Case study · FounderPact · 2026 · Beta
People start building together before they have talked about roles, ownership, money, decision rights, or what happens when priorities change. FounderPact puts Solomon, a conversational AI mediator, between them: it interviews each founder separately, structures what they care about, and shows them where they agree, where the gaps are negotiable, and where they are not.
I have been a co-founder more than once. The conversations that matter most, who owns what, who decides, what happens to the equity if someone leaves, are the ones everyone postpones, because they are uncomfortable and because a static agreement form arrives too late to help. By the time a template asks the question, the founders already have positions they did not know they held.
That is a product hypothesis drawn from lived experience, not a measured market. The bet is that a conversation, held separately with each person and then reconciled, surfaces those positions early enough to be useful.
Solomon, named for history's fairest judge, interviews each founder on video or in chat, helps them structure their own position, and then produces two things: a private Position Report for each person and one shared Alignment Report for the pair. The report separates three kinds of items: agreements, negotiable gaps, and unresolved conflicts.
flowchart LR A["Founder A
interview with Solomon"]:::mine --> PA["Position Report A
(private)"]:::mid B["Founder B
interview with Solomon"]:::mine --> PB["Position Report B
(private)"]:::mid PA --> AL["Alignment Report
agreements · gaps · conflicts"]:::good PB --> AL AL --> PACT["Proposed pact
founders accept, revise, or leave open"]:::good classDef mine fill:#dcfce7,stroke:#16a34a,color:#14532d; classDef mid fill:#eef2ff,stroke:#6366f1,color:#1e1b4b; classDef good fill:#dbeafe,stroke:#2563eb,color:#0b3a8f;
The first shape was short interviews followed by a shared questionnaire and the two reports. I tested the questionnaire on my own collaboration and watched another founder go through it. The feedback was specific and humbling: personality questions repeated themselves, the equity labels confused people, and several sections were too dense to finish in one sitting. Those went back into the design.
By October the Solomon experience was mobile first, with chat alongside voice and an in-context way to ask Solomon about any single question. A later candidate that prefills the questionnaire from the interview was built and tested but, at the time of writing, had not been merged, so I do not count it as shipped.
Two complete, fictional two-founder journeys were run end to end. One reached a jointly signed pact. The other exhausted its revisions and preserved the disagreement rather than papering over it, which is the behavior I wanted. Reload and account-switch checks kept the outcomes intact, and private objection phrases did not appear in the partner's view or the shared view.
The honest boundary. Those are workflow and privacy checks. They do not establish adoption, real-speech accuracy, or better founder outcomes. One founder has said he would use it with a collaborator, and a handful of people were waiting to test. That is a beta, and I describe it as one.
FounderPact is the project where I learned, hands on, what a conversational agent needs in order to be useful across a sensitive, multi-step, two-person workflow: orchestration of the model, structured data behind the conversation, interaction design for chat and voice, privacy boundaries enforced in code, state that survives days and devices, and a clear line where the human decides. None of that is specific to founders. It is the shape of most serious conversational AI products.