Who owns your AI spend? Martech Futurist | August 31, 2026
If your AI spend hasn't moved EBIT, start by asking who owns it. In most marketing organizations, nobody does.
Four research teams published inside 72 hours last week, working from different samples, different methods, and different agendas. They converged anyway. Alex Singla and his McKinsey coauthors report that 94% of businesses have yet to create meaningful value from AI. Faisal Hoque, Tom Davenport, and Paul Scade argue in HBR that most AI-attributed layoffs are ordinary restructuring in new packaging. Christine Moorman and her coauthors show, through the 35th CMO Survey, that marketing AI adoption has nearly doubled in two years while no marketing technology activity scores above 5 on a 7-point performance scale. And Charlotte Rogers reports that under half of brands have a well-defined effectiveness function.
Read those together and you get one diagnosis. Organizations bought capability and skipped accountability.
I have watched this pattern inside Fortune 500 marketing organizations, and the tell is always the same. Ask who owns the AI-driven personalization program and you get a list of six names. Six names is zero names.
Pick three domains and defend the list against everything else
Singla, Alexander Sukharevsky, Kate Smaje, Eric Lamarre, and Robert Levin published The new management playbook for AI in McKinsey Quarterly on August 28. They studied 20 companies that have produced returns — roughly 20% EBITDA improvement after three years, about $3 of incremental EBITDA per $1 invested, cash-flow positive inside one to two years.
The number that matters for planning season: two-thirds of those companies concentrated on three business domains or fewer.
Focus discipline is the finding. The authors name six foundational capabilities — C-suite fluency, front-line leaders with genuine technical depth, an operating model built for speed, enterprise platforms, consumable data, and design for scale from the first prototype. Only 10% of companies have adopted the distributed operating model they describe. The capability list gives you a maturity check. The scoping rule gives you something to act on Monday.
Commentary: Singla and his coauthors put a hard number on something most CMOs already suspect — the sprawling pilot portfolio is the problem, and the fix is subtraction.
Decompose the work before you decide anything about headcount
Hoque, Davenport, and Scade published AI Transformation Requires Redesigning Work, Not Cutting Roles the same day. Their term for the practice is AI-washing: "Many layoff announcements are better understood as AI-washing: ordinary restructuring initiatives packaged in AI language to reassure investors." They cite Goldman Sachs analysts who attribute roughly 16,000 jobs of reduced monthly US payroll growth to AI over the past year, and about 0.1 percentage points of unemployment.
Their method is task-level decomposition. Break the role into discrete tasks, find where a model measurably improves performance on a specific task, then reconfigure the human-machine combination around that finding.
It reminds me of what Don Schuerman, CTO and Head of Marketing at Pega, said when I interviewed him on The Agile Brand podcast: "I think more than asking what the latest model version brings, the real unlock for AI value is redesigning processes and operating models, not just getting the newest model."
Commentary: The authors give you the sequence — decompose, test, reconfigure — and the sequence matters because teams that start from a headcount target skip straight to the reconfiguration step and get a smaller team doing the same badly designed work.
Treat the CMO Survey numbers as a diagnosis of your own operating model
Moorman, Mara Michel, and Elise Romola published Marketing Has an AI Problem, and It Has Nothing to Do with AI on August 27, drawing on 308 senior marketing leaders. Adoption is broad: content creation at 73.9%, personalization at 65.4%, automation at 48.9%, data analysis at 46.3%, targeting at 45.2%, generative engine optimization already at 41.5%. Leaders report gains: sales productivity up 14.1%, customer satisfaction up 10.8%, marketing overhead down 14.6%.
Then the structural causes. Training budgets fell from 5.8% of marketing spend in 2019 to 3.8% today while 60% of organizations still build capability internally. Teams shrank. Deployment stayed siloed. The CMO–CFO relationship scores 4.5 out of 7.
Of the five causes, I weigh that last one heaviest. A marketing leader who cannot make a joint case with finance cannot fund a multi-year capability build, and the McKinsey returns arrive on exactly that timeline.
Compare the service side. On August 26, Kim Hedlin and Daniel O'Sullivan reported that AI spending by customer service leaders surged 38% while overall service budgets rose 2%. Service leaders funded AI by reallocating from labor and overhead inside a flat envelope. They made a portfolio decision and defended it to finance. Marketing leaders, on the evidence of a 4.5, largely have not — which is one reason agentic CX programs keep landing in service organizations first.
Commentary: Moorman's team supplies the missing variable in every other study this cycle — the specific organizational conditions that have to exist before AI adoption converts to margin, and the survey data showing marketing has been defunding those conditions for seven years.
Give effectiveness an owner, because outcomes do not distribute
Rogers reported on August 28 that under half of brands have a well-defined effectiveness function, and that marketers increasingly treat effectiveness as a shared responsibility across the organization. The day before, she reported that 70% of marketers say their business prioritizes effectiveness while two-thirds remain tasked with delivering cost efficiencies.
Here is where my own framework has hardened. Effectiveness is an outcome. Growth, agility, and retention are outcomes too. Capabilities produce outcomes, and only capabilities can be assigned — which is why I keep intelligence, operations, and experience as the horizontal capability layer in the framework I am building on top of Coherence at Velocity, and keep outcomes out of it entirely. Assign an outcome to everyone and you have assigned it to no one. Assign the marketing operations capability that produces it to a named executive and you have something a CFO will fund.
The other principle from that work applies directly to the HBR piece: humans stay accountable for direction even when the analysis is fully automated. A model can score a workflow. A person decides what the workflow is for.
It reminds me of what Christine Royston, Chief Marketing Officer at Wrike, said when I interviewed her on The Agile Brand podcast: "As you think about your workflows, your processes, certainly you want them to be robust and measurable. Now we need to make sure they are flexible because technology is changing so much."
Commentary: Rogers documents accountability diffusion in progress, and the two-thirds-still-cutting-costs figure explains why — marketers carrying an effectiveness mandate on top of an unchanged efficiency mandate have every reason to leave the function undefined.
Run the math on a portfolio you recognize
A $2B specialty retailer runs eleven AI initiatives across marketing: email subject-line generation, paid social creative variants, service deflection, product content localization, lookalike modeling, forecast enrichment, brief automation, review summarization, SEO clustering, media mix modeling, and a chatbot on the help center. Nine directors own pieces. Marketing spend is $84M. After fourteen months, attributable EBIT contribution is zero, and the CFO has started asking about it in QBRs.
Apply the scoping rule. Three domains, chosen for margin proximity: service deflection, lifecycle personalization, content production. The other eight initiatives go dormant, and someone senior says so out loud.
Now decompose one of them. Content production runs 14 people producing roughly 400 assets a quarter. The tasks inside that: intake and brief, first draft, brand review, legal review, channel adaptation, trafficking, performance tagging. Your team gets measurably better output on first draft and channel adaptation with a model in the loop. Brand review and legal review get slower under that volume, because someone has to read more output. Trafficking and tagging are deterministic, so you run them on rules rather than a model — the lightest mechanism that clears the accuracy bar.
The redesign follows the task map. Two roles shift from drafting to brief quality and review throughput. Adaptation capacity triples. Quarterly assets go from 400 to 1,100 with the same 14 people, and the constraint moves to review, which is a staffing decision you can now make with evidence.
That is a defensible three-year case. The eleven-initiative portfolio produced fourteen months of activity and no case at all.
Do these four things before the next planning cycle
Cut to three domains. Rank every AI initiative by distance to a P&L line. Keep three. Say the word "dormant" about the rest in a room with your CFO in it, and put a date on revisiting them.
Name one accountable executive per domain. Give each a capability to own and a number to hit. Shared ownership of an outcome is the failure mode Rogers is documenting.
Rebuild the CMO–CFO joint case first. A 4.5 out of 7 is the binding constraint on everything else in this brief. You cannot fund a multi-year capability build across a relationship that weak, and no amount of pilot velocity substitutes for it.
Decompose before you restructure. Map the tasks, test where your team performs measurably better with a model in the loop, then design the roles. Teams that reverse this order buy the org chart they already had, at lower headcount.
The seat with the name on it is the whole finding this week. Four research teams, four methods, one answer. Fill it before Q4 planning closes, and put the person's name in the deck.
Featured this cycle:
The new management playbook for AI: How to move faster and create more value — Singla, Sukharevsky, Smaje, Lamarre, Levin, McKinsey Quarterly, August 28, 2026
AI Transformation Requires Redesigning Work, Not Cutting Roles — Hoque, Davenport, Scade, Harvard Business Review, August 28, 2026
Marketing Has an AI Problem, and It Has Nothing to Do with AI — Moorman, Michel, Romola, American Marketing Association, August 27, 2026
Under half of brands boast 'well-defined' effectiveness function — Charlotte Rogers, Marketing Week, August 28, 2026