Lesson objective

Describe the marketers likely to attend an AI session by role and by the output each one owns, using survey figures only with their population and caveats.

Plain-language explanation

Why this matters now

Levels 1 to 5 taught you to use AI in agency work, to set rules for it, to govern it and to automate parts of it behind approval gates. Level 6 asks you to teach other marketers to do the same. You may be the agency's training lead running a lunchtime session, a creative director coaching a team through a new tool, a faculty member teaching a marketing course, a bootcamp or association instructor, or a consultant brought in to train a client's in-house team.

The people in front of you are not beginners in the usual sense. They are strategists, copywriters, art directors, media buyers, social and community managers, influencer managers, PR staff, email marketers, analysts and account leads. Most have years of practice. Most bill their time, or answer to someone who does. Many have already tried an AI tool, formed an opinion of it, and brought that opinion into your session.

Two sources describe the conditions these learners work in. Each must be taught with its limits.

Survey evidence (not binding): The CMO Survey, 35th edition, results posted 31 March 2026. The results are based on 308 marketing leaders at for-profit US companies, 97% of them at vice-president level or above. The post reports that training budgets stand at 3.8% of marketing spending, down from a pre-pandemic high of 5.8%. It reports that AI use in marketing "has more than tripled since 2022", and that companies project AI will account for more than half of all marketing activities within three years. It also reports that no marketing technology activity scored above 5 on a 7-point performance scale, and that the barriers cited were organisational: budget, integration, bandwidth and talent.

Read that survey carefully before you quote it. It is a sponsored, self-selected survey of executives, not a probability sample; its partners, Deloitte, the American Marketing Association and Duke University's Fuqua School of Business, provide "intellectual and financial resources"; and the post does not give its fielding dates. The three-year projection is what respondents expect, not a measurement. The "more than tripled" line is a comparison; the underlying percentages sit in a full report this programme did not read. Most important for this profession: it describes marketing leaders at companies, not agencies. This programme holds no AI adoption figure for agencies or PR firms. Never state one.

Research (not binding): Humlum and Vestergaard, NBER Working Paper 33777, issued May 2025 and revised March 2026. Linking adoption surveys to administrative labour-market records in Denmark, the authors report that employers absorb AI through task reorganisation, including new tasks in content generation, AI oversight and AI integration. It is a working paper, not peer reviewed; the data are Danish; and the abstract does not say whether marketing professionals were among the occupations studied. It is a reason to expect that people in your room may have been handed new checking work alongside new drafting tools. It is not proof that any particular agency has done so.

Put together with their caveats, the picture is this: on one sponsored survey's figures, the share of marketing spend set aside for training is smaller than it was, while checking AI output is a kind of work organisations are adding. Your hour has to earn its place.

A question to sit with

A senior copywriter has been told to attend your session. She has a client call in ninety minutes and a draft still open. What must she leave with for your hour to have been worth more than finishing that draft?

What goes wrong

1. The room is treated as beginners. The session opens with "what is a large language model" for people who have used one daily for a year. The experienced learners stop listening, and they are often the ones whose habits matter most.

2. The adoption statistic opener. "Most agencies already use AI" or "80% of PR firms have adopted it" appears on the first slide. No verified figure exists for agencies or PR firms. The CMO Survey's "more than tripled since 2022" is a comparison from a sponsored executive sample about company marketing, not an agency adoption rate.

3. The projection becomes a fact. "AI will do half of all marketing within three years" is presented as a finding. It is what surveyed executives expect.

4. The session competes with billable work and loses. Fifty minutes of slides, with laptops open and client messages arriving. The session was scheduled, but the learners were never really there.

5. One session for every role. A media buyer, a copywriter and a PR lead are shown the same demonstration. They own different outputs, face different checks and answer to different rules, so each hears mostly about someone else's work.

6. The new checking work stays invisible. The session teaches how to get a draft from a tool and stops there. The checking, which is the part that protects the client and the agency, is never practised.

The practical answer

This programme holds no education or learning-science research. Nothing below is a finding about how adults learn, how much they retain, or which teaching method works best. It is practice, labelled as practice, built on the working conditions of marketing, which anyone in your room can check against their own week.

Best practice: describe the room by role and by the output each role owns. Titles vary from agency to agency; outputs do not. Check your registration list against a table like this one.

Role AI-assisted output they own The check that matters most What loses them
Copywriter or content writer Claim-bearing copy Each claim against the substantiation file A tour of prompt tricks
Art director or designer Synthetic images and edits Rights, likeness and a record of how the asset was made Being told the tool will replace them
Media buyer or trader Audience and bid suggestions Targeting against consent and data-use limits Examples drawn only from creative work
Social or community manager Replies and posts Disclosure, and whether a review or reply is genuine A long lecture before any practice
Influencer or creator manager Creator briefs The disclosure written into the brief Rules with no brief to apply them to
PR or media relations Pitches and statements Every fact and attribution before it goes to a journalist Advertising-only examples
Email or CRM marketer Scheduled sends Consent and opt-out state before the send Content with no channel in it
Account lead The deliverable the client sees The named review, recorded Detail with no decision attached

AI Coalition Network recommendation: design every session for one thing each role will do on the next working day. For a copywriter that might be checking three claims against the file before a draft leaves. For a media buyer, confirming the consent basis of an audience before it is uploaded. Module 02 builds the whole session backwards from that behaviour. This is our educational standard for sessions in this programme, not a requirement from any authority.

Best practice: respect the clock in the design, not only in the apology. Put the finish time on the first slide and keep it. Start from a piece of work the learners recognise, such as a fictional brief or a fictional draft for a fictional client, rather than from the tool. Move to practice early. If the session must be short, teach one check well rather than six checks badly.

Best practice: ask the room instead of assuming. A short form before the session can ask each person's role, which AI tool they use at work if any, what they use it for, and one time it produced something wrong that someone had to catch. Ask for no client names and no client data. The answers describe this room. They are not a statistic, and you never report them as one.

Best practice: name the new work. If the people in the room are now expected to check what a tool drafts, say so out loud and give that checking time in the session. The Danish working paper is a reason to expect oversight tasks to appear; your own registration form tells you whether they have appeared here.

Remember this

  • Marketers arrive experienced, busy and already opinionated; design for their work, not for a beginner.
  • There is no verified AI adoption figure for agencies or PR firms; never state one.
  • The CMO Survey is a sponsored, self-selected executive survey of company marketing leaders; its projections are expectations.
  • Describe the room by role and by the output each role owns, and ask the room rather than assume.
  • Every teaching approach here is practice or this programme's recommendation, never a learning-science finding.

Marketing agency example

Renata Solberg, a fictional training lead at the fictional Kestrel Media Partners, is asked to run a one-hour AI session for the whole agency. Her first draft opens with a slide reading "73% of agencies now use AI: don't fall behind." Before finalising, she looks for the source. She cannot find one, and this programme records no adoption figure for agencies or PR firms, so she deletes the slide. She considers replacing it with the CMO Survey line that AI use in marketing has more than tripled since 2022, then notes its limits: 308 marketing leaders at for-profit US companies, a sponsored, self-selected sample, and a comparison rather than a rate. She decides her room does not need a market statistic at all. She sends a four-question form instead: role, which tool if any, used for what, and one time it was wrong. Twenty-two of twenty-six people reply. Nine are copywriters or content writers, five are media buyers, four are social managers, three are account leads and one is the agency's only PR specialist. Nine plus five plus four plus three plus one makes twenty-two. Eleven say a tool once produced a claim or a figure that someone had to remove. Renata rebuilds the hour around three roles. Copywriters practise checking claims in a fictional draft for a fictional snack brand against its substantiation file. Media buyers check a fictional audience list against its recorded consent basis. Account leads practise writing the review record. She puts "we finish at 1:00" on the first slide and keeps to it. When the PR specialist says the session had little for her, Renata agrees and books a thirty-minute follow-up on fact-checking pitches. She tells the room that the structure is her practice, not a research finding, and that the form answers describe this agency, not the industry.

Responsible-use warning

Never open a session with an AI adoption figure for agencies or PR firms; none is verified. Never quote the CMO Survey without saying it is a sponsored, self-selected survey of company marketing leaders, and never present its three-year projection as a measurement. Never describe the room as one audience when its roles own different outputs. Never report your own pre-session form answers as a statistic about the industry.

Human-review requirement

The instructor owns every figure and description of the audience in the session material. Before delivery, a second reviewer, such as the agency's head of training or a fellow instructor, checks each figure against its source, confirms the population, sampling caveat and date are stated, and confirms no company or survey figure is presented as an agency adoption rate. The reviewer also confirms the pre-session form asks for no client names or client data. Each learner remains accountable for their own work under the agency's named-reviewer process.

Practical exercise

Profile a fictional room by role and output. Use a fictional agency, fictional staff and fictional form answers throughout.

  1. Write a fictional registration list of fifteen to twenty-five attendees for a one-hour AI session at a fictional agency. For each, record the role and the output they put in front of a client or the public.
  2. Count the attendees by role and check that your counts add up to the total on the list.
  3. For each role, write the one check that matters most for the AI-assisted output that role owns.
  4. Write one sentence you may truthfully say about the CMO Survey, with its population, its sampling caveat and its date.
  5. Write the sentence you will say when an agency principal asks what share of agencies use AI.
  6. Draft a four-question pre-session form that tells you who is in the room without asking for any client name, client data or personal information about customers.
  7. Choose the three roles your session will serve best, and write one behaviour each will perform on the next working day.
  8. Label every recommendation in your plan as best practice or as an AI Coalition Network recommendation, and delete any sentence that claims a research finding about how adults learn.

Knowledge check

  1. What may an instructor say when asked what share of agencies or PR firms use AI?

    Answer guidance: That there is no verified adoption figure for agencies or PR firms, and that the CMO Survey covers marketing leaders at companies in a sponsored, self-selected sample.

  2. What does the CMO Survey post report about training budgets?

    Answer guidance: That training budgets stand at 3.8% of marketing spending, down from a pre-pandemic high of 5.8%, among 308 marketing leaders in a non-probability executive survey.

  3. How should the survey's claim that AI will account for more than half of marketing activities within three years be described?

    Answer guidance: As what the surveyed executives project or expect, not as a measurement or a finding about what will happen.

  4. Why describe the room by role and by the output each role owns?

    Answer guidance: Because roles own different AI-assisted outputs with different checks and rules, so one generic session serves mostly someone else's work.

Sources and further reading