Outcom.AIAsk anything about an account.
Get an answer you can trust.
Plain-language questions, grounded answers. Copilot is an analyst for your entire book — it reads the same evidence-backed context your team does and answers traced to real fields and real quotes, with its confidence shown. Never a hallucinated number. Copilot is how you interrogate the Success Agent™.
Grounded answers, in the shape the question needs.
Answers from the agent's freshly-computed finding — states, metrics, diagnoses.
Retrieves the raw evidence — quotes and events — via RAG over 16 entity types.
Analyst for the number. Research for the quote.
Some questions want a computed answer; others want the source. Copilot runs both on one uniform question framework — 23 chat agents, ~230 questions — and picks the right mode for what you asked.
- Analyst — grounded on a live finding, computed on demand.
- Research — quote-backed retrieval over calls, emails and docs.
- "What are they saying about a competitor?" → the actual snippets.
Every claim binds to a real field.
Copilot answers only on the fields projected for the question — it can't reference a number it wasn't handed. A faithfulness evaluation machine-checks every claim against its source, so groundedness is measured, not hoped for. Hallucinated figures are structurally impossible.
- Answers projected strictly from required fields — no free-hand numbers.
- A faithfulness eval verifies each claim against its evidence.
- Confidence shown on every answer — the same one from the finding.
Eleven archetypes, one right route.
A question isn't just words — it has a shape. Copilot classifies each into one of 11 archetypes — state, metric, diagnosis, trajectory, forecast, comparison, recommendation and more — and routes it to the right template and mode.
- Each archetype picks its template and analyst-vs-research mode.
- The NLU pipeline turns intent into the exact node that can answer it.
- Consistent behavior across every agent in the platform.
An analyst on every account.
Account research that once spanned six tabs returns as one grounded answer, in seconds, on any account in the book. The point was never a chatbot — it was an answer you can stake a renewal on.
- Normalize → intent → plan → node → synthesize — the core query path.
- Reads the same Context360 truth — no separate store to drift.
- Turn any answer into a document through Compose, grounded end to end.