Sample AI agent · Win/loss analysis
Ask why you're winning and losing. Watch it investigate.
Sellers write "price" because it's the easiest reason to give. Pick a question, watch the agent dig through the CRM and buyer interviews, then decide which findings go to leadership.
1 · Ask the agent
What the agent can see
| Deal | Segment | Region | Outcome | Competitor | CRM loss reason | Days | Compliance champion |
|---|---|---|---|---|---|---|---|
| D-1001 | Mid-size bank | UK | lost | ChatGPT Enterprise | Price | 181 | no |
| D-1002 | Insurer | Nordics | lost | Gemini | Price | 142 | no |
| D-1011 | Mid-size bank | UK | won | Microsoft Copilot | 90 | yes |
“Price is just the easiest thing to write in an email when a deal stops.” Head of Operations Risk, mid-size bank, lost
2 · Investigating
Each question replays a real run on made-up data. Claude chose every step, and every quote and number passed the agent's checks.
Why it matters
Turning what buyers actually say into decisions runs through all my work. For example, at an enterprise AI company, every readiness assessment taught us what banks really needed, and we fed it straight back into positioning, packaging and pricing. Win/loss is the same habit, applied to every deal.
How it works
The agent gets a question and a set of tools, and decides for itself what to look at next and when it has enough. The code checks every finding: quotes must appear word for word in what a buyer said, and numbers must come from a query it actually ran. Anything that fails is sent back to be fixed. A person approves each finding before it's shared.
query_dealsWin rates, cycle times and loss reasons, by any CRM fieldsearch_interviewsFind what buyers said about a topicread_interviewRead a transcript in fullrecord_findingSubmit a finding, checked before it's acceptedflag_gapRecord what it couldn't answerfinishWrite the summary and stop