One keen early adopter doesn't change a team. A champion with a lane, a handover and a manager who protects the time can.
Pick champions with five yes or no signals, give each one lane and a recurring deliverable, and turn their habit into a recipe a second person can run. Put the hours in writing, hold a thirty-minute weekly share-out of real work, and measure use, persistence and impact on one named deliverable per lane.
Most marketing teams already have one person who is good at AI. She has a few prompts that work, a folder of results she’s proud of, and a reputation as the one to ask. What the team doesn’t have is a way for that skill to travel, so it stays with her, and everyone else’s week looks the way it did last quarter.
A lifecycle marketer builds a prompt that rewrites a winback email around the reason a customer lapsed. It works, and a colleague asks for it in the team chat. She pastes it in. Two weeks later there are three slightly different copies, the paid team hasn’t heard of it, and nobody can say which version the good results came from. When she’s on leave, the prompt goes quiet with her.
That’s what depending on one person’s initiative looks like. Nothing in the team turns a personal habit into something the next person can pick up and run, however hard the individual works. A champion, in the sense I mean here, is the person given the job of making that handoff happen, inside one part of the marketing work, with the time and standing to do it.
This piece sits next to three others on the site, and it’s worth being clear about what each one covers:
An early adopter and a champion can look the same in week one. Both are keen, both have tried most of the tools, both are happy to demo. The difference shows up after the demo: whether anyone else changes how they work.
In marketing that usually comes down to craft credibility. Marketers adopt a method when they’ve watched someone whose work they respect do it on a brief like their own. The person who tries every new tool sometimes has the least of it, because colleagues have learned to wait and see which tool survives the month.
So I’d stop asking for volunteers and look for five signals, which you can answer from what you already know about your team.
| Signal | What to look for in a marketing team | A warning sign |
|---|---|---|
| 1. Peers ask how | When their work comes up in a review, colleagues ask how they did it, and they can name something a teammate changed as a result. | People admire the output and move on. |
| 2. Owns a recurring deliverable in the lane | The weekly newsletter, the monthly campaign report, the paid creative refresh. Something that comes round again, so a new method has somewhere to live. | Their work is mostly one-off projects. |
| 3. Shows the failures too | They’ll say where the tool got it wrong, including the draft that went out with a claim nobody had checked. | Every demo is a highlight reel. |
| 4. Respects the brand and claims line | Their AI-assisted work comes back from brand or legal review needing about the same fixes as anyone’s, not more. | They treat review as a speed bump. |
| 5. Their manager will protect the hours | A named block each week, agreed with the person they report to. | The time would come from evenings. |
A Future Factors scoring aid based on our own judgment, not a validated instrument. Four or five yes answers: ask them. Three: ask them, and pair them with a teammate who already has peer trust. Two or fewer: keep them as a power user and ask again in a quarter.
Signal four is the one that’s specific to marketing, and it’s often skipped. Marketing work goes out under the company’s name. A champion who is fast but loose with claims teaches the rest of the team to be fast but loose, and the habit spreads just as readily as a good one would.
The last signal sounds administrative, yet it decides a lot. If the hours come from evenings, you’ve found someone willing to do unpaid work for a few months, not a champion.
The usual instruction is “be our AI champion,” which gives a person the whole of AI to be responsible for. Nobody can hold that. What works better is a lane: one part of the marketing work, with its own recurring deliverables, its own risks and its own people to teach.
The lanes differ more than they look from outside. A content marketer wants a brief to become a draft without losing the point of view, a paid buyer wants variants that can be tested honestly, and a lifecycle marketer is thinking about segments and consent. A champion from one lane can’t credibly teach another, so a single general champion for a mixed team tends to reach only people who think like them.
| Lane | What the champion shows | What the champion holds | The usual trap |
|---|---|---|---|
| Content | Turning a brief and source material into a first draft, and repurposing one piece into several formats | The edit pass: facts checked against sources, the point of view kept | Volume with no edit, so more drafts and the same quality |
| Paid media | Generating ad and search variants, reading search terms, drafting creative briefs | The testing rules, the platform policy limits and every budget decision | Variants nobody tests properly |
| Social | Calendar drafts, caption options, sorting comments into reply, escalate or ignore | Voice, and the list of topics the team doesn’t comment on | Chasing every trend because drafting got fast |
| Brand | A voice guide the AI can actually be given, and a review checklist for AI-assisted work | The guardrails: claims, disclosure, what needs legal review | Becoming the approver of everything |
| Lifecycle and email | Subject line and body variants by segment, flow audits, winback rewrites | Consent rules, send logic and what customer data may go into a tool | Personalising with data the tool shouldn’t have |
| Marketing ops | Reporting summaries, campaign naming and tracking checks, cleaning exports | Data access, the approved tool list, the numbers going into any report | Turning into the team’s AI help desk |
A Future Factors working view based on our own judgment. In a small team one person will hold two lanes, and that’s fine. A lane without a recurring deliverable probably doesn’t need its own champion yet.
Give every champion one lane of the work, not the whole of AI.
The brand lane is the one other departments don’t have an exact equivalent for. If the brand champion reviewed everyone’s AI-assisted work, they’d become the bottleneck and the team would route around them. Their job is to write the guardrails once, in a form the other champions can apply to their own lane, and to take the grey cases.
A champion holds the standard for their lane. They don’t sign off everyone’s work.
A small team can start with fewer than six lanes: content, lifecycle and ops, say, borrowing guardrails from whoever already owns the voice guide.
Lanes tell you who. The harder question is what moves from the champion to everyone else. I think of it as three levels, and most teams stall between the first and the second.
A Future Factors working model, based on our own judgment and not a research-validated framework. It describes where one piece of work stands, not how mature your whole team is.
A recipe is real when a teammate who wasn’t there gets a usable result from it. A standard is real when it shows up in a template or checklist the team already uses. Here’s each level in three lanes.
| Lane | Personal habit | Shared recipe | Team standard |
|---|---|---|---|
| Content | Pastes interview notes into a chat and asks for a draft in her own style | A one-page recipe: which notes to include, the prompt, and the three things to check before it goes to edit | The brief template has a line for source material, and the editor’s checklist includes the three checks |
| Lifecycle | Rewrites a winback email around the reason the customer lapsed | The prompt, the lapse reasons it handles and the data it must not be given | The flow-review checklist includes a consent and data check |
| Marketing ops | Asks for a summary of the weekly performance export | The prompt plus a step to confirm the totals against the source before sharing | The weekly report template carries a ‘totals checked by’ line |
Examples of the shape, not a prescribed set. The level is about whether other people can run the work, not about how clever the prompt is.
Moving a habit to level two is where a champion earns the title. Writing a recipe forces the question the habit didn’t have to answer: what does this person know that makes it work? Often it’s context, such as which audience, which tone, which data to leave out, and that’s what the next person is missing. A prompt can help pull it out of the champion’s head and onto a page.
| I’m a [lane] champion on a marketing team and I want to hand over a way of working. Here is what I do today, in my own words: [paste your rough steps and the prompt you actually use]. Ask me up to five questions about anything a colleague would need to know to get the same result, such as audience, tone, data to include or leave out, and where it usually goes wrong. Then write a one-page recipe with these headings: the task and when to use it, what to gather first, the prompt, what to check before the output is used, what it must not be given, who owns it. Do not add steps I haven’t described, and mark anything you’re unsure of as [CHECK]. |
Only paste material you’re allowed to put into your approved tool. The result is a draft that a second person should test on their own work before it counts as a recipe.
The step from recipe to standard is mostly workflow design. In McKinsey’s 2026 State of AI survey, nearly three-quarters of the respondents it classed as high performers said they were fundamentally redesigning workflows because of AI, against about a quarter of everyone else.[1] The survey spans functions, not just marketing, and the link isn’t proof of cause. What I take from it is that organisations getting value tend to change the work itself. For a marketing team, that means the brief, the review checklist and the report template, which is where a level-three standard lives.
A habit isn’t adopted until someone else can run it without the person who invented it.
To keep recipes somewhere other than a chat thread, our guide to building an AI prompt library for your team covers where to keep them and how to stop the library going stale.
A champion can show, write down and answer questions. Two things sit outside that, and teams tend to expect both.
The first is capability across the whole team. Showing a method in a thirty-minute session isn’t training, though a team can certainly start with its own champions. Structured, expert-led training on the team’s own marketing work shortens the learning curve and gives everyone the same baseline, so the champion isn’t the only teacher. Training builds the capability. The champion’s lane and the recipes give it somewhere to be used, which is workflow design. Reinforcement is what turns it into behaviour, and that’s the second thing a champion can’t do alone.
A champion is a peer, so they can’t set expectations for the people they sit next to. That’s the manager’s job. Microsoft’s 2026 Work Trend Index surveyed 20,000 AI users and found that the people it classed as its most advanced users were more likely to say their manager sets quality standards for AI work: 83% compared with 57% of everyone else.[2] It’s a self-reported comparison, so it shows what goes together, not what causes what, but it fits what I’d expect: people use AI more carefully when someone has said what good looks like.
In practice the manager’s part is small and specific:
One more line is worth drawing before the programme starts: what the champion’s recipes do, and what a person still owns. Writing it down protects the champion from being the person blamed when an AI-assisted piece goes wrong.
| Work in the lane | What AI does | What the human still owns | How it gets checked |
|---|---|---|---|
| Content: a thought-leadership draft | Produces a first draft from the brief and the source notes | The argument, every factual claim and the byline | Editor checks claims against the sources, every piece |
| Paid: a creative refresh | Proposes headline and description variants tied to a stated hypothesis | Which variants run, the budget and any claim about the product | The buyer reviews against platform policy and the claims list before launch |
| Ops: the weekly performance summary | Drafts the narrative from the export | The numbers, and the call on what the movement means | Totals confirmed against the source, named on the report |
Our suggested split for champion-run recipes. The checks in the last column are what turn a recipe from a personal habit into a team standard.
The ask differs by seat. A head of marketing protects the time and asks for examples. A team lead makes the recipes fit the weekly cycle. A brand or legal reviewer wants the guardrails in writing before AI-assisted volume goes up.
Champion programmes that fade are usually missing a calendar, not a plan. The thirty days below are deliberately small: one champion per lane, one recurring task each, one shared session a week.
Pick champions with the scorecard. Agree the hours with each champion’s manager and write them down. Name one person, usually the brand champion, for the grey cases.
Each champion picks one recurring deliverable in their lane and records how they do it today. No tool changes yet; the aim is a plain description of the current work.
Each champion runs the task with AI on a real piece of work, then writes a one-page recipe using the prompt above. The brand champion drafts the lane guardrails.
First share-out. Each champion presents one piece of real work, including what went wrong.
A second person in each lane runs the recipe on their own work. The champion fixes the recipe wherever the second person got stuck.
Review in 30 minutes: keep, change or drop each recipe, and decide which one earns a place in a brief or checklist.
Our suggested starting plan. Treat the dates as a rhythm to adjust, not a requirement.
The weekly session is where recipes get tested in front of people. A content lead might bring the newsletter draft that took too many rounds of edits, and a lifecycle champion might show how her winback recipe handled similar copy. People start borrowing across lanes without being asked. Keep it to thirty minutes in the same shape each week.
| Field | Example for a Thursday session |
|---|---|
| Length and slot | Thirty minutes, same time each week, for the first eight weeks. |
| First ten minutes: one champion shows real work | The content champion walks through this week’s newsletter: the notes, the draft, the edit and the two claims that needed fixing. |
| Next ten: bring something stuck | Anyone brings one piece of work where AI didn’t help. The room tries a fix live. |
| Last ten: decide what goes in the recipe list | Keep, change or drop. A recipe needs a named owner and a next review date. |
| What gets written down | The task, the recipe used, the checks done, what went wrong, and who runs it next week. About five lines. |
The two fields people leave blank are what went wrong and who runs it next week. Without them, a share-out is a demo.
Teams most often trip by skipping days six to ten, because describing today’s work feels like a delay. It’s the baseline you’ll want later. The other slip is a share-out that drifts into showing tools. When the examples stop coming from real work, pause and find out why.
The obvious thing to count in a champion programme is activity: sessions held, prompts shared, people who attended. Those numbers tell you the programme is running. They don’t tell you whether the team works differently. I’d use the same lens we use for any adoption effort: use, then persistence, then impact.
| Stage | Signal to track | What healthy looks like | Warning sign |
|---|---|---|---|
| Use | People outside the champion group who ran a recipe on their own work | Grows each month, across more than one lane | Only the champions and their friends run anything |
| Persistence | Recipes still being run a month or two later without a reminder | Recipes get edited by people other than their author | Recipes were written, shared once and left untouched |
| Impact | Revision rounds or time to first usable draft on one named deliverable | Down for the deliverables with recipes | No change, or faster output with more brand fixes |
| Impact | Items in the brief template, report template or review checklist that came from a recipe | A few by day 30, more each quarter | Recipes exist but the templates are unchanged |
A Future Factors measurement lens based on our own judgment. Measure one named deliverable per lane before you start, or you only have impressions. Pick measures that fit the lane, such as revision rounds for content or send errors caught in review for lifecycle.
A fast first draft that comes back from brand review with more fixes isn’t an improvement, and a champion who says so is more useful than one who reports speed alone. For a longer treatment of what to count and what to ignore, our guide to AI usage metrics goes through it.
I’d start with the smallest version of this on Monday. Choose one lane and one champion, using the five signals. Get their manager to protect two or three hours a week in writing. Have them describe one recurring deliverable as it is today. In a month, you’ll know whether a second person can run what they wrote down, and that’s most of what you need to decide what to do next.
One per lane that has a recurring deliverable, which in a small team usually means two or three people covering more than one lane each. Start narrower than feels right. Two champions who’ve each handed a recipe to a teammate beat six who attended a kickoff.
Keep them as a power user and pair them with someone colleagues already trust. The power user finds the methods, and the trusted colleague carries them to the rest. The second person needs credibility more than tool skill.
The head of marketing is better placed as the sponsor who protects hours and asks to see the work. The brand lead is a natural fit for the guardrail lane, so long as they write the guardrails once and consult on grey cases, not review everything. A senior leader as a lane champion is harder, because people below them tend to copy to please.
They do different jobs. Champions give peers a credible example in their own lane and keep the practice going. Expert-led training builds a shared baseline across the team and shortens the curve, because people practise on their own briefs. A team can start with champions alone, and adding training should help each champion’s recipes land faster.
Put the hours in their objectives, name them as the owner of the recipe in the templates it ends up in, and let their manager say publicly that this is part of the job. Recognition that adds work, such as more sessions, tends to backfire. Our champions network guide covers how a written role definition protects the person.
Sources were read on October 8, 2026: the McKinsey State of AI survey page (published August 25, 2026) and Microsoft’s 2026 Work Trend Index annual report (published May 5, 2026). Both figures are cross-function comparisons between groups, so they show what goes together, not what causes what, and neither is specific to marketing teams. I tried to read the Cochrane review of local opinion leaders for the peer-influence evidence but the page wouldn’t load, so I haven’t quoted it. The champion lanes, the three levels, the selection scorecard, the 30-day plan and the share-out template are Future Factors working tools based on our own judgment, not research-validated models. The scenes are illustrations, not client stories.
Weekly tips. Real examples. Practical help for busy professionals.
We care about your data, check out our privacy policy.