The Five AI Moves Every Agency Needs to Make Now
These are practical, people-first steps to build your capability, without turning your agency into an AI petri dish.
Most independent agencies are doing the sensible thing right now: experimenting, learning, keeping an eye on the market, trying not to freak the team out, and getting ready for the inevitable pressure which will be coming from clients in the not distant future.
This post is here to help you take some meaningful first steps.
These five moves aren’t about how to implement the latest and most hyped tool, but rather start to build durable capability. The kind which can improve the work, protect your people and lay the foundations for keeping your agency strong over the next five to ten.
Move 1: Decide what AI is for your agency
You need your team to not be guessing the agenda. Same with your clients.
From what we can see, most agencies have skipped this step. They signed up to some tools, encouraged self-learning sessions and let different teams figure it out (or not) in their own way. No surprises that the result wasn’t all that predictable - inconsistent output, mixed confidence, uneven quality, a sense that everyone is improvising and a AI Slack channel which is equal parts outrage and enthusiasm.
Of course you need a clear strategy, but for now, a clear stance will do the trick.
A simple way to get to your stance is choosing your primary intent for the next 6-12 months:
AI Acceleration - remove friction and speed up the work you already do
AI Expansion - unlock new kinds of ideas, experiences and outputs you couldn’t deliver before
AI Evolution - start building structured, repeatable AI capability across the agency that can improve as people and the technology does
Don’t feel like this is a lifelong contract, you can always change your approach as your agency and the world moves forward. All you have to do right now is choose a direction to row in, so everyone can pull in the same direction. And once you have this clear intent, everything becomes easier. Training becomes more focused, workflow choices become obvious and governance becomes practical instead of uninformed paranoid.
Move 2: Create your minimum viable AI capability model
One page that sets the rules of engagement, lowers anxiety and starts setting the scene for better work.
This is when you make a decision to move away from ‘AI is happening’ to something your team can get behind.
It should just be a single page that defines how AI will be used in your agency. Not a policy document nobody reads (that has been debated through 8 rounds of refinement in your management meetings), but a working agreement that creates psychological safety, clarity and protects the craft your agency is known for.
It should cover:
What we use AI for - the tasks and workflows we actively want to improve
What we won’t use AI for - creative, ethical, legal or reputational red lines
Who owns it - one accountable lead, with a support person. Definitely not ‘everyone’
What good looks like - quality standards and creative non-negotiables
How we handle risk - data boundaries, IP checks, disclosure expectations, review gates
This is your agency’s initial operating guardrail. It needs to reduce fear, prevent chaos and create the space for your team to use AI well. You should be able to explain your AI stance in under a minute.
Move 3: Pick 2-3 high value workflows and upgrade them properly
Start small. Make them real and work. And then let the wins build the momentum for the next thing you implement.
Let’s be clear. You really shouldn’t just dive into rolling out AI everywhere. It’s better to choose a few workflows that matter and redesign them intentionally. Low hanging fruit and low budget exercises should always be your first point of call. And they should be created with the people who do the work. Think about moving through the stages of crawl, walk, run. You can’t do the later stages effectively without being secure in the earlier ones.
Pick workflows that are:
Frequent - preferably weekly, not annually
High drag or effort
High leverage (improving them will improve everything downstream)
Visible (you people to see and feel the wins
Some examples could be:
Supercharging your visual prompts, so you get better results faster
Going from automated meeting notes to action brief faster
Final work QA covering off proof reading and adherence to BVI
Think about:
Where and how could AI speed this up?
Where and when do humans take over (taste, judgement, decision-making)
What the output is for each
What checks need to happen before anything goes client-facing
This is also where job protection becomes real in practice. You’re using AI to remove the grind and expand potential, not shrink teams or compromise craft. When your teams see this clearly, adoption is less scary and more useful.
Move 4: Upskill by role, not hype
Train people on what changes in their job. Ignore the rest.
A lot of AI training in agencies is well intentioned, but unfocused. There are a couple of Lunch & Learns, documents filled with the latest tools and an expectation that teams can self-learn their way to fluency.
It’s much simpler and easier to train people on the parts of AI that specifically increase the value they have in the role they already have. (Obviously, once they have a basic grasp, which almost everyone seems to have now.)
Some thought-starters of what to train teams to use AI for:
Strategy - document and research synthesis, sharper integration and testing of framing, stronger decision support, synthetic audience creation and tuning
Creative - AI-enhanced prompting for visual Gen AI tools, expanded creative exploration with agentic tools, stronger work evaluation with synthetic audiences, checking partner, copy structure accelerator
Design - consistent style control, roll-out automation, rapid design-system / brand world exploration
Account / client leadership - automated meeting to action brief, accelerated cost estimates, risk literacy, trust and disclosure, proper governance
Production and project management - workflow redesign, QA gates and enhanced checking, AI-enhanced production methods, creating repeatable delivery systems
Leadership - capability mapping, governance, making AI adoption feel safe and coherent
You’re looking to create capability and a team that can partner with AI to enhance what they do. They should still be focused on pushing their craft, directing AI with taste, knowing how to evaluate output quickly with the right level of trust in AI work (always treated as something to be checked), and consistently produce work that still feels human and distinct.
We focus on a 10 | 80 | 10 approach. The first 10% is your high quality human thinking to get initial concepts or ideas down, a super sharp brief on how you want to work with the AI, and an idea of what you want out at the end - you’re setting the bar. You then shift into the 80% where the human + machine partnership kicks off with exploration, research, sounding boards, idea generation (but rather get the AI to help get better ideas out of you); all greatly expanded and accelerated. Finally the last 10% is all you and your exquisite human human refinement, discernment and craft.
This is also how you protect careers, particularly junior pathways. AI changes the ‘grunt work’, but increases the need for people who can orchestrate, curate, craft and judge. Training is how you steer that shift in a positive direction.
Move 5: Start building agentic workflows
This is where capability compounds, whether it’s small systems or complex workflows
Gen AI is really useful, and most of us have our head around it.
Agentic workflows are the next step. This is where systems don’t just generate off your prompt, but acts independently to support a process. It breaks work into steps, routing tasks, checking quality and handing off to humans at key moments to ensure things have been done right.
Doing this honestly doesn’t need a giant bespoke platform. There are great systems you can already plug into from the creative side, like Figma Weave or Flora, or a plethora of simple agent builders from the likes of Monday.com (Taka), ClickUp (Super Agents) and more. What’s even better is building smaller, bespoke modular chains that are unique to the way your team works, and makes specific workflows more reliable and repeatable, ideally improving over time.
Some examples could be:
Brief builder - Automatically generates structured agency briefs from client inputs and call notes, instantly editable and linked to resourcing, removing admin and freeing teams to focus on the work.
Auto estimate - Automatically generates first-draft estimates using past project data, giving teams an editable baseline that dramatically reduces scoping time and stress.
Checking buddy - Automatically checks work against brand and channel guidelines, catching issues early so humans can focus on nuance, craft and final polish.
Resizing or rollout creative workflow in a platform like Pletor or Weavy, helping reduce laborious tasks
Synthetic Client - An agent that emulates one of your tricky clients, so you can test work or responses
Done well, these systems:
Reduce work
Increase consistency
Help teams move faster without losing quality
Make best practice repeatable
Give juniors structure and seniors leverage
Create space for human judgement and craft
Start with one workflow. Test and prove it works. Get it adopted. Then bring in the next. Going iteratively will get you there better. A few well-adopted and well-tuned agents will be far more valuable than an army of super agents which are barely adopted. It’s not about having the most tools, but rather the most usable capability. Keep in mind that the more complex an agent, the more likely it is to go wrong. Rather have a number of focused agents who do one thing really well, and then an overseer agent who can triage requests and send them to the best agent for the job.
What you’re trying to achieve
All five moves keep people front and centre. And the point isn’t to replace them, but make them more than they are, by:
Removing the grind
Raising the ceiling on creative output
Protecting quality and trust
Improving margins through smarter delivery
Creating a healthier agency that’s better prepared for the next decade, where people want to work.
Like some help with this? You can find out more here about how we can help.