Lesson objective
Build a session plan in which every objective has an activity that rehearses it on fictional material and a piece of evidence the instructor can inspect before the session ends.
Plain-language explanation
Why this matters now
Lesson 1 gave you the Monday behaviour. This lesson turns it into a plan. The plan has three columns, and they must line up: what the learner will be able to do (the objective), what they do in the session to rehearse it (the activity), and what you will see that shows the rehearsal happened (the evidence).
When the columns do not line up, the session looks busy and proves nothing. The objective says "check claims against the file", the activity is a talk about substantiation, and the evidence is the attendance sheet. Everyone was present. Nobody checked a claim.
This is practice, and this programme labels it as practice. It holds no research on how adults learn or which methods work best, so the argument here is practical: you cannot tell whether a behaviour was rehearsed unless it left something behind that you can look at.
The evidence columns also connect to how professional bodies describe responsibility. Professional guidance (PRSA, for members): ESA #22, July 2026, applies the PRSA Code's Independence provision to say practitioners "remain responsible for evaluating all AI outputs." Evaluating is an action. In a session, you can watch a learner evaluate a fictional output, and you can read what they wrote about it. That is evidence. A nod is not.
A question to sit with
If your session were interrupted halfway through and a senior partner asked "how do you know anyone can do this yet?", what could you hand them?
What goes wrong
1. The objective and the activity drift apart. The objective is about briefing creators; the activity is watching a tool generate captions. Interesting, but it rehearses a different behaviour.
2. The activity uses real material. Learners practise on a live client's draft or a real customer list. The practice becomes client work, and client data enters a training room. Module 06 covers labs in depth; here the rule is simple: fictional material only.
3. The evidence is attendance or a feeling. "Participants said the session was useful" and "all twenty attended" describe the session, not the behaviour.
4. The evidence arrives too late. A form sent a week later measures what people remember of the session, not whether they could perform the behaviour in it.
5. The minutes do not add up. The plan gives ninety minutes of content to a sixty-minute slot. Something gets cut on the day, and it is usually the practice.
The practical answer
Best practice: use observable verbs in every objective. An objective you cannot watch is a hope. The table below lists verbs you can see, and the ones to replace.
| Verbs you can observe | Verbs to replace |
|---|---|
| Mark, tick, strike, check against | Understand, appreciate |
| Write, draft, rewrite | Be aware of, know about |
| Stop, pause, escalate, route | Recognise the importance of |
| Record, sign, date | Be familiar with |
| Sort, classify, choose | Explore, consider |
Best practice: build the plan in three columns that line up, row by row.
| Objective (observable) | Activity on fictional material | Evidence you can inspect in the session |
|---|---|---|
| Check each factual claim in an AI draft against the client's substantiation file | Learners mark every claim in a fictional draft for a fictional brand against a fictional file, with two planted unsupported claims | The marked draft: which claims were ticked and which struck |
| Write a disclosure into a creator brief | Learners complete a fictional creator brief, including the disclosure wording and where it appears | The brief's disclosure section |
| Stop a send when the consent record is missing | Learners review three fictional scheduled sends, one with no consent record | A written stop note for the right send |
| Record a named review | Learners review a fictional deliverable and complete the review record | The record, with name, date and what was checked |
Best practice: plant errors you know are there. An activity with no planted error cannot show whether a learner would catch one. Build the fictional material with a known number of problems, write them down privately, and compare what learners found with your list.
Best practice: collect the evidence before people leave. Walk the room during the activity, or collect the marked drafts and briefs at the end. Look at them while the learners are still there, and correct misses on the spot. The behaviour you corrected in the room is the one they take away.
Best practice: split the human and the tool in each objective. NIST's AI Risk Management Framework 1.0, a voluntary framework that binds no agency and that NIST stated in September 2026 was being revised, says in its Appendix C that human roles in oversight "need to be clearly defined and differentiated." In an objective, that means naming what the tool does ("drafts the brief") and what the person does ("writes and checks the disclosure section, and signs the review"). A learner should never leave unsure which part is theirs.
Best practice: make the minutes add up, with practice protected. Write the minutes beside each row and total them. If the total exceeds the slot, cut information, not practice. A workable shape for a sixty-minute session is ten minutes of setting up the behaviour, thirty-five minutes of activity with planted errors, and fifteen minutes of reviewing the evidence together. Ten plus thirty-five plus fifteen makes sixty. That is a shape from running sessions, not a researched ratio.
Link to the endorsement controls. The creator-brief row is not only a teaching device. The FTC staff FAQ describes a reasonable training and monitoring programme that includes instructions to endorsers and guidance on disclosures. A brief with a disclosure section is the artefact that programme produces. The FAQ is staff guidance and not binding; Lesson 3 builds this out.
Remember this
- Every objective has an activity that rehearses it and evidence you can inspect.
- Use verbs you can watch: mark, write, stop, record.
- Practise on fictional material with planted errors you have listed privately.
- Collect and read the evidence before people leave, and correct misses in the room.
- Total the minutes and protect the practice; the split is practice, not a researched ratio.
Marketing agency example
Dario Venn, a fictional L&D lead at the fictional Kestrel Media Partners, reviews a colleague's plan for a sixty-minute session for account teams. The plan reads: "Objective: understand AI risks. Activity: presentation and discussion. Evidence: feedback form." Dario rebuilds it in three columns. Objective: "Check each factual claim in an AI-drafted client email against the client's approved fact sheet." Activity: each learner receives a fictional email for the fictional Lantern Bay Coffee, drafted with an AI tool, containing nine factual claims. Dario has planted three that are not in the fictional fact sheet: a sourcing percentage, an award and a delivery promise. Evidence: the marked email. He writes the minutes: ten to set up the behaviour, thirty-five for the activity, fifteen to review. Ten plus thirty-five plus fifteen makes sixty. On the day there are twelve learners. Dario collects the marked emails at the thirty-five-minute mark and sorts them during a short break in the activity. Nine learners struck all three planted claims. Two missed the award. One missed the delivery promise and ticked it as supported because it "sounded like something they would say". Nine plus two plus one makes twelve. In the review segment, Dario shows the delivery promise beside the fact sheet and asks the room where the sentence came from. Nobody can find it. He then asks each learner to write the review record for their email: their name, the date, what was checked and against which file. He keeps the marked emails as the session's evidence, notes that they are practice on fictional material, and tells the account director the result as a description of this group, not a statistic.
Responsible-use warning
Never write an objective whose verb nobody can observe, and never accept attendance or a feedback form as evidence that a behaviour was rehearsed. Never let learners practise on a real client's draft or a real customer list. Never let the plan run longer than the slot, because the practice is what gets cut. Never present a time split as a research finding; it is a shape from running sessions. Never reuse a learner's marked practice material as client work.
Human-review requirement
The instructor owns the session plan, the fictional material and the planted-error list. Before delivery, a second reviewer confirms that every objective uses an observable verb, that each has a matching activity and inspectable evidence, that all material is fictional, and that the minutes total the slot. After delivery, the reviewer reads a sample of the evidence against the planted-error list. The agency's named reviewers remain accountable for all real client deliverables.
Practical exercise
Build a three-column plan for a sixty-minute session. Use a fictional agency, fictional brands and fictional documents throughout.
- Choose two of the four Monday behaviours from Lesson 1 and write each as an objective with an observable verb.
- For each objective, design an activity on fictional material: a fictional draft, brief, send list or deliverable.
- Build the fictional material and plant a known number of errors in it. Write the planted-error list on a separate page.
- For each objective, write the evidence you will inspect in the session and when you will collect it.
- Name, for each objective, what the tool does and what the person does.
- Write the minutes beside each row, total them, and cut information, not practice, until the total equals sixty.
- Swap your material with a colleague, have them complete the activity, and compare their findings with your planted-error list.
- Record what they found, what they missed and what you would change in the material or the instructions.
- Label each design choice in your plan as best practice or as an AI Coalition Network recommendation.
- Write the two sentences you would say to a senior partner who interrupts halfway through and asks how you know anyone can perform the behaviour yet, pointing to the evidence you have collected so far.
Knowledge check
-
What are the three columns of a session plan, and what must be true of them?
Answer guidance: The objective, the activity and the evidence; they must line up so each activity rehearses its objective and leaves evidence you can inspect.
-
Why plant known errors in the fictional practice material?
Answer guidance: Because an activity with no known error cannot show whether a learner would catch one; comparing findings with a private list shows the behaviour.
-
Why is a feedback form sent a week later weak evidence of the behaviour?
Answer guidance: It records opinions or recall after the event, not whether the learner could perform the behaviour during the session.
-
What should be cut when the plan runs longer than the slot?
Answer guidance: Information, not practice, because the practice is where the behaviour is rehearsed and the evidence is produced.
Sources and further reading
- Public Relations Society of America (PRSA), Board of Ethics and Professional Standards — AI Agents in Public Relations, Ethical Standard Advisory (ESA) #22 (2026-07-01). Accessed 2026-09-15.
- National Institute of Standards and Technology — Artificial Intelligence Risk Management Framework (AI RMF 1.0), NIST AI 100-1 (2023-01-26). Accessed 2026-09-15.
- Federal Trade Commission — FTC's Endorsement Guides: What People Are Asking. Accessed 2026-09-15.
- AI Coalition Network — Certification Educational Standard: Human Review of AI-Assisted Professional Work. Accessed 2026-09-15.