How do you build a 90-day AI implementation plan for your agency?
Summary: Most AI programmes fail not because the technology doesn't work, but because the learning never gets captured. Here is the 90-day AI implementation plan Spark AI uses with agency clients to turn a workshop into a working system.
Date: 04.08.2026
Author: Emma Wharton Love
Read length: 8 minutes
Spark AI is the AI performance partner that builds organisation-wide AI capability for creative agencies and brand teams. We take ambitious teams from fragmented experimentation to AI embedded across your people, processes and systems ā so it delivers real commercial impact.
A 90-day AI implementation plan works in four three-week blocks: set the rails, build the rhythm, scale what works, then commercialise. Each block ends in a named deliverable, and one person owns the whole thing.
A question that comes up constantly in our conversations with agency leaders: we're doing loads of AI stuff, so why hasn't anything changed?
Research from Harvard Business Review puts data behind that feeling. Their study, A Systematic Approach to Experimenting with Gen AI, describes a widening gap between people using AI and organisations embedding it. The researchers call it the productivity J-curve ā an initial dip as organisations adapt, followed by sustained gains once the right investments pay off.
Honestly, I think "J-curve" undersells it. It matches what 149 agency leaders told Spark AI in our research with The Industry Club: lots of experimentation, pockets of enthusiasm, and no systematic way to turn individual wins into organisational capability.
Why does AI use get stuck in agencies?
Because the learning never gets captured. Someone figures out a brilliant workflow, keeps it to themselves, then moves on, and the insight leaves with them.
Ask yourself: is your team systematically learning what works with AI, or just accumulating anecdotes? If you're not sure, it's probably anecdotes.
The HBR research points to three factors separating organisations making real progress from those stuck in the dip:
Clear accountability. Someone owns AI progress, not as a side project.
Regular sharing rhythms. Teams talk about what's working.
Captured learnings. Insights get documented, not just discussed.
None of this requires a large budget. It requires clear leadership intention, and a sequence. Here's the agency-specific 90-day framework Spark AI uses with clients.
Weeks 1ā3: set the rails
The first block establishes ownership, a goal and three target workflows.
Set a clear goal. Get clear on what AI is for in your agency. Efficiency? New services? Better work? Scale? Write it down in a single sentence and share it with your team. A goal that takes a paragraph to explain won't survive contact with a busy fortnight.
Level the playing field. Make sure everyone has foundational AI fluency. Not a lunch and learn ā proper, agency-specific training that gives people the skills to use AI in a sophisticated way. The bar is building AI assistants that support real work, not redrafting emails.
Assign ownership. Appoint an AI lead: one person with protected time and actual authority. Not a committee. A name on the whiteboard. When everyone is responsible, no one is responsible.
Choose three initiatives. Pick three familiar workflows where AI could make a measurable difference ā perhaps one each for client services, strategy and creative. Map those workflows step by step, identify where an AI assistant could support specific tasks, and define what success looks like: higher quality output, faster turnaround, or greater consistency.
Deliverable by end of week 3: a one-page AI charter with your goal, your AI lead's name, and three initiative briefs with success metrics.
Weeks 4ā6: build your rhythm
The second block turns the plan into a habit and puts three working assistants into live use.
Create a sharing ritual. Establish a recurring slot for learning ā ten minutes in the Monday meeting, a Friday lunch and learn, or a dedicated Slack channel. Keep it brief and specific: a clever prompt technique, a new feature someone's found, an application that saved hours, a client win. The ritual signals that continuous learning is part of how you work.
Build and test. Develop one AI assistant for each of your three workflows, and give the build to someone in the relevant team. If they build it, they own it, and they'll use it. Then get them to share it at the next team meeting. That's how the flywheel starts turning.
Measure what matters. Track the metrics you defined in week 1. Time saved, quality improvements, client feedback. If you said you'd reduce briefing time by 30%, start counting
Deliverable by end of week 6: three working AI assistants in active use, initial metrics captured, and at least one documented "what we learned" from each workflow.
Which AI agents should an agency build first?
You don't need dozens of tools. You need three that get used. Some examples by function:
Client services ā a brief interrogator that reviews incoming briefs against a checklist and generates clarification questions before work starts; a meeting follow-up generator that turns transcripts into structured emails with action points; a client feedback translator that converts vague feedback into specific revisions.
Strategy ā an audience persona assistant trained on your segmentation data, so positioning can be tested without a research sprint; a competitive intelligence assistant that returns relevant insight from competitor audits without the web noise; a brief builder that drafts or refines creative briefs against your templates and brand background.
Creative ā a concept territory generator that produces initial directions for brainstorms based on campaign history; a visual prompt translator that converts written concepts into detailed image prompts; a copy feedback assistant that assesses drafts against the brief and tone guidelines.
Weeks 7ā9: scale what works
The third block converts proven workflows into documentation, client positioning and governance.
Turn experiments into SOPs. Once you've proven something in one area, you've built more than an efficient process ā you've built a template for change. Write up the workflow so a new joiner could follow it on day one.
Package the client story. Start thinking about how you'll talk to clients about AI. Not as a cost-cutting measure, but as a capability that makes the work better. Prepare your team for the inevitable question: how are you using AI? Your client services team should be confident answering it.
Review your governance. Check your AI policies. Review your contracts with clients, freelancers and partners. Can you answer compliance questions with confidence? How are you using client data? Are your AI tools on enterprise-grade plans? A one-page responsible AI statement, ready for your website and your proposals, s now a hygine.
Deliverable by end of week 9: at least one documented SOP ready for wider rollout, draft client-facing AI positioning, and an updated AI policy reviewed by leadership.
Weeks 10ā12: commercialise
The final block changes what you sell and how you price it.
Update proposals. Reframe how you present value. Lead with how you solve the client's business problem rather than "we build websites" or "we're a communications agency". Show your with-AI edge: the additional creative routes explored, the deeper research, the faster iteration.
Tier your value. Consider what AI changes about what you can offer. Does it let you deliver more quickly at lower cost, or explore more routes and do more detailed research? Does it let you sell systems that create outputs rather than the outputs themselves? Outcome-based or subscription pricing is worth weighing for the right work. Our post on why AI breaks the billable hour goes through the four options.
Publish your approach. Put a responsible AI page on your website: how you use AI, what guardrails you have, and why it makes the work better. In our experience, agencies that discuss their AI approach openly build stronger client trust.
Deliverable by end of week 12: updated proposal templates with AI positioning, pricing options reviewed, and a public-facing AI statement published.
The difference between busy and better
The agencies that capitalise on AI over the next couple of years won't be the ones with the biggest budgets. They'll be the teams that deliberately turned experimentation into structure.
We've seen it with production partners like Tuncarp and POD LDN, who are reshaping pricing models and delivery workflows by systematically upskilling their teams. Mindset over money.
There's a people benefit too. The HBR research found that initial sceptics often changed their minds within weeks of using a well-designed system for embedding AI. We see the same thing. Some of the people least interested in AI at the start of a Spark AI programme end up as the power users, exploring everything it can do and changing how they work entirely.
That only happens when there's a real system in place, rather than enthusiasm alone.
Frequently asked questions
Is 90 days realistic for AI implementation in an agency?
Yes, because the plan assumes one named lead with protected time and three workflows, not a whole-agency transformation. Most of the work sits inside existing meetings and existing client projects. What makes it slip is usually the absence of an owner rather than the pace.
What if we've already done AI training and nothing stuck?
That's the normal starting point, and it's what the plan is designed for. Training builds individual fluency. The sharing ritual, the documented SOPs and the named owner are what turn that fluency into something the business keeps when someone leaves.
Do we need an AI lead if we're a small agency?
Yes, and it matters more at small scale rather than less. It doesn't have to be a full-time role ā it has to be one person with a name, protected hours and the authority to make decisions. Committees produce discussion; individuals produce deliverables.
How many AI assistants should we build in the first 90 days?
Three, one each for client services, strategy and creative. Spark AI's experience across agency programmes is that three assistants that get used daily beat a dozen that sit unopened.
Want help running this across your team?
Spark AI's Accelerator takes a team from scattered experiments to structured capability in less than 90 days. Book a call to talk it through.
āWhere can I find more AI insights like this?
Spark Intelligence is Spark AI's fortnightly newsletter for creative leaders, with a practical exercise in every issue.
Spark AI ā AI training, consultancy and agent building for agencies and brands. Your AI performance partner. https://www.wearespark.ai/
