‹ ALL PROJECTS

Win-Loss Agent.

An agent that interrogates the deal record, from win-loss calls to CRM notes to email threads, and explains why deals close or die. It turns post-mortems from a quarterly ritual into a living playbook that updates with every closed deal.

LOG ENTRY002 / 2025-09
STATUS● In progress
TERRAINllmagentsrevops
MAPTHE SYSTEM AT A GLANCE
DEAL RECORDwin-loss callsCRM notes · emailAGENTinterrogates evidence,extracts decision driversPATTERNSpricing · champion· competitor playsPLAYBOOKliving — updateswith every dealEVERY CLOSED DEAL RE-ENTERS THE RECORD

The problem

Win-loss analysis is the highest-leverage, least-done job in sales. The evidence already exists in call recordings, CRM notes, and email threads, but nobody has time to read it all, so teams repeat the same losses. Consultancies sell the analysis as a quarterly project for five figures, and by the time the report lands, the quarter that produced it is over.

Who it's for

  • The CRO buys pattern visibility: which objections, competitors, and deal shapes kill revenue, in the buyer's words rather than the rep's recollection.
  • Product marketing gets a competitive playbook grounded in real deals instead of anecdotes from the loudest rep.
  • Sales managers get loss reasons they can coach against, deal by deal.
  • Reps contribute nothing extra. The agent reads what already exists, which is the only way win-loss ever gets done consistently.

The approach

An agent that treats the deal record as a witness to be interrogated:

  • Ingests win-loss calls, CRM history, and threads per deal.
  • Extracts the decision drivers: what moved the buyer, in their words.
  • Aggregates across deals into patterns like pricing objections, champion strength, and competitor plays.
  • Publishes a living playbook that updates with every closed deal.

Architecture decisions

Being made now, with the same rules as the GTM Operating System: evidence-cited extraction (no claim without a quote), structured outputs at every step, and aggregation that keeps "what one buyer said" separate from "what the pattern says."

Business impact

The target is replacing the quarterly win-loss consulting engagement with an always-current playbook at API-call prices, and shortening the loop from "loss happened" to "team adjusted" from a quarter to a week. Numbers land here once the build does.

Risks & guardrails

The known failure mode of automated win-loss is confident storytelling, a plausible narrative stitched from thin evidence. The design answer is inherited from the GTM work: every driver cites its source passage, and low-evidence deals get flagged as low-evidence rather than narrated anyway.

Where it goes next

In active development. This entry updates as the ascent continues.