After an AI keynote: which follow-up is worth commissioning?

The event is over. The organiser sends the slides, the recording and a satisfaction survey. Each can be useful. Yet the next procurement decision remains open: what additional work would help participants use the talk, and what evidence would justify paying for it?

For an AI keynote, that decision belongs in the preparation conversation. A pharmaceutical leadership meeting may seek a common vocabulary for investment choices. A medical association may want members to scrutinise an AI claim. A company may want a particular team to examine a proposed change. Those objectives call for different follow-up deliverables.

Choose the intended result before choosing the format

If the objective is orientation, commission a concise decision brief: the central distinction, the supporting sources and the questions left open. Readers can return to it when the subject arises. A recording provides access to the talk; a short brief makes the relevant reasoning easier to locate. Neither download counts nor positive ratings demonstrate that readers can apply it.

If the objective is learning, agree a small follow-up activity with an explanation of the answer and a way to identify unresolved questions. Link it to one capability from the talk. The existing guide to an AI learning session explains how to construct a practice task; the organiser’s decision here concerns who prepares, delivers and interprets the follow-up.

If the objective is an organisational decision, nominate the team that will own a subsequent discussion. Its output could be a short record of the assumption to examine, the evidence still missing and the person responsible for obtaining it. A general keynote does not itself approve a project. The organisation retains responsibility for the resulting decision.

These are proposed options for the preparation brief, rather than a validated package. Selecting one can be more useful than commissioning all three without a recipient or purpose.

What research supports, and where it stops

Larsen’s 2015 report analysed 35 participants in medical-conference courses. Within each person, material was assigned to repeated quizzes with feedback, restudy or no follow-up. Quizzed material was better retained 5.5 months later. The test repeated the questions; substantial attrition limits generalisation. Changed practice was not measured.[1]

Van Eersel and colleagues’ 2016 experiment analysed 54 psychology undergraduates. Retrieval with feedback outperformed rereading on a transfer test one week later. The advantage shrank when rereading also included focused key statements, but remained. Unequal study time limits interpretation. This was a laboratory task, not an event follow-up.[2]

For commissioning, my inference is to specify the explanation and comparison material as carefully as the questions. A question without a useful answer discussion is a different deliverable from the intervention examined in these studies. Neither paper establishes an optimal reminder schedule for an AI keynote or a financial return from one.

Decide what the report will be allowed to claim

CDC’s training-evaluation guidance distinguishes learning from workplace application. It notes that a posttest alone cannot establish improvement from the starting point. Follow-up timing depends on the objective and the opportunity to use the material. These distinctions are useful when agreeing an event’s evaluation scope.[3]

Consider a fictional event with 200 invited participants. Forty submit a later exercise; 30 meet the agreed reasoning criteria. The report can say that 30 of 40 respondents met those criteria. It cannot describe 75% of the invited audience as having acquired the capability. The other 160 provided no evidence either way. Existing knowledge could also explain a good answer.

Even if the same people improve between comparable tasks before and after the event, that observation alone does not isolate the talk’s effect. Other learning, practice and the follow-up itself may contribute. A causal question needs a suitable comparison and an evaluation design capable of addressing it.

For routine planning, keep four observations separate: whom the follow-up reached, who completed it, what their work demonstrated and whether a documented workplace action followed. An action record shows that something happened; it does not by itself demonstrate that the decision was better. This is my proposed reporting structure, not a new measurement standard.

Put responsibility and effort into the booking brief

Before the event, record the intended result, the deliverable, its recipient, the responsible person and the evidence to collect. Include the preparation effort, delivery date and time participants are being asked to give. Choose the date around a realistic opportunity to use the content, rather than treating a particular interval as scientifically optimal.

Explain whether participation is voluntary, who will see responses and how long they will be kept. An optional exercise should not quietly become an individual performance assessment. Use fictional or explicitly cleared cases. Participants need an accessible way to take part and a clear route for questions; confidential work examples need not be collected to make the activity useful.

As the founder of aiomics, I value the connection between a discussion about AI and a decision someone can responsibly take afterwards. For event preparation, one sentence makes that connection concrete: after the talk, this person will receive this deliverable, use it for this decision and report these observations. That sentence is a useful addition to the booking brief.