Jolene is the marketing executive at a 35-person furniture retailer. She asked an AI tool for a LinkedIn post about the year-end sale and got ten drafts in under a minute.
Then the problems appeared. One draft promised 30% off when the sale was 20%. Another included a customer quote nobody had said. None of them sounded like the brand. Her manager spent longer fixing the drafts than writing one from scratch.
The problem wasn't the AI. It was the brief. Here's a simple AI marketing workflow that starts in the right place.
I. Start with one approved brief
Before anyone drafts anything, get the campaign owner and the person who approves the offer to agree on the basics:
| Brief field | What to record |
|---|---|
| Audience | Who you're talking to, and why the offer matters to them now |
| Objective | The one action you want people to take |
| Offer | Approved product details, price, dates and conditions |
| Evidence | Sources for any facts, plus approved customer quotes |
| Voice | A few on-brand examples, preferred words and words to avoid |
| Formats | The channels you need and their specifications |
| Approver | The person who signs off the final version |
Keep guesses separate from facts. If the price or launch date isn't confirmed, leave a gap. Don't let AI fill it in.
II. Ask for real alternatives, not ten versions
Start with one core message and two genuinely different angles. For example, one draft explains the problem the product solves, and another shows how it works. Keep the facts identical, so the team is only comparing the approach.
Here's a prompt Jolene could use:
"Using only the approved brief, draft two LinkedIn post options for the stated audience. Keep the offer and factual claims unchanged. Follow the voice examples. List any missing information separately. Don't invent customer quotes, performance figures, deadlines or guarantees. Label both as drafts for review."
AI can sound confident and still get details wrong, so every draft still needs checking.
III. Review before you adapt
Check the core message before turning it into a carousel, email or landing page. Otherwise, one mistake spreads across every format.
A simple review checklist:
Does it speak to the right audience and objective?
Can every claim be traced back to the brief?
Are the offer, dates and conditions correct?
Does it sound like the brand?
Are the images approved for this use?
Is the next step clear, with the right link?
When you reject a draft, say why. "Adds a claim that isn't in the brief" is something the team can fix. "Doesn't sound right" isn't.
IV. Use templates for the predictable parts
Not every step needs AI. Layouts and formatting work well as reusable templates, while AI helps with the words.
Once Jolene's sale announcement is approved, she turns it into a LinkedIn post, a six-slide carousel outline and a short email. The facts stay fixed. Only the format and explanation change for each channel.
V. Measure the whole process
Track the full journey, not just drafting: briefing time, drafting time, review time, rounds of revisions and how many pieces get approved.
If drafting gets faster but reviews take longer, fix the brief and the prompt before producing more. Once the campaign runs, measure clicks and enquiries separately. Faster production doesn't mean a more effective campaign.
VI. Make it a team habit
Keep the brief, prompt, checklist and final assets together. Ask a colleague to repeat the process on the next campaign and note where they get stuck. That's how one person's experiment becomes the team's way of working.
Comparing training options for your team? Our corporate AI training checklist can help.
If you'd like your team to build a workflow like this with guidance, explore our AI for Marketing Teams programme or book a discovery call.
Jolene is an illustrative example based on a typical SME.