How to run a 90-day AI visibility pilot with a go or no-go decision
Run the pilot to prove the system works, not to prove pipeline. In 90 days you can set a weekly AI visibility baseline in every priority market, run the first wave of agents with reviewers in the loop, and measure the hours saved against your estimate. Write the go, adjust and stop criteria before day 0, review at day 45, and decide at day 90. Pipeline is counted from the following quarter.
Key takeaways
- Ninety days is enough to prove the measurement, the workflow and the hours; it is too short to judge pipeline.
- Spend two weeks on readiness: reviewers named, read-only access granted, baselines set.
- Start with the agents that save the most hours and carry the least risk; add the rest in waves.
- Write go, adjust and stop criteria before the pilot starts so the decision is about facts.
What a pilot should prove
- The measurement works: a fixed set of buyer questions asked weekly in each assistant and market, with results you trust.
- The workflow works: agents prepare, a second check catches errors, the owning function signs off, and nothing is published without a person.
- The hours are real: time tracked for four weeks against the estimate in your business case.
If those three hold, scaling is a resourcing decision. If they do not, more spend will not fix it.
The 90-day plan
| When | Milestone | Exit criteria |
|---|---|---|
| Day -14 to 0 | Readiness | Reviewers named by function, read-only access to search console, analytics, tag manager and CRM reports, baselines set |
| Day 0 to 30 | Start and baseline | Weekly cycle running; first AI visibility read-out; first backlog of gaps |
| Day 45 | Mid-pilot review | Early signals, hours saved so far, review load per function |
| Day 46 to 90 | Run and learn | Gaps closed through briefs and site fixes; time tracking complete |
| Day 90 | Go or no-go | Results against the written criteria; decision to scale, adjust or stop |
Get an AI visibility read across ChatGPT, Perplexity, Gemini, Claude and Google AI Overviews.
Which agents go in the first wave
Start with work that is repeatable, measurable and internal: the AI answer checks, search and analytics reporting, keyword and question research, brief drafting and pre-publish quality checks. Leave anything that touches customers, the CRM or the live site for later waves. Our human-in-the-loop approval workflow explains how sign-off works by function.
Go, adjust or stop: write it down first
- Go: a trusted weekly baseline exists for every priority market, measured hours saved reach at least half of the estimate, and review time stayed within what each function agreed.
- Adjust: the baseline exists but review load ran over; keep the agents that saved the most time and cut the rest.
- Stop: no reliable baseline, or review time cancels the hours saved.
Do not judge the pilot on traffic or pipeline alone. Count pipeline from the quarter after, as a conservative, base and ambitious range. Measuring share of answer covers the baseline method in detail.
Get it approved
The pilot is the ask at the end of a wider program brief. Our AI visibility program brief template includes the pilot table, the criteria and the decisions to request. Our generative engine optimization service can run the pilot with your team. Talk to our team.
Frequently asked questions
How long should an AI visibility pilot run?
Ninety days, with two weeks of readiness before it. That is long enough to set baselines and measure hours saved, and short enough to keep leadership attention.
What should we measure in the pilot?
A weekly share of answer baseline by market, the accuracy of AI descriptions of your company, hours saved against the estimate and review time per function.
Should we expect pipeline from the pilot?
Not as the decision criterion. Visibility gains take longer to show in pipeline; count it from the quarter after the pilot.
Related service: Generative engine optimization. See also pricing and client results.
We run enterprise B2B demand generation for teams in Montreal, Toronto, Ottawa and the United States.
