In the Agentic AI Asymmetry Dilemma series Stuart Bailey, co-founder and managing director of AgencyFit, named a structural problem opening up across our industry: agency AI accelerating while advertiser and publisher systems stayed still. That problem hasn’t gone away, but naming it isn’t enough on its own, so this series asks the other half of the question: what actually closes the gap? He has called it the Connected Advantage. Build together, or drift apart.
It’s still an ecosystem problem, and it still needs all three parties to fix it. This series looks at what each one needs to do, and where collaborative AI can genuinely bring this industry closer together instead of pulling it further apart. The first article focuses on advertisers, and on a workflow join that sits between advertiser and agency that nobody owns.
A quick note on sourcing. I’ve drawn on ACAM and Kantar’s ‘Closing the AI Maturity Gap’ (July 2026) research, which benchmarked 126 Australian CMOs across seven maturity drivers and six levels of AI marketing maturity.
I think the most useful number isn’t the average; it’s the gap between the top and bottom of that list. Leadership and Culture scores 3.09, the strongest driver. Roadmap, Planning and Investment scores 1.70, the weakest, unmoved since 2025. Team Design and Workflow sits at 2.19. Eighty-three per cent of organisations are in the first three of six levels of maturity, and not one has reached the fifth.
Most interesting is that the two weakest drivers don’t sit inside the marketing organisation at all.
Workflow Doesn’t Stop At Your Team
Only nine per cent of CMOs hold a clear view of how their marketing team should be structured around AI. But workflow isn’t defined inside a marketing team. It’s defined at the join between two organisations. You can optimise your half, the agency can optimise its half, and what you get is two efficient halves that don’t meet. Nobody is measured on the join. Nobody owns it.
So, start here. Is your agency workflow delivering great work and strategy, or is it a source of tension? If it’s the latter, the harder questions are about your own half: do you brief well, and do you hand over last-minute upweights as routine?
The same fault shows up in Roadmap, Planning and Investment, the weakest driver. Just 16 per cent of CMOs are happy with what their AI plan is delivering. The discipline is real, the scope is the problem: a roadmap that stops at the edge of the department is a roadmap for the half of the work you can see.
The solution is simple to say and hard to do: build one workflow to rule them all.
Whose Maturity Is This
Almost 90 per cent of advertisers say they’re already using or building agentic AI (Digiday, State of Agentic Advertising, 2026). Of those, 61 per cent are specifically applying it to media planning, data analysis and personalisation (Clutch, Marketing Budget Planning in 2026). Set that against the ACAM research, where 83 per cent of marketing organisations sit in the bottom half of the maturity scale, and a fair question follows. If the advertiser isn’t mature, whose maturity is running the account?
It’s really four questions. Are you owning AI on your account, or is your agency owning it on your behalf? Are your KPIs and your data feeding the models being run for you? Do you know what these systems optimise toward? And how much of what AI does on your behalf can you actually see?
I would wager most advertisers can’t answer all four, and that isn’t a failure of effort. Agencies are building proprietary systems that plan, allocate and evaluate using data and logic advertisers can’t see. This creates what Bain calls proprietary intelligence: unique data, encoded workflows, a learning architecture that improves with every deployment. That’s a legitimate strategy, not a criticism, it’s the predictable result of one side investing and the other watching. Only 51 per cent of organisations have the cloud infrastructure agentic AI requires (Adobe, 2026).
If your KPIs and your data aren’t in the model, are you benefiting from that investment, or quietly subsidising it?
The Pitch Is Not A Diagnostic
When the join stops working, the reflex is to change the counterpart. Here’s the trouble. A traditional pitch replaces the people on one side of a join neither side has ever mapped. The incoming agency inherits the same undefined workflow and invisible criteria. Soon enough, the same fault returns with new faces attached to it.
Before deciding the incumbent partnership is the problem, ask two questions instead. How well is the partnership structured against your business objectives, rather than whether you like the team? And how ready is each side to work with agentic systems rather than around them, not just tolerating what’s already there.
Design Before You Build
The order matters more than the technology. Map the joint workflow end to end, agree the objectives it exists to serve, and make them visible to both sides. Then decide who owns each decision, which decisions can be delegated to AI and which need to remain human. Only then decide what’s worth building. Get the order wrong and it shows: 53 per cent of organisations run fewer than five live use cases, and only 41 per cent show a clear return.
This is the signature of a use case chosen from a menu rather than derived from a problem.
AI Amplifies What It’s Given
Skip the redesign and the outcome is predictable. Where the objectives aren’t unified, AI doesn’t resolve the disagreement, it accelerates it. BCG’s AI at Work research found people with a clear AI strategy and limited tools outperform people with strong tools and no strategic direction by 25 percentage points on measurable business impact. Strategy beats tools. Sophistication doesn’t fix bad inputs; it amplifies them. AI doesn’t repair a badly designed relationship. It industrialises it.
That’s why the research shows that Use Cases and ROI scores 2.41, with 31 per cent of CMOs saying fewer than one in five use cases delivers a clear return.
You can’t evidence a workflow neither side owns end to end.
Learn To Drive
None of this argues for spending less, or for measuring AI by what it saves. The measure that matters is whether AI moves business metrics: share, revenue, lifetime value, effectiveness over time etc. That almost certainly means investing more in the short term, in strategy and capability as well as technology, because that’s the work that compounds.
The advertisers who thrive won’t be the ones who outsource their AI to their agency and hope, or the ones who resist because the current model still “works”.
They’ll be the ones who own the workflow redesign, and who build enough capability to act as an intelligent co-pilot, enough to understand what the system optimises toward, interrogate it, and redirect it when it drifts. You don’t need to build the engine. But you do need to learn to drive.
The winners in the next era will be the advertisers who understand that workflow doesn’t stop at their team, and that AI can only amplify the process it’s given.
In the next piece I’ll turn to publishers, who are living the same asymmetry from further down the workflow, optimising for criteria no agency has ever shown them.
Written by Stuart Bailey, co-founder and managing director of AgencyFit.
