A readiness assessment report should fit on two pages and answer three questions: how well does the team understand governed agents, concept by concept; which gaps would cause harm at the depth you plan to run; and which agent should come first. Lead with the recommendation, then show the scores and gaps that justify it.
What is the report for and who reads it?
It turns a test result into a decision. A team takes an assessment, usually a short test set in its own industry, and the scores arrive as numbers. On their own they mean little. The report connects them to the work: if the team scores poorly on provenance, it should not supervise an agent that pays invoices, whatever its average.
The readers are the executive who approves the first deployment, the operations head who will own it, and whoever funds the training. Write for them, not for the learning team. Background on the method is in what is a readiness assessment? and what does a readiness assessment for AI agents measure?.
How do you fill it in?
- Recommendation first. One paragraph: the first agent, its starting depth, and the training needed before it acts.
- Scope. Who took the assessment, how many, which roles.
- Scores by concept. Not by person. Readers need to see where the team is weak, not who is.
- Gaps that matter. Link each weak concept to a specific risk in the planned agent's work.
- First agent. Why this one: low depth to start, clear sources, a process the team knows well.
- Training plan. What closes each gap and by when.
- What would change the recommendation. The conditions under which you would move faster or slower.
- Sign-off. Process owner and sponsor.
The template
- Recommendation (one paragraph)
- Scope: team, roles, number assessed, assessment used
- Scores by concept
| Concept | Team score | Range across the team | Needed for the planned agent? |
|---|---|---|---|
| Why agents must be governed | |||
| Depth and the decision chain | |||
| Verdicts: held, refused, withheld | |||
| Provenance of sources | |||
| Certification and lapse | |||
| The record and ownership |
- Gaps that matter
| Gap | Risk in the planned work | Training to close it | Owner | By |
|---|---|---|---|---|
- First agent: process, capabilities, starting depth, sources, who reviews
- Training plan: who takes which track, practicum timing, certificates needed before depth 3
- What would change this recommendation
- Sign-off: process owner, sponsor, date
A worked example
A hypothetical contact centre of 22 people takes a readiness test set in customer service. Scores are sample values out of 10.
Recommendation. Start with a reply-drafting agent at depth 2 for order status queries. Certify four team leads as operators before any capability goes to depth 3. Do not start with refunds.
| Concept | Team score | Range | Needed? |
|---|---|---|---|
| Why agents must be governed | 8 | 6 to 10 | Yes |
| Depth and the decision chain | 7 | 4 to 9 | Yes |
| Verdicts | 6 | 3 to 9 | Yes |
| Provenance of sources | 4 | 2 to 8 | Yes, critical for refunds |
| Certification and lapse | 5 | 2 to 8 | Yes for leads |
| The record and ownership | 7 | 5 to 9 | Yes |
| Gap | Risk | Training | Owner | By |
|---|---|---|---|---|
| Provenance | Reviewers trusting a customer's screenshot over the order system | Provenance module plus calibration on refund cases | Training lead | Week 6 |
| Lapse | Leads unaware that a lapsed certificate moves work back to people | Lead briefing and a drill | Ops manager | Week 4 |
The provenance score drove the recommendation. Order status replies draw on one sealed source, the order system, so weak provenance skills do little harm there. Refunds rely on customers' evidence, which is declared or external, and need reviewers who can tell the difference. A step-by-step version of the assessment is in how to run an AI agent readiness assessment in a contact centre, and a different sector's shape in how to run one in a university. Once training is done, report the result with the training impact report template for AI agents. The wider order of training is in training teams for AI agents.
Where the Academy fits
Pryme Intelligence Academy runs a free readiness assessment for teams of up to 25 people. The team takes the Foundations test set in its industry, and the report shows scores by concept, the gaps, and the first agent to start with, which is the structure this template follows.
Questions people ask
Should individual scores appear in the report?
No. Show team scores and ranges. Individual results belong in each person's training record and their manager's conversation with them.
What score is good enough to start?
There is no single number. A team can start at depth 2 with modest scores if a few certified reviewers check every draft. Depth 3 and above needs the concepts tied to that work to be strong across everyone who approves.
How often should we reassess?
After the first round of training, and again before any capability moves up a depth. New joiners should take the same assessment in their first month.