What marketers actually need to change to make AI work for engagement

Almost every marketing team now has AI somewhere in its stack. Fewer of them have anything to show for it. That gap, between adoption and actual results, is where most of the AI conversation in marketing needs to move next.

The numbers back this up. Adobe’s 2026 AI and Digital Trends report, based on a survey of 3,000 executives and practitioners, found that ambitions for AI-powered, real-time personalisation are running well ahead of what most organisations are actually equipped to deliver. Data quality, unified customer profiles, and the analytics infrastructure to support any of it are still missing at most companies. Barely a third have moved agentic AI for customer support past the pilot stage.

It means most marketers are using it the wrong way, and the fix has less to do with the technology than with how teams are set up around it.

Spend is rising faster than readiness

According to StackAdapt and Ascend2’s February 2026 survey of brand and agency marketers, 87 percent of brands plan to increase personalisation spend this year. At the same time, 68 percent describe themselves as still in the early stages of implementation. Budgets are moving faster than capability, which is roughly the opposite order most successful technology rollouts follow.

This shows up most clearly in how AI gets deployed day to day. Twilio’s State of Customer Engagement research found that 56 percent of brands now use AI to tailor customer interactions, and 96 percent say AI has improved customer-facing operations in some way. Encouraging on the surface. Less encouraging once you ask what “improved” actually meant in each case, because the same research also found that only 45 percent of consumers feel genuinely understood by the brands they interact with, a figure that has been falling.

Somewhere between the AI being deployed and the customer feeling understood, something is getting lost.

Personalisation done badly can actively cost you

There’s a sharper warning buried in Gartner’s June 2025 research: customers exposed to poorly executed personalisation were 3.2 times more likely to regret a purchase, and 53 percent reported a negative experience from it overall. Bad AI personalisation actively damages trust in ways a generic, un-personalised experience never would.

This is the part of the AI conversation marketers tend to skip. Most guidance focuses on capability: what the tool can technically do. Far less attention goes to restraint: knowing when a recommendation is confident enough to act on, and when it’s better to say nothing at all. A subject line test that doesn’t beat its control by a meaningful margin should get retired.

What separates the brands actually seeing results

CMO Council research offers a useful split here. Marketing teams it classifies as AI “Power Partners”, meaning those who’ve redesigned their workflows around AI-human collaboration rather than bolting AI onto existing processes, are six times more likely to report a major impact on personalisation than teams still treating AI as a productivity add-on. Seventy percent of Power Partners are actively redesigning how work gets done. Among everyone else, that figure sits at 7 percent.

Most of these teams are working with similar platforms. The difference is whether AI sits inside a redesigned workflow with clear ownership and a feedback loop, or whether it’s been dropped into the old workflow and expected to improve results on its own.

What this actually requires from a marketing team

Three shifts show up consistently among teams getting real engagement gains from AI, rather than just AI activity.

First, a genuine first-party data foundation. Without that, AI is working from partial information no matter how sophisticated the model behind it is.

Second, a habit of measuring incremental lift. New AI-generated content or a new recommendation engine should be judged against a control group and a specific threshold.

Third, workflows that are actually rebuilt around AI rather than layered with it. That means clear rules for when AI acts independently, when a human reviews first, and what happens when the model gets something wrong. The Power Partner data suggests this structural change matters more than any individual tool decision.

None of this is about doing more with AI. It’s about doing less, more precisely, with a clearer sense of when it’s actually earning its place in the customer relationship.

The real adoption gap

The marketers seeing genuine engagement improvements from AI are the ones treating it as a change to how the team works. Everyone else is layering a fast, confident tool on top of a slow, uncertain process, and wondering why the results don’t match the ambition.