Will AI take your marketing job, and does everyone in your agency need to learn to build with it? The honest answer to the second question is no, and getting that right makes the first question far less frightening. Everyone benefits from AI. Not everyone needs to build it.
The anxiety is well founded. At a town hall during its AI Marketing Strategies event, executives admitted to Digiday that some employers are actively avoiding entry-level hires as AI absorbs junior work. One small-to-mid-sized performance agency owner said: "We haven't hired college-level grads in some time, and I don't anticipate that we will be." Meanwhile, Gartner research reported by Marketing Dive found that of 1,303 senior marketing leaders, 18% have already eliminated some positions because of AI and 16% have created new roles.
Faced with that, many agency leaders reach for the same fix: make everyone an AI builder. One agency founder tried exactly that, and told an AI adoption session in Singapore this week why he wouldn't do it again.
What happens when an agency makes AI everyone's job?
His agency of around 20 people gave everyone their own code account: strategists and creatives, but also HR and finance. They set aside 20% of the week for it. Fridays became "build Fridays".
The result split the team down the middle. Half got it. Half didn't. And the half who got it produced a sprawl of random skills and automations: one person's shortcut for a performance client, another's fix for a different account, none documented, none maintained, and nobody quite sure what triggered what.
Any agency person will recognise the pattern. It is the shared drive, except now it runs. Picture a media planner's bidding script that only works on her laptop, or a copywriter's prompt chain that lives in a personal account and leaves when he does.
What are builders, users and approvers?
The fix was to appoint roles. As the diagram shows, there are three, under one line from his slide: "Everyone benefits from AI. Not everyone needs to build it."
| Role | Who they are | What they do | Agency example |
|---|---|---|---|
| Builders | Not engineers: people with passion who dedicate more time | Design, build and improve the agency's AI workflows, and document them | A systems-minded planner who turns the monthly media report into a repeatable workflow |
| Users | Everyone who works in the systems day to day | Run the work and feed back bugs and requests | Account executives producing client updates who flag when an output looks wrong |
| Approvers | Subject-matter experts who own quality control | Check the output; nothing ships until they have signed it off | The head of strategy who reviews every insight deck before it reaches a client |
The most important detail is the first one: builders are not engineers. They are the curious, patient people willing to give the time. In a creative shop that might be a producer. In a CRM agency, a data-minded campaign manager. In a PR firm, the account director who already wrote the agency's press-release templates.
Approvers matter just as much. They are the people who define what good looks like for the agency's work, and they are the reason a half-finished AI draft never lands in a client's inbox.
Why does "everyone builds" feed imposter syndrome?
The imposter phenomenon is the persistent feeling that you are less competent than others believe, and that sooner or later you will be found out. It thrives when the rules of competence change overnight.
A blanket "everyone builds" mandate does exactly that. A senior copywriter with fifteen years of craft is suddenly judged on whether she can wire up an automation. The half of the team who "didn't get it" don't feel freed from chores; they feel behind. The half who did get it start to wonder whether they are still strategists or have quietly become junior engineers.
Clear roles reverse the pressure. A user isn't failing at building; they are doing a defined, valuable job. An approver's experience becomes the agency's quality standard. That is the logic behind the insight: authenticity thrives where insecurity ebbs. People do their most honest work when they know what they are there for.
What happens when your AI hybrids leave?
The founder asked a question every agency MD should sit with. If you turn people into part-engineer, part-expert, part-client-facing hybrids, what happens when they leave? And what happens when you need to hire and scale?
You can't advertise for that person. The market doesn't produce account directors who also maintain your custom automations. When one resigns, the skills they built either break or become archaeology.
Roles solve the succession problem. Builders document what they build because other people depend on it. Users can be hired for their client or craft skills and trained on the system quickly. Approvers stay anchored in subject expertise, which is what clients pay for.
This matters for the junior question too. Gartner predicts most high-performing teams will eliminate traditional entry-level jobs by 2030, but it urges CMOs to give juniors higher-value work earlier rather than cut the pipeline. Defined roles give that advice a shape: a graduate can start as a user with clear feedback duties, then grow into a builder or, with time and judgement, an approver.
How do I assign AI roles in my agency?
- Drop the blanket mandate. Keep access for everyone; drop the expectation that everyone builds.
- Audit what has been built. List every skill, prompt chain and automation in use. Keep, merge or retire each one, and give every survivor an owner.
- Choose builders by appetite, not job title. Look for the people who already tinker. Give them protected time, rather than spreading the same slice across everybody.
- Name an approver for every workflow. A subject-matter expert who defines what good looks like and signs off before anything reaches a client.
- Give users one feedback channel. A single place to log bugs and requests, reviewed every week by the builders.
- Write the roles into job descriptions. Roles that live only in a Slack message disappear at the next reorganisation.
So is AI a threat to marketing job security?
To tasks, yes. To judgement, much less. The work most exposed is the chores: monthly reports, competitor audits, manual resizing, status meetings. The work least exposed is deciding what matters, what is good and what to tell the client.
That is also the right split between people and software. Defy Agents run agentic workflows for repetitive delivery, such as reporting, monitoring and resizing, so people can focus on the work that wins. SAMY, the global influencer agency, adopted Defy Agents to streamline performance reporting, reducing three days of work to one. That is the kind of chore no strategist went into advertising to do.
If you want to see how agents fit alongside monitoring and data access, the Defy platform shows the whole stack. Whatever you use, the principle holds: give people roles, not homework.
