What does completed AI training tell a manager?
Completion, explanation and a reviewed work sample can support different management decisions.
An AI training completion report tells a manager which accounts or named learners are recorded as having completed an activity. It does not, by itself, show what people can do with AI at work or whether they are ready to take responsibility for a particular task.
Those are different management decisions. A completion report can help you follow up on an assigned course. A review of actual work can help you decide what support or permission a person needs next.
Start with the decision, not the completion percentage
Suppose a manager wants a team to use AI to prepare customer research. The team completes an introductory course, and the learning system reports that everyone passed. That is a useful record of the course activity. It leaves open whether the research contains the right evidence, how the team checks it and what happens when sources disagree.
Before asking for another metric, write down the decision you want to make. Are you deciding who still needs an introduction, whether the team can prepare a draft for review, or whether someone can send the result to a customer? Each decision needs a different amount of evidence.
The customer-research example is hypothetical. My experiment did not involve a company training programme or employees, so it cannot tell us how often a completion report is used this way inside a business.
What the experiment can and cannot show
From 25 to 28 October, I authorised AI agents to take free courses and assessments through my account. The experiment recorded 424 credentials issued to the account and verified on providers' public pages. The count included 23 aggregate awards issued on top of component awards. It was a deliberately selected set of free offerings, not a representative sample of professional training.
The provider verification established that the awards were issued. It did not independently establish what I knew or who supplied every answer. The register records AI-performed assessments, but human help with them was not measured. Some Saylor finals departed from that provider's expectation of individual work; my authorisation to use the account did not settle the provider's rules.
The complete experiment record shows the results and limits. My account of the 424 credentials explains the count and what remains untested. The management question here is narrower: what should a completion report be used to decide?
Match the evidence to the decision
Use a completion record to decide whose assigned course needs follow-up. Ask the person to explain one claim and its checks to decide what needs clarification or practice. Review a work sample against the intended responsibility to decide whether the work can proceed and with what supervision. These observations inform different decisions; none guarantees future performance.
Look at one piece of work
For the hypothetical research task, ask the person to prepare a short briefing on a company using the tools they are allowed to use. Agree on the task and the information they may access. Then look at the sources, the distinction between fact and inference, and how they handled something the sources did not settle.
Ask the person to walk you through one claim. Where did it come from? What did they check before relying on it? What would make them stop and ask someone else? If the briefing is intended for a customer, compare it with the approval and confidentiality rules that apply to that work.
This is a proposed management exercise, not a test validated by my experiment. The person may use AI well, need help checking a particular kind of claim, or need a clearer description of the task. A completed course alone cannot distinguish those explanations.
Decide what happens next
Use the result to choose a next step, not to give someone a permanent label. If the person can prepare a useful draft but misses source checks, give them a review step and a chance to practise that part. If access to customer information is uncertain, settle the permission before asking them to use it. If a result will be sent externally, decide who reviews it and what would trigger a pause.
The evidence should be proportionate to the responsibility. A manager may accept a draft as an input to an internal discussion while requiring a stronger check before the same information is used in a customer decision. Keep the example and feedback with the work where the team can return to it, subject to the organisation's information-handling rules.
Completion still has a place. It can tell you whether an assigned introduction has been recorded as complete and help you identify missing completion records for follow-up. It should not carry the extra claim that the team has adopted a tool, saved time or can perform a job independently unless you have evidence for those outcomes.
For a leadership team deciding where AI belongs in the business, AI consulting starts with the work, the decision and the responsibility the organisation needs to keep.