AI use is becoming easier to observe and harder to assess. Licences, pilots and feature announcements can create an impression of maturity that the underlying operating model does not support.
The latest figures from the Office for National Statistics offer a useful corrective. Adoption has accelerated, but extensive use remains unusual. For marketing leaders, this is a favourable moment: interest is established while the operating model remains open to design.
Adoption is broadening faster than it is deepening
Among UK businesses with at least 10 employees, the proportion reporting use of at least one AI technology rose from approximately 12% in late 2023 to 35% in June 2026. That is a substantial change, but far from a settled market. Around two-thirds of businesses in scope were still not reporting formal use.
Among businesses already using AI, the average number of technologies used increased from approximately 1.4 to 1.6. The count is only a proxy for depth, but it suggests that breadth is moving slowly.
Only 10% of AI-using businesses described their use as extensive. Just 15% said AI featured in most employees’ daily work. These new measures are early indicators, but together they suggest widespread organisational adoption remains some distance away.
Close to three-fifths of AI-using businesses named improving existing operations as one purpose. The larger gain comes when AI starts moving a complete piece of work forward.
The ONS research covers the wider economy rather than regulated marketing or communications specifically. Its relevance is the pattern: access is spreading much faster than operating models are changing.
Value sits in constrained workflows
Regulated teams often have no shortage of ideas. Capacity is lost in the work between brief and approval: assembling substantiation, adapting approved materials, checking recurring claims and preparing variants for distinct customer groups.
Reducing friction across those steps offers a stronger business case than accelerating the first draft. AI can take on repeatable work from a structured brief to a review-ready output, leaving experts more time for cases that need interpretation, challenge or a genuinely new decision.
For a regulated marketing team, that means bringing the rules into the work earlier. The system should know the product, audience and channel, draw from approved evidence and surface comparable decisions. New judgements should move clearly to the right person.
Start with a workflow that can be improved
A useful candidate has recurring volume, a visible queue, reliable inputs and a clear definition of acceptable output. Channel adaptation, claims triage and recurring customer communications often meet those conditions.
A capable workflow assembles the brief and evidence, produces or checks the work, routes genuine exceptions and records the outcome. Otherwise the next job begins with the same search for context.
This approach gives marketers room to be more ambitious. Ideas do not need to be rationed simply because review capacity is fragmented. Routine work can move faster while legal, compliance and subject-matter experts remain accountable for the judgements that matter.
Employee use is moving ahead of formal operating models
The ONS found that 55% of employees reported using AI for work or education, compared with 35% of businesses with at least 10 employees reporting use of AI technology.
The surveys cover different populations and questions, so the percentages are not directly comparable and do not prove unauthorised use. Together, they raise the possibility that some AI activity is informal, ad hoc or missing from business reporting.
For leaders, the implication is practical. Staff need a credible route to use AI with authoritative information, clear boundaries and an approval path. The FCA’s Mills Review also indicates that consumers are increasingly willing to use general-purpose AI for financial information.
Adoption deepens when decisions become reusable
The durable advantage comes from connecting AI to the organisation’s regulation, internal policy, brand standards, product and customer context, approved evidence and previous decisions. Human judgement remains visible, but it no longer has to start from a blank page.
A useful system shows which sources informed a suggestion, identifies missing substantiation and distinguishes a settled requirement from an issue needing authorised review. The final decision, including its reasoning and conditions, becomes part of the next brief.
When a claim such as “clinically proven” is narrowed to match the evidence, or a qualification is approved only for a particular audience and placement, the reasoning should be available to the next marketer facing a similar decision. This is central to brett’s view of review: ambitious teams should be able to move faster because the organisation remembers what it has already learnt.
The ONS figures describe a market with growing participation and limited depth. That gives marketing leaders space to design carefully: choose a material workflow, connect it to authoritative context and use each decision to improve the next.
What marketers need to know
How mature is business AI adoption today?
Participation is increasing quickly, but organisational depth remains limited. Only 10% of AI-using businesses describe their use as extensive, leaving significant room to build more capable operating models.
What separates experimentation from operational adoption?
Operational adoption creates a repeatable path from a structured brief through generation, evidence checks, review and approval. It produces a defined outcome and preserves the decisions needed to improve the next piece of work.
Which marketing workflow should be addressed first?
Prioritise work with recurring volume, a visible queue, reliable evidence and a clear definition of acceptable output. Channel adaptation, claims triage and recurring product communications often meet those conditions.
What context turns a general AI model into organisational capability?
Connect it to the product, audience and channel, applicable regulation, internal policy, brand standards, approved evidence and previous decisions. Better context allows more routine work to move before an expert steps in.
Which measures show meaningful progress?
Track cycle time, rework, backlog completed, consistency across comparable claims and the quality of escalations. Prompt counts and licence activation show activity rather than organisational capability.