1. Understand the existing workflow
We begin by learning how a team currently collects SEO findings, reviews them, prioritizes them, and assigns follow-up. The goal is to identify a repeated handoff worth testing—not to assume a problem from an agency’s website.
2. Define the boundary
Agree which public site may be checked, the page limit, what signals are in scope, and what data must not be accessed. The current prototype uses anonymous public-page requests and does not require a CMS login.
3. Review observations together
Each finding remains a proposal until a qualified person checks the page and decides whether it is accurate and useful. A technical signal does not automatically become a client recommendation.
4. Measure a real baseline
Before a pilot, agree on a representative workflow, a baseline measure, and a numeric success threshold. Potential measures include reviewer time, accepted findings, false positives, or time to assignment. Results are reported only after they are observed.
5. Close out and decide
Document the sample, limitations, findings accepted or rejected, and whether the workflow merits further work. Any future integration or write access requires separate scope, permission, and approval controls.
Current prototype boundaries
The prototype is early, deterministic, read-only, and manually run. It does not monitor continuously, connect to a CMS, change content, or provide a ranking forecast.