Most agencies know they should be using AI. Few know where to start. The ones that do start often pick the wrong thing, get underwhelming results, and conclude that AI is not ready for agency work.
It is. They just started in the wrong place.
Here is a practical roadmap for implementing AI in your agency, based on what we have seen work across 100+ agencies of different sizes and specialisms.
Start with the bottleneck, not the buzzword
The biggest mistake is picking a tool and looking for a problem it solves. That is backwards. Start with the bottleneck. (If you are not sure where to look, our AI agency audit guide can help.)
Every agency has one. It might be:
- Proposals take too long. You spend 6-8 hours on a pitch that might not land.
- Scoping is inconsistent. Every project starts with a different level of detail depending on who runs the discovery call.
- Admin is eating your margins. Status updates, timesheets, meeting notes, and follow-ups consume hours that should be billable.
Pick the one that costs you the most time or money. That is where AI goes first.
The three layers of agency AI
AI in an agency is not one thing. It works across three layers, and you should implement them in order.
Layer 1: Sales and new business. This is where most agencies see the fastest return. AI can research prospects before a call, draft tailored proposals, refine pricing language, and track how prospects engage with your documents. The time savings are immediate and measurable.
Layer 2: Delivery and production. This is where AI handles the repetitive parts of client work. Structured briefs from client inputs, first-draft copy, code generation, image creation, meeting summaries. The goal is not to replace your team. It is to remove the low-value work so they can focus on the thinking.
Layer 3: Operations and internal. This is where AI runs the back office. Automated reporting, resource planning, invoice generation, internal knowledge bases. This layer takes longer to set up but compounds over time.
Most agencies try to start at Layer 3 because it feels safe (it is internal). But Layer 1 delivers faster results and builds confidence across the team.
Implementation in practice
Here is the process we use with agencies we work with:
- Audit your current workflows. Map out every step in your sales, delivery, and operations processes. Mark the ones that are repetitive, time-consuming, or inconsistent.
- Pick one workflow. Not three. Not five. One. The one where the time saving is clearest and the risk is lowest.
- Build the system. This means choosing the right tool, writing the prompts or setting up the automation, and testing it on real work. Not a demo. Real client work.
- Measure the result. Time saved, quality improvement, team feedback. If it works, document it and move to the next workflow. If it does not, adjust or move on.
- Train the team. AI systems only work if your team actually uses them. That means documentation, walkthroughs, and giving people time to get comfortable.
The agencies that succeed treat this as a continuous cycle, not a one-off project. They implement one system, measure it, then move to the next.
Common mistakes
Trying to do everything at once. You do not need an AI strategy that covers every department. You need one working system that proves the value, then you build from there.
Choosing tools before defining the problem. Every week there is a new AI tool promising to transform your agency. Ignore the noise. Define the problem first, then find the tool that solves it.
Not involving the team. If your team sees AI as a threat rather than a tool, adoption will fail. Bring them in early. Our guide on how to train your team on AI covers how to do this without the resistance.
Expecting perfection from day one. AI outputs are a starting point. They need human review, refinement, and judgement. The agencies that get the best results treat AI as a capable junior, not an autopilot.
Where to start this week
If you have not started yet, do this:
- Write down the three tasks in your agency that take the most time relative to their value.
- Pick the one that is most repetitive and least creative.
- Spend an hour testing whether AI can handle the first 80% of that task.
That is your starting point. Not a strategy document. Not a board presentation. One hour, one task, one test.
The agencies pulling ahead right now are not the ones with the biggest AI budgets. They are the ones that started. And the ROI data backs it up.
This is part of Delivery Notes, a series on implementing AI inside your agency. Subscribe to the newsletter to get new articles weekly.
