Most marketing teams already use AI. Far fewer have a short list of AI plays they will run, measure and keep or kill. January is the best month to fix that: budgets reset, targets are fresh and the team has headspace for something new. But if you wait until the first week back to decide what to try, you lose the month to set-up.
This article is for marketing leaders, founders and heads of growth who want January’s AI work to produce results rather than experiments nobody reports on. It gives you five playbooks, each with the steps, a prompt or template, what to measure and when to stop. By the end you’ll be able to pick two, prepare them in December and have early numbers by the end of January.
Why should you plan your January AI playbooks now?
Because the money is already moving and the readiness isn’t. Gartner’s 2026 CMO Spend Survey of 401 marketing leaders found that 15.3% of marketing budgets on average now go to AI, yet only 30% say their organisations have mature AI readiness capabilities, as MarTech reported. The same survey, covered by Marketing Dive, found marketing budgets flat at 7.8% of company revenue, so AI spend is being found by cutting something else.
Use is no longer the gap either. HubSpot’s 2026 State of Marketing survey of more than 1,500 marketers found 86.4% of marketing teams use AI in at least a few areas. The gap is turning that use into a repeatable system with an owner and a number attached.
Do this in December, for each playbook you choose:
- Name one owner and one metric.
- Gather the inputs (data exports, access, examples) before the break.
- Write the prompts and templates, and test them once on last month’s data.
- Book a 30-minute review for the last week of January.
Playbook 1: How do you turn customer research into sharper messaging with AI?
Feed AI the words your customers already use, then test the messages it finds against your current ones. Most teams sit on hundreds of reviews, sales call notes, support tickets and survey answers that nobody has read end to end. AI is very good at that reading.
Steps
- Export 6–12 months of customer language: reviews, win/loss notes, call transcripts, support tickets, open survey answers. Remove names and personal data first.
- Run the prompt below on each source, then once more across all the outputs.
- Pick the top three pains and write one headline and one line of body copy for each.
- Test them against your current message on your highest-traffic landing page or a paid campaign.
Prompt
You are analysing customer language for a marketing team. From the text below, list the 10 most frequent problems customers describe, in their own words. For each, give: how often it appears, two verbatim quotes, the outcome they wanted, and any competitor or alternative they mention. Do not summarise beyond what the text says.
Measure: click-through rate or conversion rate of the new message against the old. Stop rule: if no new message beats the control after two weeks at normal traffic, go back to step 3 with the next three pains.
Playbook 2: How do you run an AI creative testing sprint on paid social?
Use your ad platform’s generative tools to make more variants than your team could design by hand, but control the test so you learn something. Meta said on 6 October 2026 that generating video from static assets in Advantage+ creative is now generally available, and that product-image-to-video for catalogue advertisers is in beta. That turns your existing image library into video tests without a shoot.
Steps
- Choose one campaign with steady spend and a clear conversion event.
- Pick your three best-performing static ads from the last quarter.
- Create two AI variants of each (for example, a short video from the static and a new headline drawn from Playbook 1).
- Run them in one ad set against the originals for 14 days. Change nothing else.
- Check every AI variant before it goes live: brand colours, claims, legal lines and anything a regulator would care about.
| Week | What to do | Decision |
|---|---|---|
| 1 | Launch 3 originals + 6 variants | None. Let delivery settle |
| 2 | Review cost per result by variant | Pause anything 30%+ worse than its original |
| 3 | Scale the winners | Roll winning formats into other campaigns |
Measure: cost per result against the originals. Stop rule: if no AI variant beats its original within 14 days, keep the originals and test a different format next month.
Playbook 3: How do you set up a weekly competitor brief that writes itself?
Give an AI assistant a fixed list of sources and a fixed template, and have it produce the same brief every Monday. The value comes from consistency: the same five competitors, the same sections, every week, so changes stand out.
Steps
- List 3–5 competitors and, for each, the pages to watch: pricing, product updates, blog, careers, press.
- Use an assistant with web research and scheduled tasks (ChatGPT, Gemini, Copilot or Claude all offer versions of this) and paste in the template below.
- Send the output to one channel your team already reads, and give one person five minutes on Monday to sense-check it.
Template
Weekly competitor brief, week of [date]. For each competitor: (1) what changed on their pricing or packaging pages, (2) new products or features announced, (3) new content themes, (4) roles they are hiring for, (5) anything we should respond to, with a suggested action. Link every point to its source. If nothing changed, say “No change”.
Measure: the number of actions the brief triggers in a month and the hours it saves against doing it by hand. Stop rule: if it triggers no actions in four weeks, cut the list to the two competitors that matter most or stop it.
Playbook 4: How do you baseline your brand’s visibility in AI search?
Write down the questions your buyers ask, put them to the main AI assistants each month, and record whether you appear. You can’t improve what you haven’t measured, and most brands have never checked.
Google’s own guidance on AI features and your website says there are no additional requirements or special optimisations needed to appear in AI Overviews or AI Mode, and that SEO best practice still applies. So the work is about being the best, clearest answer, not about tricks.
Steps
- Write 20 prompts a buyer would use, across the funnel: “best [category] for [sector]”, “[your brand] vs [competitor]”, “how to solve [problem]”.
- Run each in ChatGPT, Gemini, Perplexity and Google AI Mode. Record in a sheet: mentioned (yes/no), position, which sources were cited, and whether the description of you is accurate.
- For each prompt where a competitor appears and you don’t, find the page they’re cited from and ask what your equivalent page is missing.
- Fix the three biggest gaps in January, then rerun the same 20 prompts in February.
Measure: share of prompts where you’re mentioned, and the accuracy of what’s said. Stop rule: none. This becomes a monthly habit, and it gets more useful each time you run it. If you want to go further, our AI marketing systems work covers this tracking end to end.
Playbook 5: How do you win back lapsed customers with AI-personalised email?
Segment lapsed customers by what they bought and why they might have left, use AI to draft a message for each segment, and hold back a control group so you know what the campaign earned. Lapsed customers already know you, which makes them cheaper to win than new ones.
Check the rules first. In the UK, the ICO’s electronic mail marketing guidance allows a limited “soft opt-in” for existing customers who bought a similar product or service from you, as long as you gave them a simple way to opt out when you collected their details and in every message. It doesn’t cover prospects or bought-in lists. The ICO notes the guidance is under review after the Data (Use and Access) Act, so check the current version, and check local rules outside the UK.
Steps
- Pull customers with no purchase or activity in 6–18 months who can lawfully be emailed.
- Split them into 3–5 segments by product bought or reason for leaving (from Playbook 1 if you ran it).
- For each segment, prompt: “Write a short, plain email to a customer who bought [product] and hasn’t returned in [time]. Lead with what’s changed since they left: [list]. One clear call to action. No discounts unless I add one.”
- Hold back 10–20% of each segment as a control that gets nothing.
- Send, then compare reactivation rates between those who got the email and the control.
For personalisation beyond email, see five ways to use AI to personalise your customer experience.
Measure: reactivation rate (or revenue per contact) against the control. Stop rule: if a segment shows no lift over control after two sends, drop it and rework the offer.
Which playbook should you run first?
Run the one closest to revenue that your data can support today, then add one more. Running all five at once spreads attention too thin to learn anything.
| If this is true | Start with |
|---|---|
| Your messaging feels generic or hasn’t changed in a year | Playbook 1 (customer messaging) |
| Paid social is your biggest channel and creative is the bottleneck | Playbook 2 (creative sprint) |
| You keep getting surprised by competitor moves | Playbook 3 (competitor brief) |
| Organic search traffic is falling or buyers say they “asked ChatGPT” | Playbook 4 (AI search baseline) |
| You have a large customer base with clear churn | Playbook 5 (win-back email) |
Playbooks 1 and 2 pair well, because the messaging work feeds the creative tests. If you haven’t set out how AI fits your wider plan yet, start with how to build an AI marketing strategy.
What does a playbook look like with real numbers?
Here is a hypothetical example, with illustrative numbers, of a B2B software company running Playbooks 1 and 4 in January.
Playbook 1. The team exports 400 support tickets and 60 sales call summaries. The prompt surfaces “we can’t see which campaigns create pipeline” as the most-mentioned problem, appearing in 1 in 5 conversations. They rewrite their demo page headline around it. Over three weeks the page gets 2,000 visits per version: the old headline converts at 2.0% (40 demo requests) and the new one at 2.6% (52). That’s 12 extra demo requests in three weeks. With the team’s own figure of 1 in 4 demos becoming an opportunity, that’s 3 extra opportunities, for about a day of work.
Playbook 4. Of 20 buyer prompts across four assistants (80 checks), the brand is mentioned in 9, or 11%. A competitor is mentioned in 31 (39%), mostly cited from a comparison page and a pricing explainer. The team publishes both pages in January. In February they rerun the checks and are mentioned in 17 of 80 (21%). They now have a number to report monthly and a list of the next pages to write.
The point isn’t the numbers. It’s that each playbook ends with a before, an after and a decision.
How do you prove the playbooks worked by February?
Report three numbers for each playbook: the business metric against its control, the hours saved or spent, and the decision (keep, change or stop). Keep it to one slide.
| Playbook | Metric vs control | Hours saved per month | Decision |
|---|---|---|---|
| Customer messaging | +0.6 pts conversion | 6 | Keep, extend to paid |
| AI search baseline | 11% to 21% mentions | -4 (new work) | Keep, monthly |
Two rules make the report credible. First, every result is measured against a control or a baseline taken before you started, not against a hunch. Second, count the time the playbook costs as well as the time it saves, including the human review. If you want a quick read on where your team sits before you start, take the AI Maturity Assessment; it takes a few minutes and shows where the gaps are.
Summary
- Pick two playbooks, not five: one owner and one metric each, chosen in December.
- Start with your customers’ words: AI reading your reviews and call notes is the cheapest way to sharper messaging.
- Control your creative tests: use generative ad tools for volume, but change one thing and set a 14-day stop rule.
- Make competitor tracking a fixed weekly brief: same sources, same template, every Monday.
- Baseline your AI search visibility now: 20 buyer prompts, four assistants, rerun monthly.
- Win back lapsed customers lawfully: check soft opt-in rules, segment, and always hold back a control.
- Report three numbers in February: result against control, hours saved, and keep, change or stop.
If you’d like these playbooks built and running for your team, talk to us at SK.