Start with one team. Expand when it works.
Most enterprise AI programs that try to roll out everywhere at once tend to stall. The ones that work start narrow. A Cogniplane pilot is the smallest useful version of a governed AI workspace: one team, a handful of approved skills, an approval policy you can defend. By the end of week one you have real usage, real approval decisions, and a session history your security team can actually open. From there you either expand or you don't, and either way you keep what you learned.
What a pilot includes
A pilot is scoped tightly so it can finish. We pick one team, two or three real workflows, and the integrations they actually need. Anything the team doesn't need on day one stays off.
One team, one workflow set
Pick a team with a clear use case. Finance reporting, sales research, HR Q&A, recurring legal review. The "enable AI for everyone" pilots don't usually produce decisions, so we don't do them.
The integrations they actually use
Notion, GitHub, Microsoft / SharePoint out of the box; one or two custom MCP servers if their workflow needs them. Reads and writes are toggled separately so the team can start safely.
Approval rules tuned to their risk
Read-only actions can auto-approve. Anything that writes, sends, or exports waits for a human. We set those rules with you on day one, so they're not defaults you find out about later.
A reviewable audit trail from minute one
Every session, tool call, approval decision, and PII finding is logged. Your security or compliance lead can replay any session. This is the artifact most pilots forget to produce.
What the first month looks like
Most pilots run a little slower than what's below. The thing that usually slips is the customer team's availability, not ours, so a week-long step often takes ten days in practice. Treat the timeline as the optimistic end of the range.
Setup call and tenant provisioning
One 60-minute call. We agree on the team, the workflows, the integrations, and the approval policy. Your tenant comes online within a few business days, depending on what your IT team needs to verify on the SSO side. We send the security posture page ahead of the call so your security lead has time to read it before we get into the details.
Real usage, real approval decisions
The pilot team starts working in Cogniplane. We sit in the loop for the first few sessions. By Friday, you have:
- A working tenant with the team's approved skills and integrations enabled
- Real session transcripts, including approvals that fired
- Token usage and cost data per user
- A first read on what the team actually uses AI for, which is rarely what they predicted
Tune, expand, or stop
You decide. Common moves: tighten an approval rule that's noisier than expected, add a second integration, draft a custom skill for one of the team's workflows. The tenant policy updates live; no redeploys, no migration windows.
Pilot review
One 60-minute review call. We bring usage data, the audit trail, and a recommendation. You bring the team's feedback. From there we agree on what expansion looks like, whether that's a second team, a broader skill catalog, or a long-term contract. Sometimes the answer is that the fit isn't right; if so, you keep the data and we wrap things up cleanly.
What a pilot costs
Pilots are a paid engagement, not a free trial. The conversation tends to go better when both sides have something at stake, and the kind of pilot we run takes real configuration time on our end. Self-serve signup wouldn't get you the same outcome.
Pricing depends on the team size, which integrations are involved, and how much custom skill work the pilot needs. Before week 0 we send a one-page scoping doc with a fixed fee on it. No per-token surprise billing, no open-ended consulting hours. If the pilot rolls into a long-term contract, the pilot fee credits against the first quarter.
For most teams the pilot lands somewhere in the range of a single seat-license year on a comparable enterprise tool. We'll quote precisely once we've talked through the scope. If the budget framing isn't going to work, tell us early. We'd rather hear that up front than negotiate it down at the end.
What we need from your team
A pilot only works if a few things are in place on your side. None of them are heavy. They just have to exist before week one starts, otherwise the pilot drifts.
An executive sponsor
Someone senior who has decided they want this to work. Without a sponsor, a pilot tends to fade out the second a competing priority lands on someone's calendar. With one, the team finishes.
An identified pilot team
One team with a clear, recurring set of workflows. Ideally a team that's already using unsanctioned AI tools, because they have the most to gain and the most concrete feedback.
SSO available
WorkOS supports SAML, OIDC, Google Workspace, and Microsoft Entra. We need credentials or a connector contact from your IT team.
A security review window
Most security teams need 1–3 weeks. We can run that review in parallel with week 0 setup so it doesn't push the pilot. Start with the security posture page; it covers what most reviews ask first.
Ready to scope a pilot?
Email us with the team you have in mind. We'll send a one-page scoping doc and book the setup call. If a pilot isn't the right fit yet, we'll tell you. They aren't free and they aren't for every situation, and we'd rather lose a deal early than waste both sides' quarter.