Every agency owner I speak to asks the same question: “Should I hire someone for AI?”
The answer is almost always yes. But the role they are imagining is almost always wrong.
You do not need a Head of AI. You do not need a Chief AI Officer. You do not need a machine learning engineer with a PhD. What you need are people who can use AI to make your existing work better, faster, and more profitable. That is a very different hire.
The roles that actually matter
There are three types of AI-capable people your agency needs. None of them require “AI” in the job title.
1. AI-literate strategists. These are your account directors, planners, and senior creatives who understand where AI fits in a client’s business. They do not need to build automations. They need to know what is possible, what is practical, and what is nonsense. When a client asks “should we use AI for this?”, your strategist needs to give a credible, specific answer.
2. Prompt-skilled creatives. Copywriters, designers, and content producers who can use AI as a creative tool, not a crutch. They know how to write prompts that produce useful starting points, how to iterate on outputs, and how to blend AI-generated material with original thinking. The difference between a creative who uses AI well and one who does not is roughly 2-3x output at comparable quality.
3. Automation-capable ops people. Operations managers, project managers, and production staff who can build and maintain AI-powered workflows. They connect tools together, build prompt templates, set up automated reporting, and keep the systems running. These are the people who turn one-off AI experiments into repeatable processes. If you have read our piece on how to implement AI in your agency, these are the people who make that implementation stick.
Hire or upskill?
For most agencies under 50 people, upskilling is the better first move. You already have people who know your clients, your processes, and your culture. Teaching them AI skills is faster and cheaper than hiring someone with AI skills and teaching them your business.
The exception: if nobody on your team has the curiosity or aptitude to pick up technical workflows (connecting APIs, building automations, writing structured prompts), you need to hire for it. One strong operations hire with automation experience can transform an agency of 15-20 people.
A practical approach:
- Upskill your strategists through structured learning (workshops, guided experimentation, real project application). Our guide on how to train your agency team on AI covers this in detail.
- Upskill your creatives through prompt engineering training and a shared prompt library.
- Hire (or contract) an automation specialist if you do not have one internally. This is the hardest skill to develop from scratch.
What to look for in interviews
Forget asking candidates to define “large language models” or name five AI tools. That tests knowledge, not capability.
Instead, ask these questions:
“Show me something you built with AI.” You want to see actual work. A workflow they automated. A prompt system they created. A process they improved. If they cannot show you something tangible, they are a user, not a builder.
“Walk me through how you would use AI to [specific task from your agency].” Give them a real brief. Watch their thinking process. Do they jump to a tool, or do they start with the problem? The best candidates think about the workflow first and the tool second.
“What has AI been bad at in your experience?” Anyone who says “nothing” does not actually use AI seriously. You want someone who understands the limitations: hallucinations, inconsistent quality, the need for human review. This is someone who will use AI responsibly with your clients.
“How do you quality-check AI output?” This separates the competent from the reckless. Someone who pastes AI output straight into a client deliverable is a liability. Someone who has a systematic review process is an asset.
The ChatGPT user vs. the workflow builder
This is the critical distinction most agencies miss. There is an enormous gap between someone who uses ChatGPT to write emails and someone who can build AI into your agency’s operations.
A ChatGPT user can write a prompt, get an output, and copy it into a document. Useful, but limited. They are essentially using AI as a search engine with better answers.
A workflow builder can design a system where client briefs are automatically structured, research is pulled from multiple sources, first drafts are generated with consistent brand voice, and the whole thing runs with minimal manual input. They think in systems, not prompts.
When you are hiring, test for the second type. The first type you can train internally in a few weeks. The second type takes months to develop, and some people never get there because it requires a particular kind of systems thinking.
Realistic salary expectations
AI skills carry a premium, but it is not as extreme as the hype suggests. In the UK market for 2026:
- AI-literate strategist/account director: £55,000-£80,000. This is not much above standard rates because the AI skills are layered on top of existing strategic capability.
- Prompt-skilled creative (mid-level): £35,000-£50,000. Again, not a huge premium. The value shows in output, not salary demands.
- Automation/AI operations specialist: £45,000-£65,000. This carries the biggest premium because it is the hardest role to fill. People who can genuinely build and maintain AI workflows are in high demand.
For contract/freelance rates, expect to pay £400-£800 per day for strong AI operations people. They are worth it if they set up systems your team can maintain.
What not to hire for
Do not hire a “Head of AI” as your first AI role. You do not have enough AI infrastructure for someone to lead. They will spend six months writing strategy documents and leave because nothing got implemented. Start with doers, not directors.
Do not hire based on certifications alone. The AI certification market is flooded with courses that teach theory, not practice. A candidate with no certifications but a portfolio of real AI workflows is more valuable than someone with five certificates and no applied work.
Do not hire someone to “figure out AI” for you. That is your job as the agency owner. You need to understand the strategy. You are hiring people to execute it. If you need help figuring out the strategy itself, that is what consultants (like us) are for.
The bottom line
The AI skills gap in agencies is real, but it is not solved by creating a new department. It is solved by embedding AI capability across your existing team, with targeted hires to fill the gaps your current people cannot close.
Start by assessing where your team is today. Then decide what you can train and what you need to hire. Get that balance right, and you will build an agency that uses AI as naturally as it uses email.
This is part of Margin Watch, a series on how AI is reshaping the business of running an agency. Subscribe to the newsletter to get new articles weekly.
