Most marketing teams already have AI. Few have changed how they work.
Some of the companies we work with have rolled out Copilot or Claude Enterprise. Licenses are handed out, training is held and usage rates go to the leadership team every month. Then you ask which marketing process has changed, and it gets quiet. Emails get written a bit faster. Everything else runs as before.
Buying a tool is easy. Changing how a team works is much harder.
This is a practical AI marketing strategy for marketing leaders who want the second one done within a quarter. It’s based on what we’ve seen work with teams like Attendo, Sanoma and Eficode, and in our own marketing.
1. Name the person who owns the change
Who in your company is responsible for changing how marketing works with AI?
In most companies we meet, IT owns the tool and HR owns the training. Marketing is too busy shipping the next campaign to change its own processes. So nobody does it.
The fix is simple to say. One named person owns the change in marketing, with time in their calendar for it and a target to report on.
Attendo’s Finnish marketing and communications team started with a clear vision: every team member would start using AI for at least one task. That’s a good first target. It’s specific, everyone can check it, and it includes the whole team. Read the Attendo case.
2. Fix the basics before you automate anything
A piano lesson starts with middle C and rhythm. Chopin comes much later. At cooking school, you first learn to chop onions, fast and evenly, over and over.
The reason is the same in both. Once the basic move happens without thinking, your head is free for what actually matters.
In marketing, the basics are pretty boring:
- You know who you sell to, down to account names.
- You can say in one sentence why customers choose you.
- Your customer data is clean.
- You measure sales, not clicks.
These are exactly the things that get skipped, because AI agents are so much more interesting. I get it. I’d also rather build agents than write a target group definition.
But without the basics, AI mostly speeds up the mess. Content written for a vague target group stays vague. There’s just more of it.
Clean the customer data with AI before you automate anything that uses it. Every later project depends on it.
Sini Korpinen, Chief CX and Communications Officer at Attendo Finland, spent her first year there fixing workflows and defining roles. Processes came first, automation later. Read how she got her entire team using AI.
3. Map the use cases, then pick a few
Most teams can list dozens of tasks where AI could help. The hard question isn’t how to use AI in marketing in general. It’s which few tasks to start with.
Attendo’s team did it in three phases. A 90-minute workshop with live demos built around their own work came first. Then came interviews across the marketing and communications team, which turned up more than 50 potential AI use cases in content creation, social media, internal communications and analytics. Each one was scored on impact and feasibility. The top four became concrete implementation plans, and the potential added up to more than 9,000 hours a year.
Sanoma ran a similar mapping at a larger scale: more than 20 interviews across seven teams and over 100 content creators. Some teams were already 10x faster with AI on tasks like versioning, translations and routine edits. The Sanoma case shows how that became a two-year roadmap.
Eficode’s assessment ended in an à la carte menu of opportunities, each with an estimated cost and impact, so the team could choose and schedule the next steps. Many options didn’t need any significant technology investment. See the Eficode case.
What makes a good first use case? Look for tasks with these traits:
- Clear rules. You can describe the task as “first X, then Y, then Z.”
- Repetition. The same pattern happens dozens or hundreds of times.
- Volume. The more material there is to process, the bigger the gain.
- Some reading or analysis. AI understands content in a way spreadsheet formulas can’t.
- Boredom. If the task makes someone daydream about other work, it’s a good candidate.
A good example is rating 15,000 keywords for relevance to one company: same process, clear result, huge volume. A poor example is setting a client’s strategic direction, because it needs interpretation and creativity.
Not sure where your team stands overall? Our AI maturity model for marketing describes five stages and what it takes to move between them.
4. Decide where the freed time goes
AI productivity gains are real. They also disappear easily.
You roll out AI and get a clear efficiency gain. Six months later, nothing has changed. People have simply found new ways to fill the time.
There’s a name for this: Parkinson’s law. Work expands to fill the time available for it.
If AI frees up half of a marketing team’s capacity and nobody decides where that capacity goes, it fills itself. More meetings. More reports. More of something.
So decide before you start. Where does the freed time go?
- Producing the same output at lower cost?
- Getting to market faster?
- Testing new markets?
- Raising quality?
Sanoma’s teams tested every decision against one question: will this help us get “five days of work done in four days”? That’s the kind of clear target that makes the gain visible.
Deciding this up front also makes the gain easier to measure and the business case much easier to write.
5. Ship the first use case within a month of picking it
Planning feels productive, but the team learns the most from the first working system.
Singa’s inbound lead automation went live in week 3, after two weeks of mapping the process and setting up the technical base. Leads that used to wait for a morning sorting session now reach the right rep within minutes. Read the Singa case.
Good first projects for a marketing team:
- Visual production. Olo-apteekki’s campaign banners took 20-30 minutes per set. A custom app now makes every size with one click. See the Olo-apteekki case.
- Content versioning and translation. The tasks where Sanoma’s teams saw 10x gains.
- Lead handling. Enrichment, scoring and routing, like Singa.
- Brand-consistent assets. Keep your brand book in a format AI can read, and a skill can produce on-brand banner variations in minutes.
- Weekly reporting. The pipeline and campaign report that takes an hour every Monday.
Build feedback in from day one. At Singa, anyone who spots a wrong decision flags it in a shared Slack channel, and the prompts get adjusted weekly.
Every first version needs a person at the approval step. That’s where the team learns to trust the system, and where it catches the mistakes.
6. Keep the human parts human
I recently made a 15-second brand animation for us with Claude. The first version was ready in about 40 minutes, and I didn’t open a single animation program.
That first version was fine. And fine is basically free now, because anyone can get it in 40 minutes.
It only got good when someone looked at it and said what was wrong. Too fast. Wrong mood. That part looks like an ad.
When code and content get cheap, what’s left is taste, and taste is hard to copy. In marketing, I’d describe taste as the ability to say “this is wrong” before you can explain why. It comes from seeing a lot of good work and even more bad work.
For experienced marketers this is good news. You have exactly what the machine lacks. For juniors it’s harder. How do you learn taste if the machine does all the practice work? I don’t have a good answer to that yet.
Every marketing team should also write its own AI writing guide, for internal and external use. Clay has published a good example. Two points from it are worth copying straight away:
- You have to be able to stand behind every sentence, whoever wrote it, AI or human.
- The reader’s time is worth more than the writer’s. If you made a document in a minute and it takes the reader half an hour, you’ve just moved your work onto someone else.
For marketing I’d add one more: read the text out loud before you publish it. If it doesn’t sound like the person whose name is on it, keep editing.
7. Train people on their own work
Training works best when it’s built around the team’s own tasks. Attendo’s 90-minute workshop used live demos designed for their use cases. As one of their digital marketing specialists put it, “We saw actual tools doing our everyday work.”
Sanoma’s roadmap includes a two-year AI training program for all 100+ creators, alongside governance, a dedicated speedboat team for image, audio and video production, process automation and change management. Governance came first: AI guidelines and licenses had to be in place before the other streams could move.
At Olo-apteekki, the commercial team got hands-on Claude training while we built their visual production app. The team started building Claude Projects for its own daily tasks almost immediately and ran into the usage limits during the first week.
The best skills and automations we’ve seen were built by the person who does the work. Training should get people to that point.
Your AI marketing strategy as a 13-week plan
Put together, a quarter looks like this:
- Weeks 1-2: Name the owner. Write down the four basics. Measure how many hours the team spends on its ten most repeated tasks.
- Weeks 3-5: Interview the team and map the use cases. Score them on impact and feasibility.
- Week 6: Pick two or three use cases. Decide where the freed time goes and write the business case.
- Weeks 6-9: Build and ship the first one, with a person at the approval step and a channel for feedback.
- Weeks 6-13: Train people on their own tasks.
- Weeks 10-13: Ship the second use case. Measure against the baseline from week 2 and decide what comes next.
None of this needs a big budget. It needs an owner, a baseline and the discipline to stop doing the old version of the work.
Our AI strategy work follows the same order, from mapping use cases to getting the first projects into production. For the tools themselves, see AI marketing tools worth your budget.
The teams ahead of you next year will be the ones that changed one process this quarter.
