A single, mandatory front door for every AI idea. The intake form filters bad ideas before any build effort.
Minimum intake fields:
- Workflow being targeted, and current pain point
- Department / business unit / champion
- Expected business value (cycle time saved, error reduction, revenue impact, cost saved)
- Data the agent will see (PII? regulated? confidential?)
- Decisions the agent will make (advisory? automated? consequential about people?)
- Systems the agent will write to
- Estimated volume per month
- Proposed autonomy level (Assistive / Validated / Autonomous)
- Known constraints (regulatory, contractual, customer-facing)
This produces the risk classification almost for free, and feeds directly into the Agent Card if approved.