§ Playbook · Marketing
The AI playbook for Marketing
Marketing has the highest AI adoption of any function and some of the least to show for it. This playbook covers what genuinely works, which roles change, what the stack costs you, and how to get value in 90 days without shipping slop.
Free · no form, no follow-up sequence
Nearly nine in ten marketers now use AI every week. MIT found 95% of enterprise AI pilots produce no measurable P&L impact. Both numbers are true, and the gap between them is your actual to-do list.
95%
of enterprise GenAI pilots deliver no measurable P&L impact.
MIT State of AI in Business, 2025
49%
of licensed martech capability actually gets used.
Gartner Marketing Technology Survey, 2025
6wk → 7d
Klarna's image production cycle, saving $10m a year.
Klarna / Forbes, 2024
§ Part one
What actually changes.
Start with the honest observation: AI is not very good at your strategy, and it is extremely good at your production line. The evidence from the brands doing this at scale is remarkably consistent. Klarna cut its image cycle from six weeks to seven days and took $10m a year out of production and agency spend. Unilever's AI studios now turn out around 400 assets per product where a campaign used to produce 20.
None of that is brand thinking. All of it used to be paid for as if it were. The work that moves first is volume work: versioning, adaptation, translation, first drafts, resizing, retouching. The work that moves last - or not at all - is judgement: what the brand stands for, which audience matters, what not to say.
01
Asset production becomes a factory
Versioning, adaptation and localisation across formats and markets is now machine work with human sign-off. This is where Klarna and Unilever found their money, and it is where you will find yours. It is unglamorous, which is why it works.
02
First drafts stop being a job
Copy, briefs, post variants, email flows - the blank page is gone. What remains, and grows, is the edit: someone senior enough to know which of the forty drafts is wrong in a way that matters.
03
Personalisation gets real, slowly
McKinsey puts GenAI personalisation at 5-8% revenue lift where it works. Where it works is the qualifier doing the heavy lifting: it needs usable customer data, which most marketing departments discover they do not quite have.
04
Media spend gets an audit it won't enjoy
Predictive models are cutting wasted media spend by 15-20% in the consultancy tellings. Directionally right. The bigger shift is that AI makes your media agency's margin visible, which changes that conversation permanently.
05
The agency relationship inverts
WPP now sells its AI platform directly to brands. Publicis put €300m into CoreAI. Your agencies are industrialising the same production work you are - so you are now negotiating with someone whose costs just fell and whose headcount is following.
06
Strategy stays stubbornly human
AI cannot tell you what the brand should mean, who it is for, or which trade-off to make when growth and margin argue. Anyone selling you autonomous strategy is selling you a random number generator with good manners.
§ Part two - the honest bit
What it costs you.
This is the part most playbooks skip, so let us not. The production win is a people change. WPP shed roughly 7,000 people in 2025 and its CFO said plainly that the savings come from people. Around 23% of agencies cut junior copywriting roles in the same year. If your plan assumes nobody's job changes, you do not have a plan - you have a communications problem on a delay.
There is also a money conversation. Gartner found marketing teams use 49% of the martech capability they license. AI does not fix that; it makes it worse, because the tools you underuse are now also the tools AI quietly replaces.
01
Junior production roles shrink
The honest version: fewer people making more assets. The entry-level rungs - junior copy, production design, versioning - are where the reduction lands. If you care about your talent pipeline, you now have to design one deliberately, because the old apprenticeship just automated itself.
02
The editor becomes the scarce role
The person who can tell good from plausible is now your production bottleneck and your quality control. Senior creatives who can direct machines do not appear by magic; they were the mid-level people you were about to stop hiring.
03
Your martech stack owes you money
Half your licensed capability is unused and AI is eating the point tools from both ends. Every renewal is now a genuine decision. Expect to retire tools you fought hard to buy, and expect the people who championed them to mind.
04
Agency contracts need rewriting
Paying campaign rates for asset production that costs your agency a fraction of what it did is not a partnership, it is a subsidy. Retainers, scopes and rate cards all need reopening - kindly, but this year.
05
Brand safety gets a new owner
Coca-Cola shipped AI Christmas ads two years running and got mauled twice; 'AI slop' was 2025's word of the year. A third of consumers say visible AI use damages their view of a brand. Someone senior now owns the question 'should a machine have made this?' - and it cannot be the intern running the tool.
06
Some sunk costs are just sunk
The DAM migration, the CMS rebuild, the personalisation engine that never quite launched - some of it will be superseded before it pays back. Grieve briefly, write it down honestly, and stop funding things because they were expensive.
§ Part three
What good looks like.
A good marketing function on the other side of this is smaller in the middle, more senior at the edges, and much clearer about what humans are for. Production runs like a factory with taste as the quality gate. Strategy, brand and judgement get more human time, not less - that was meant to be the point.
01
A production engine with a taste gate
High-volume asset work runs through AI with clear brand guardrails, approval steps, and a named human who owns quality. Output goes up, cost per asset falls, and nothing ships that a customer could mistake for slop.
02
A smaller, sharper stack
Fewer tools, properly used, with AI capability bought where it changes work rather than where the demo was nice. Utilisation is a KPI. Renewals are decisions, not habits.
03
Measurement that would survive a CFO
Only 41% of marketers can point to AI ROI. Good looks like the other kind: each AI use case has an owner, a baseline, and a number that moved. If it cannot show one, it stops.
04
A deliberate talent pipeline
You still hire juniors - but into an apprenticeship you designed, learning judgement on real work, not surviving on production tasks that no longer exist. The team knows which decisions stay human, and why.
The change, across all five pillars
People
Production shrinks, editors rise - and the junior pathway gets rebuilt on purpose rather than quietly starved.
Processes
Asset work runs as a measured production line with a taste gate, not a pilot with a demo day.
Data
Customer data made fit for personalisation before you buy the tools that assume it already is.
Platforms
A smaller martech stack where utilisation decides renewals and nothing survives on sunk cost.
Partners
Agency retainers, scopes and rate cards repriced to what production actually costs now.
§ Part four
The first 90 days.
Not a transformation programme. One production job shipped properly, measured honestly, and used to earn the right to the next one.
Ⅰ
Weeks 1-2
Pick one production job
Choose the highest-volume, lowest-judgement work you have - versioning, localisation, resizing. Baseline its cost and cycle time now, while nobody is defensive about the numbers.
Ⅱ
Weeks 3-6
Ship it, with a taste gate
Stand the workflow up in production with brand guardrails and a named senior owner for quality. Real assets, real campaigns, approval on. Resist the urge to pilot forever.
Ⅲ
Weeks 7-10
Measure, then defund something
Compare against the baseline in money and days. Then act on it: retire a tool, reopen an agency scope, redeploy the hours. A win that changes no budget line is a demo.
Ⅳ
Weeks 11-13
Widen, and have the people conversation
Take the second and third jobs. Tell the team honestly which roles change and what the new senior-editor path looks like. They have read the same headlines you have; silence is not kindness.
The day-90 scorecard · what you should be able to show
§ Part five
The traps, so you can skip them.
Trap 01
Pilot theatre
Eighteen months of experiments, a lovely showcase, nothing in production. MIT's 95% were not idle - they were busy. Busy is not the same as live.
Trap 02
Shipping slop
The cost saving is visible in your budget; the brand damage is visible in everyone else's feed. If customers can tell a machine made it and you did not want them to, you have already lost the trade.
Trap 03
Buying before choosing
Adding AI tools to a stack you use half of is how the 49% number happens. Choose the job first. The tool is the last decision, not the first.
Trap 04
Letting the interns run it
Handing AI to the most junior people because they are 'good with technology' puts your least experienced judgement at the point of maximum leverage. Backwards. Senior people direct the machines.
Trap 05
Pretending nobody's job changes
Your team can read. Klarna, WPP and the agency layoffs were front-page news. An honest account of what shrinks, what grows and what you will retrain beats a reassuring one that expires.
Trap 06
Keeping every contract
If your agency's production costs fell 27% and your fees did not, the saving happened - it just went to someone else. Renegotiate like you mean it.
§ Before you go
Six questions worth an honest hour.
Score yourself before someone with a rate card does. If more than two answers are uncomfortable, that is not a crisis - it is an agenda.
The self-assessment
§ Start with the workshop
Book your
Catalyst Workshop.
Bring the strategic change you have been circling. In 1-2 executive days, we will turn it into the first move, the case for it, and the mobilisation plan.