Buyers moving to AI faster than the function built to reach them needs a shift.

A CMO sat down with us this year, leading marketing and sales for an international media business, the kind of company its audience had relied on for decades.

“Our brand is trusted. Our content is good,” she said. “But somehow that isn’t showing up as first choice. And our people are already using AI everywhere. I just don’t know if we’re using it on purpose and in the right way.”

What she described was familiar too. Sales people and marketers across the team were individually experimenting with tools like ChatGPT and CoPilot in their day-to-day work, with no shared framework, no governance, and no clear link back to what the business was actually trying to achieve commercially. Meanwhile, the market had delivered its verdict before internal debate caught up with it. Brand tracking showed audience preference for the portfolio sitting at roughly half the level of its two biggest competitors, with a meaningful share of the audience naming no preferred brand at all.

On AI specifically, the gap was starker: almost none of the customers surveyed associated the business with AI, against a runaway share for a single AI-native rival.

Her brand was trusted. Her people were capable and already adopting the tools. Her function, in effect, was reacting to a market that had already moved on, without a shared plan for how to move with it.

This is The Shift.

The market and the function must move together

Buying behaviour changes because buyers themselves now use AI to research and evaluate faster, which raises the bar on what “first choice” requires and rewards the vendors who are visibly fluent in it. Internally, functions default to individual, ungoverned adoption, because no one has decided what’s approved, what’s governed, and who owns it. The shift is not just what customers now expect. It’s whether the operating model behind the function can meet that expectation coherently, at scale, without every individual reinventing the approach for themselves.

Most leaders assume that once individuals start adopting AI, the organisation is adapting. It rarely is. Individual adoption without governance produces inconsistency and risk exposure, and none of it compounds: a clever prompt or workflow dies with the person who found it. Meanwhile, the market doesn’t wait for internal governance debates to resolve. Competitors moving first on both fronts, visible AI fluency and genuine buyer relevance, set the new bar for preference while the rest of the field is still deciding who owns the conversation.

“Left alone, it splits a team in two,” she said, describing what she’d already seen. “Half of them barely touch it. The other half treat it like an intern with internet access and no supervisor, and just believe whatever it tells them.” That split is the internal mirror of the external problem. A function can’t present a coherent, AI-fluent face to a market that’s already moved, while half its own people ignore the tools and the other half trust them uncritically.

In her business, the symptoms were textbook. Marketers were already fluent individually, but there was no shared tool stack, no risk framework, and no current view of which channels buyers now actually used to evaluate the category. Brand awareness stayed high while preference and share of consideration quietly slipped. The one thing the function could not do was point to a coherent, function-wide answer to “what’s our AI strategy” that connected to commercial outcomes.

Not a talent problem. Not a tooling problem. A translation problem.

Where to start

1. Start with the market as the fixed point. Before deciding anything about tools or structure, get a current, evidence-led read of how buyers actually research and choose now: where AI has changed what they pay attention to, which channels have quietly gained or lost relevance, and where the brand’s equity is and isn’t converting into preference. Anything decided about capability has to be tested against this, not against internal assumptions about what used to work.
2. Then translate the group-level AI position into a function-specific strategy. A corporate stance on AI is not the same as a working plan for marketing and sales. Decide which tools are approved for which tasks, what governance applies at what level of risk, and who owns the capability centrally, rather than leaving it to whoever happens to be most enthusiastic. Ownership sounds like an org chart question. It isn’t. Leave it unanswered and the function defaults to one of two outcomes: rogue experimentation happening in the shadows while most of the team waits for permission, or one unlucky enthusiast turned into the permanent “AI person” everyone routes through instead of a capability everyone builds. Neither is capability. Both are dependency.
3. Finally, redesign the operating model around it. AI capability has to become a shared, structural asset, not one specialist’s job or forty individual habits. That usually means a new owning role, a shared tool stack and prompt library, and workflows rebuilt around how work actually needs to move between marketing, sales, product and customer success.

Use the moment to question the workflow, not just automate it.

The instinct is to point AI at how things are already done. That’s a mistake if how things are already done isn’t working. AI adoption is the reason to ask which workflows earn their place and which have simply always been there, and to redesign before automating rather than after. Done well, it’s also the moment to codify what currently exists only in individual heads: the account context one seller carries, the pitch judgement one strategist has built up over years, the customer relationship one account manager owns personally. Left uncodified, that knowledge, access and relationship equity is a single point of failure for the business. Captured properly, as shared playbooks, prompts and structured account knowledge, it becomes an asset the whole function can draw on, not a dependency on whoever happens to hold it.

Be deliberate about how the change is framed.

Leadership teams default to talking about AI as an efficiency play, saving time, doing more with less, because that’s the easiest business case to write down. It’s also the fastest way to stall adoption. Frame it that way, even unintentionally, and the team hears exactly what’s being said: the business thinks it needs fewer of them. That’s a rational thing to hear, and a rational thing to quietly resist. The framing that actually moves a function is impact, not efficiency: freed-up time reinvested in the work that drives revenue and deepens relationships, not time banked against headcount.

Then direct where the time actually goes, deliberately. Protecting the reinvestment story is not the end of the work, because saved time doesn’t automatically find its way to anywhere useful. Left unmanaged, Parkinson’s Law takes over: work expands to fill the time available, and existing habits simply spread into the extra hours. The measure of success quietly shifts from whether the work performs to how fast it got done, and a business can save a great deal of time without becoming any more effective for it.

Leadership has to point the freed capacity somewhere specific. For a commercial function, not just marketing, that usually means testing activity against a short set of questions:

• Is there enough activity running from first awareness through to customer success to build relationships across the whole market, not just short-term lead generation?
• Are the commercial team’s conversations, sales, account management and marketing alike, properly supported with the personalised, targeted material that brings them closer to revenue?
• Are there enough occasions for real people to build the relationships B2B buying still depends on, rather than fewer meetings and less time in the room?
• Is anything being done to earn recommendation deliberately, content and communications built to be shared across a buying committee, rather than left to chance and last-click attribution?
• Is the business genuinely listening to customers at every touchpoint, and feeding that back in as the voice of the customer, rather than assuming satisfaction?
• Are the commercial and technical use cases for the product or service fully evidenced, so customers are convinced by proof rather than assertion?
• Are procurement and tender credentials, ESG, AI policy, security, ready in advance, rather than assembled under the pressure of a live tender?

None of this happens by default. It’s the difference between AI creating more time for a commercial function and AI creating more impact from it.

The pragmatic reality

She listened, agreed with the logic, and raised the real constraint.

“I don’t have twelve months to get this perfect,” she said. “I’ve got a global conference and our first CMO summit this year, and I need the whole function moving together in months, not next year.”

She was right. Waiting for a fully mature governance framework before doing anything visible costs credibility the function doesn’t have to spare, especially once a competitor is already being named by the exact audience you’re trying to win. The answer isn’t to slow the diagnostic work. It’s to run it fast and in public: align leadership early, a first global summit is exactly the right forum for that, then equip the front line directly once the framework exists, rather than waiting for perfect before anyone outside the project team sees a plan.

Connecting the two: the approach in outline

The work typically runs in four connected stages.

1. Diagnose. Pull together everything already known, brand tracking, prior workshops, sales strategy, customer research, and test it against fresh, mixed-method research into how buyers in each region actually behave now. This usually reframes the brief: the gap is rarely about willingness to adopt AI. It’s the missing translation from a corporate position into a working function-level strategy.
2. Design the capability. Build the governance framework, a tiered risk model, an evaluation method for any new tool, and a prioritised, task-mapped stack that becomes the function’s shared standard, replacing ungoverned individual use. Give it a named owner to drive that early work and scan the horizon, but treat the role as a catalyst, not a permanent home. The destination is a function that’s AI-literate throughout, not one specialist everyone defers to. Use the same exercise to question which workflows deserve to be carried forward at all, and to codify the account knowledge and relationship context that otherwise lives only with individuals.
3. Rebuild around the buyer. Redesign the channel and content strategy market by market, around where buyers actually pay attention now, not where they used to.
4. Redesign the structure and prove it. Give the capability a lasting home across three layers, not one role. Leadership sets governance and makes AI literacy an explicit expectation of every marketer, not an optional extra. The named owner keeps driving adoption, tool assessment and horizon-scanning. Team managers embed the practical training, evolving workflows and peer examples that make it real day to day rather than theoretical. Add a phased roadmap so leadership and the front line see the function move from reactive experimentation to proactive, sponsored execution, not just read about it.

What to watch for

The signs worth paying attention to:

• Brand awareness stays high while preference and consideration quietly slip
• Individuals across the team are already using AI tools, but no one could tell you the function’s actual policy on it
• No answer exists to “which tools are we allowed to use, for what, and who decides”
• A competitor, often one far younger, is named unprompted by the exact audience you’re trying to win
• Channel and content plans still assume how buyers researched two or three years ago
• Governance, if it exists at all, sits at group level with no translation into how marketing and sales actually work day to day
• Capability lives with one enthusiastic specialist rather than with the team, and everyone else routes requests through them
• Leadership can describe the corporate AI position confidently, but not the function’s
• The team is quietly split in two: half ignoring the tools, half trusting them uncritically
• Leadership talks about AI mainly as a way to save time or do more with less, and adoption is stalling anyway
• Nobody can point to where the time AI has freed up is actually being spent
• AI is being pointed at workflows nobody has questioned in years, just to make the same work faster
• Account knowledge, pitch judgement or customer relationships exist only in individual heads, uncodified and unshared

None of these are fatal on their own. Together, they are the early warning that the market has already shifted and the function hasn’t moved with it.

If any of this sounds familiar, it usually is. And it is, with the right sequence, entirely fixable.