions there, so the on-screen redline and the downloadable Word file show identical changes in identical places. Generated once; both sides see the same package.
AI call 3 · draft the proposal package
AI call 4 · place redlines in the document
Private · founder only
Review & decide
The founder sees the shared argument, the full redlined proposal, and — behind a "confidential explanation" expander — the founder-only rationale. Then a binary choice: Accept or Decline the entire package. No line-item voting, no counter-offers, and no visibility into the investor's vote.
Private · investor only
Review & decide
Identical experience with the investor-only rationale. Both decisions are made blind — the double-blind vote removes any incentive to game the other side's answer. Whoever decides first waits on a second sync screen.
Step 4 · The gate
All-or-nothing unanimity
Votes are revealed only after both are in. The results screen lists each proposed change as included or not included, with the reason.
Both acceptEvery change in the package is adopted. The improved term sheet becomes the candidate deal.
Anyone declinesEvery change is dropped ("you declined" / "partner declined"). The original deal stands untouched. Nobody revealed anything, nobody lost anything.
↓
Step 5 · Final documents
Side-by-side term sheets, downloadable
- The final screen shows the original and the improved term sheet side by side, with additions in green and deletions struck through in red.
- Download original: the verbatim uploaded file — original formatting, tables, and styling intact.
- Download improved: a Word document built by applying the redlines directly to the uploaded original, so it reads like a lawyer's tracked-changes markup of the actual deal.