Centralise marketing evidence, not every decision

The campaign dashboard says a segment is responding. Customer interviews suggest the message is landing badly. Sales has a different explanation again. By the time someone reconciles the evidence, next quarter’s plan is already being approved.

AI makes it easier to put those signals in the same room. It does not decide who should believe what, or who is allowed to change the plan. For a marketing leader, that is the operating-model decision: give the team a shared way to assess evidence, while keeping campaign decisions with the people accountable for them.

More synthesis can still produce a worse decision

In September, Gartner argued that separate customer-insights, research and analytics teams can produce disconnected recommendations, and recommended bringing those capabilities under a unified insights leader. That is a useful diagnosis. If every team defines a target account, a qualified response or a successful campaign differently, asking AI to compare their work will produce a polished account of incompatible things.

But a central insights function can become another queue. A demand generation lead should not need permission from an analyst to change a test that is clearly underperforming. Nor should a shared team become the default owner of positioning, channel spend and regional judgement simply because it can assemble the evidence. Centralise the standards and the ability to challenge an interpretation. Keep the decision close to its commercial consequence.

There is good reason to sort out that distinction now. In Marketing Brew’s survey of more than 450 newsletter readers, fielded in July 2026 and published in September, 65% said AI strategy and governance had been added to their remit. This is a survey of its readers, not a representative census of B2B CMOs. Still, it describes a familiar leadership problem: responsibility for AI is growing faster than agreement on how marketing should use its output.

Give the evidence an owner and the decision a different owner

Consider a hypothetical B2B SaaS company testing a campaign for a new EMEA segment. Marketing sees higher engagement with a message about faster implementation. Recent customer interviews, however, suggest buyers choose the product for its control features. Which message should the team fund next month?

AI can help a marketer pull together interview notes, approved claims, campaign results and relevant deal feedback. It can identify which observations support each message and show where the sources disagree. That is assistance with synthesis, not a mandate to change the campaign. A click says something about an asset’s appeal; an interview says something about what a buyer reports. Neither, alone, establishes why an account purchased.

In this example, marketing operations could own the source inventory, definitions and access rules. Product marketing would check whether the proposed message is supported by customer evidence and the product’s actual capabilities. The campaign owner would decide whether to test it, how much budget to put behind the test and what result would justify a broader change. The marketing leader would step in when the choice changes the company’s positioning or moves a material amount of spend.

Those are assignments, not a prescribed org chart. In a lean team, one person may cover more than one role. What matters is that a recommendation arrives with its sources, date range, gaps and an identifiable decision owner. “AI found a trend” should never be enough to win a budget argument.

A recent Acquia-commissioned survey gives a useful warning about the inputs. In its September 2026 release, 61% of the 250 CMOs surveyed across the US, UK and France said their content was spread across too many systems to provide a single reliable source of truth. That finding concerns their content environments, not all marketing data. But it illustrates the risk of treating access to several systems as proof that an AI-generated answer has a sound foundation. The study was commissioned by a vendor, so it is best read as a reported concern, not proof that any particular platform fixes it.

Build a decision rhythm before buying a new layer

A CMO can start without reorganising the department. Pick one recurring decision that currently takes too long or attracts contradictory evidence: whether to continue a segment campaign, revise a message or shift spend between programmes. Document what the decision owner needs to see, which sources count and which uncertainties must remain visible.

Then make the rhythm explicit. At an agreed review point, the campaign owner gets a short evidence pack: the question being decided, comparable performance data, relevant qualitative observations, source dates and a proposed next test. AI may help assemble and summarise it. Marketing operations or an insights specialist checks definitions and missing inputs. The owner records the decision and the reason for it, including when they reject the AI-assisted recommendation.

This is deliberately less glamorous than an autonomous agent. If the problem is that no one agrees what “engaged account” means, a better definition beats a better prompt. If the sources are reasonably sound but scattered across notes and reports, AI-assisted synthesis may save effort and expose disagreements. More elaborate integration only earns its keep when the same decision repeatedly depends on context moving between systems and people.

Set an escalation rule too. A campaign owner can approve a bounded test, but a new product claim goes to product marketing. A decision affecting several regions or changing an agreed budget allocation goes to the marketing leader. That protects judgement without sending every email subject line through a committee.

McKinsey’s October 2026 article on AI operating models reports that, across organisations rather than marketing teams specifically, those furthest along in AI reinvention are more likely to place decisions and accountability closer to the work. It is not evidence that decentralising marketing decisions will increase revenue. It does reinforce the practical question a CMO should ask of any redesign: are we removing an unnecessary hand-off, or just giving it a newer interface?

Judge the model by what the team changes

A faster evidence pack is a useful first result, but it is not the commercial result. Review whether decisions now happen in time to affect live spend; whether campaign owners can explain the evidence behind a change; and whether subsequent tests distinguish a promising message from a misleading signal. Track reversals too. Discovering that an interview sample was too narrow should lead to a better test, not an embarrassing defence of the original recommendation.

If the team produces more summaries but makes the same late decisions, the operating model has not improved. My judgement is that marketing leaders should resist turning AI into a central brain for the department. Give evidence common standards and serious scrutiny, then give accountable marketers room to act. The expensive mistake is not a lack of answers. It is building a faster route to answers that nobody owns.

This blog was created by Jan’s AI content agent. Get in touch if you want to know how it works.