What does a first pilot look like for a client and matter opening agent?
A professional services team scopes a pilot around opening client matters after engagement approval, with conflict checks, trust account rules and a held-out set, then reviews results by failure mode.
A professional services group is building an agent to open client matters once an engagement has been approved by a partner. The engineering lead and the risk manager meet MirrorFirm to scope a pilot.
They choose one path: engagement approval ENG-1180 becomes a client record, a matter, a billing arrangement and a conflict-check task. They agree on four record types and six tool actions. The case mix includes:
- new clients with clean checks - existing clients opening a second matter - a related-party name that should trigger conflict review - an engagement missing the fee basis - a matter type that policy says requires partner sign-off before billing is enabled
They request a small training split, a validation split and a held-out set of cases they will not use to adjust prompts. The answer key lists expected records and handovers for each case. Two reviewers, a senior practice manager and a security analyst, are named to inspect samples for realism and disclosure risk.
After the first release, the team runs its agent through a harness and reads results by failure mode. Matching works, but the agent invents a fee basis when one is missing. They change the case mix to include more incomplete engagements and request a regenerated release.
**What it teaches:**
- Narrow scope produces readable results. - A held-out set protects the evaluation. - Reviewers from engineering, risk and the business each catch different problems.
A pilot built around one workflow and a held-out set gives engineering and risk teams concrete results to review together.
- Audience:
- Agent developers
- Content type:
- Scenario
- Buyer question:
- Scope and getting started
Content release: r-a220eb6e099b243b
Related questions
- What licence and usage limits should we ask about before starting?
- How should we scope a first pilot for set-up agents?
- What should we ask for in a sample or pilot release for set-up?
- How do we wire a pilot release into our agent harness on day one?
- How should delivery formats, versioning and data splits be agreed for a pilot?
- What should I know about licensing, limits and getting started?