Quality
Did the answers cite the right sources?
Retrieval scores on your own corpus, sampled from real pilot queries. Comparison against an eval set baseline we build with you in week 1.
Pilot program
A pilot is the right format for "try Archivez on your own corpus." Light commitment, scoped to one tenant, finished with a written usage report and a decision on production rollout.
Timeline
Week 1
Tenant provisioned on staging or prod. Admin users invited via email link. Initial bulk-upload of your sample document set.
Week 2
Documents finish indexing (same day for under 1,000 files). Pilot users connect Gmail. Monitored senders configured per user.
Weeks 3–6
Pilot users ask questions, flag bad answers. Feedback loop on retrieval quality, citation accuracy, mobile UX. Weekly 30-min check-in.
Weeks 7–8
Usage report — query counts, top documents cited, eval-set scores against your corpus. Pricing for production rollout.
Scope
Decision criteria
At the end of the pilot we share a usage report. It answers four questions your team needs answered before paying for production:
Quality
Retrieval scores on your own corpus, sampled from real pilot queries. Comparison against an eval set baseline we build with you in week 1.
Usage
Per-user query counts, top documents cited, retention week-over-week. If usage drops off in week 4, the pilot answers why before week 8 hits.
Cost
Projected monthly AI spend at your real query rate, with a recommended per-tenant cost cap calibrated to your usage pattern.
A 30-minute intro call is enough to decide whether a pilot makes sense. Bring an example of the kind of question you wish your documents could answer.