Last year we argued that design would decide which marketing teams got value from AI. Gartner's 2026 CMO Spend Survey found 70% of CMOs call AI leadership a critical goal for this year and only 30% have the capabilities in place to scale it across the team.
Three things will shape the 2027 marketing org. Boards have funded a year of AI and now want to see the return. Customers are sending their own AI agents to research and buy. Gartner expects more than 40% of agentic AI projects to be cancelled by the end of 2027.
The marketing teams that do best next year will be the ones that can prove what their AI returns. Every agent has an owner, every cost is tracked and every result is tested to see what it added.
Our 2026 post set out five roles and one principle: AI makes the weaknesses in how a team is organised more visible. Most of it was right.
BCG's June 2026 survey of 300 CMOs found 96% say AI is transforming their function end to end. 8% of them run campaigns where several AI agents work without human input. 42% still use generative AI only to help people with individual tasks. Most CMOs expect far more from AI than their teams currently get from it.
The post missed one thing. It treated agents as members of the marketing team and said nothing about the agents working for customers.
More customers now find brands in places marketing doesn't control. SparkToro found 68% of US Google searches ended without a click in the first four months of 2026. Checkout.com's June 2026 research found 57% of consumers would let an AI shopping agent switch brands if it found better value.
That creates a job most teams haven't assigned. Someone has to manage what AI tools see and say about the brand: the product data shopping agents read, the sources AI answers cite and the reviews they rely on. The job needs input from SEO, content, PR and ecommerce. Often none of them owns it.
UKTV's content was becoming harder to find in the places audiences now search. Our 4-week GEO diagnostic audited its presence across Google, ChatGPT, YouTube, Wikipedia, TikTok, Reddit and IMDb in days. The team came away with a prioritised action plan and a clear view of where its content was missing.
AI spend is now large enough to need defending. BCG found 43% of CMOs' companies spent more than $15m on AI in marketing this year.
Comviva's 2026 Global CMO Survey found 90% of organisations increased AI marketing investment over the past two years. Only 12% can prove it worked. 67% can't state their total AI cost.
86% of respondents say their leadership teams are demanding stronger proof of ROI. The effectiveness team becomes the one that decides whether an AI investment gets more budget.
In practice that means testing what each agent adds before it's rolled out, with a control group for every agent that contacts customers. Cost per result gets tracked alongside media cost.
Few businesses will hire seven people with these job titles. We describe them as roles because each one needs a named owner. In a smaller team one person might hold three of them.
Five carry over from last year, two with new names. Two are new entrants.
Still frames the problem and writes the brief your own agents work from. In 2027 they also decide how the brand should be described to the AI tools customers use to compare options.
Still owns how data, tools and processes connect. They now also own the first-party data every agent depends on and track what each agent costs to run.
Previously the agent product owner. With several agents live, this person manages them as a set. Each agent gets a clear test for going live, a performance target and a point at which it's switched off.
Previously the performance and effectiveness lead. This person decides whether AI spend grows, based on tests that show what it added.
Still leads adoption and training. Gartner research reported by Marketing Week found 18% of marketing leaders have already cut early-stage roles. This person now redesigns entry-level jobs so juniors learn by reviewing and correcting agent output.
New. Owns how the brand appears in AI answers and in the products shopping agents recommend. That covers product data quality, the sources AI tools cite and the reviews agents use.
New. The EU AI Act's transparency duties have applied since 2 August 2026. They cover deepfakes and some AI-generated text published on matters of public interest. For UK brands with EU audiences, this person owns disclosure, labelling and the human review before anything is published.
The skills growing in value are experiment design, checking agent output, keeping customer data clean and usable across teams, working out what AI costs per result and editorial judgement. The ones declining are manual campaign set-up, single-channel specialism and routine reporting.
The second list is the work junior marketers used to learn on. Teams that keep hiring juniors need new ways to build their judgement and reviewing and correcting agent output is a good place to start.
Start with the agents you already run. Ask four questions of each: does it do one clearly defined job, can you measure what it changes, do you know what each result costs and is a named person accountable for it?
Gartner puts its cancellation forecast down to rising costs, unclear business value and weak risk controls. An agent that fails these questions is likely to run into all three. Switch it off now. Write down the result for each agent and share it with finance.
Then agree how you'll measure any agent before it gets more budget. Ask an AI assistant about your category and see whether your brand appears. Map where your content reaches EU audiences and who reviews it before it's published. Redesign one junior role around reviewing agent output.
Every marketing team can now buy the same models. The teams that do best in 2027 will be the ones that can show what those models returned. Data is the moat.
Want to see how Data Agents can help you do more with less? Book a quick strategy call.