The Agent Readiness Checklist: 25 things to check before an AI agent touches your Salesforce data
The demo ran on 40 tidy accounts. Yours has three records for the same customer and an integration user that can delete anything. The checklist is how you find that out on a Tuesday, not in production.
An Agentforce readiness checklist in five groups of five
It’s a PDF of 25 yes/no checks. One sentence under each says why it matters. You can answer most of them in an afternoon, with a person who knows the org.
What’s inside the checklist
- Data. One record per customer, and the other basics an agent takes at face value.
- Meaning. Fields that say what they hold, tidy picklists, descriptions filled in.
- Access. What the agent’s user can see and change, least privilege, no shared admin logins.
- Actions and guardrails. Dry runs, approvals for risky actions, record ceilings, a stop condition.
- Proof. Field history, a log of what the agent did, an owner and a review date.
Scoring takes three lines: 20 or more yes answers means ready to pilot, 12 to 19 means fix the gaps first, under 12 means start with the data.
Who it’s for
The person told to “get us ready for AI”, and the person who signs off when an agent starts acting on customer records. For a full assessment, there’s the Agentforce Readiness Sprint.
Questions people ask
How do I know if my Salesforce org is ready for an AI agent?
Check five things: the data, what the fields mean, what the agent’s user can access, what limits are on its actions, and what proof you’ll have afterward. The checklist turns those into 25 yes/no questions.
Is the checklist only for Agentforce?
No. Agentforce is the example. The checks are about your data, access and guardrails, so they fit any agent in your org.
What if we score low?
A low score is information, not a verdict. Start with the data: the scorecard is the 20-minute version, and the Data Health Check is the thorough one.