What is training completion rate?

Training completion rate is the proportion of people assigned a course or programme who have finished it, usually tracked by team and deadline. For AI agent deployments, it tells you whether people went through the training. It does not tell you whether they can judge a held action, which is the question that matters.

Why does completion rate matter at all?

It is a useful operational signal. If only a third of the review team has finished training two weeks before go-live, you know the date is at risk. Completion by team shows where managers are not protecting study time and where people need a nudge.

It is also often the first figure a board or regulator sees. That is where it causes trouble. A report saying "all reviewers completed AI agent training" reads as assurance. It only says everyone reached the last page.

What does completion rate miss?

Completion counts activity, not ability. A reviewer can click through every lesson and still approve a payment backed by an emailed invoice. The figure is blind to three things:

• Whether people learned: shown by test results, not completion.
• Whether they apply it: shown by the agent's record of what they approved and sent back.
• Whether it lasts: shown by retention checks weeks later.

The argument in stop measuring training by completion rates goes further: a high completion rate can hide weak training, because a course that is easy to finish is often easy to finish without learning.

An example

Two operations teams prepare for the same agent, which drafts supplier payments for approval. Team A reports full completion; Team B reports most completed. On the internal test, Team A's median score is lower, and in the first month Team A approves three held payments that should have been escalated. Team B's manager had required a pass before completion counted. Team A's course had no test at all.

How is it different from outcome measures?

Completion rate measures the training process. Outcome measures look at the work. For agent oversight, useful outcome measures include the quality of reviewers' decisions on held actions and how often they correctly send back a draft.

The agent's own metrics are a different family again. Escalation rate measures how often the agent hands a case to a person, and resolution rate how often a case is closed without one. These describe the agent, not the people. A trained team may push an agent's escalation rate up at first, because reviewers start catching cases they used to miss; see how high an AI support agent's escalation rate should be.

Which terms are related?

• Pass rate: the share who passed the assessment, a stronger measure.
• Training evaluation: the wider assessment of whether training worked.
• Retention check: whether learning has stayed.

For the full approach, read training teams for AI agents.

Where the Academy fits

The admin view in Pryme Intelligence Academy shows each person's progress, results, certificates and expiry dates, not only who finished. A certificate there rests on a pass, so a team lead can report certified reviewers rather than completions.