TL;DR
- The government published six sector AI adoption plans on 8 June 2026, each written by an independent AI Champion from that industry. None of them is about marketing. Read together, they describe most marketing functions almost exactly.
- The line worth pinning up is from the Digital and Technologies plan. Using AI and integrating AI are not the same thing, and marketing is where the gap is widest.
- The numbers back it. In professional services, 69 per cent of firms have expanded generative AI training while 70 per cent report limited progress on redesigning how the work actually flows.
- Two ideas are worth borrowing into a marketing plan: manufacturing’s Scan, Pilot, Scale route out of the pilot trap, and the creative sector’s augmentation-first principle.
- The workforce question nobody in marketing is asking out loud is what happens to the junior work, because that work is how you make senior marketers.
In June the government quietly published six sector AI adoption plans, each one authored by an independent AI Champion, covering advanced manufacturing, clean energy, creative industries, digital and technologies, life sciences, and professional and business services. Not one of them mentions the marketing function. Read them back to back and they describe it line for line.
Here is the short answer, in case you read no further. The plans agree the problem is no longer getting hold of AI. It is integration. Firms are using AI in pockets, at individual level, and calling that adoption. The Digital and Technologies plan, written by Katie Gallagher OBE, states it plainly, that the harder challenge is “scaling these into production environments and embedding them into day-to-day workflows”. The Professional and Business Services plan, written by Shaheen Sayed, puts numbers on the same gap. Sixty-nine per cent of firms have expanded generative AI training. Three-quarters are not ready on the core enablers such as data, orchestration and monitoring, and 70 per cent report limited progress on process redesign.
Marketing is where that gap opens widest, for a simple reason. Marketing was the first function in most businesses to get the tools, and it is usually the last to change the process.
In this article
- The six plans, and which one to actually read
- Where using AI quietly becomes the ceiling
- The pilot trap has a marketing version
- Augmentation first is a risk position, not a soft one
- Automate the junior work and you stop making marketers
- The hire is not an AI person
- What to do with this in the next 90 days
- Questions I get asked about this
The six plans, and which one to actually read
You do not need to read six. You need to read one, and it should be the one for the sector you sell into rather than the one you technically sit in.
All six are on GOV.UK:
- Advanced manufacturing, by Chris Dungey
- Clean energy, interim, by Lucy Yu
- Creative industries, by Sally Davies
- Digital and technologies, by Katie Gallagher OBE
- Life sciences, by Dave Hallett
- Professional and business services, by Shaheen Sayed
There is also a summary of all six and a government response. The financial services plan is still to come.
If you sell to professional services firms, read Shaheen Sayed’s. If you sell into industry, read Chris Dungey’s. If your business is creative or content-led, read Sally Davies’. Read it for the language your buyers will be using in board meetings for the next two years, and for the anxieties sitting underneath that language. The policy itself matters far less.
Where using AI quietly becomes the ceiling
Adoption is not the problem any more. The tools are in the building. What has not moved is the shape of the work, and training has run a long way ahead of the systems it depends on.
The evidence
In professional and business services, 43.4 per cent of firms reported using AI in December 2025, up from 31.4 per cent a year earlier. Sixty-four per cent of employees there have AI provided by their employer, against 38 per cent nationally. Yet three-quarters of those same firms are not ready on data, orchestration and monitoring, and 70 per cent report limited progress on redesigning how work flows. Source: AI adoption plan, professional and business services, 2026.
This is the pattern I see most often in scaling businesses, and it long predates AI. Activity increases faster than clarity. Everyone is busier, the output looks healthier, and nothing structural has changed underneath it. AI has simply made the activity cheaper, so the gap opens faster.
I saw a sharper version of this recently in an acquisition context. The acquirer was tech-forward, the businesses it had bought were traditional and founder-led, and for the first year or so integration genuinely was the marketing function. Not campaigns. Not brand. Getting one set of systems, one data layer and one set of definitions so that anything downstream could be trusted. Marketing teams reaching for AI before they have done that work are automating a mess and speeding it up.
Here is the honest way to test where you sit.
Where marketing usually sits on the using-versus-integrating line
| Marketing activity | Using AI looks like | Integrating AI looks like | What actually changes |
|---|---|---|---|
| Content production | Individuals drafting posts in ChatGPT, quality varies by who did it | One documented route from brief to draft to expert sign-off, with a named owner at each gate | Volume stops being the measure. Citations and enquiries become it |
| Search and AI visibility | Asking a chatbot for keyword ideas | Tracking whether AI engines cite you, on which buying questions, and feeding that back into the plan | You can see where you are invisible at the moment buyers decide |
| Reporting | Pasting last month’s numbers in and asking for a summary | A data layer clean enough that the summary can be trusted without someone re-checking every figure | Board reporting stops being a two-day job |
| Lead handling | Drafting replies faster | Enrichment, routing and follow-up running without anyone remembering to do it | The leaky bucket stops leaking |
| Creative and brand | Generating images and variants | A written line between what AI drafts and what only a person signs off | Brand risk is bounded rather than hoped about |
Most marketing functions I look at sit solidly in the middle column and have never seen the right-hand one. That is fine, as long as you know it. The problem is that the activity level suggests a later stage than the business is actually at, and budget decisions get made off the wrong reading.
The pilot trap has a marketing version
Chris Dungey’s manufacturing plan carries the most transferable framework of the six. Scan, Pilot, Scale. Work out where AI would genuinely pay, test it somewhere that can absorb a failure, then prove it at operating scale. Sitting underneath that is the idea of lighthouse sites, a handful of working factories other firms can visit, so that the reference point is something running rather than a slide.
Marketing has exactly the same trap and rarely names it. The marketing version looks like eleven tools on the company card, four half-finished automations, one person who understands the reporting, and nothing that survives them going on holiday.
The useful bit to borrow is the sequencing, and specifically the Scan step, because that is the one marketing teams skip. Scanning means working out where AI would move a commercial number before anyone opens a tool. Not where it would feel modern. Where it would move the dial.
If you could only integrate one workflow properly this quarter, which one would it be? For most businesses I work with the honest answer is not content. It is the follow-up.
Augmentation first is a risk position, not a soft one
The creative industries plan takes a calmer line than the headlines do. Its stated principle is that “AI should support human creativity, not displace it”. Aardman, for example, uses a directed tool to clean up problematic frames in post-production, which now handles 70 to 80 per cent of that one job and frees skilled people for work that needs them.
It would be easy to file augmentation-first under ethics and move on. That would be a mistake. For anyone doing marketing right now it is a straightforwardly commercial position, for two reasons.
The first is trust, and the plan measures it. The sector using AI most is also the sector most careful about it. Call that squeamishness if you like. I would call it people who have learned what happens when something goes out with their name on it.
The evidence
Creative businesses are the most active adopters in the economy, with 51 per cent using AI against 33 per cent of all businesses (ONS, cited in the plan). Adoption inside the sector ranges from 60 per cent in IT and software down to 22 per cent in music and the performing arts. They are also nearly twice as likely to name doubts about AI product safety and transparency as a barrier, 10.1 per cent against 5.7 per cent economy-wide. Source: AI adoption plan, creative industries, 2026.
The second is visibility. Search now runs on whether a machine will cite you and a buyer will trust what it finds. Both reward the things AI cannot manufacture on your behalf: real experience, a named author who exists, a specific number from actual work, a position someone is prepared to put their name to. So fully automated content carries a second cost on top of the brand risk. It is built for precisely what the engines are learning to discount.
The most useful working rule I have found came out of a regulated engagement, building a client-facing calculator where a wrong figure carried legal exposure. AI did the heavy lift on structure, logic and first-pass numbers from the official scheme documents. A specialist on the client side verified every figure before anything went live, and the gate was named openly in the workflow rather than assumed. That separation, generation on one side and verification on the other, is the whole thing. When the two merge, AI produces something plausible and wrong, it ships, and the client spots it after publication. You do not recover that in a quarter.
Augmentation first reads like the slower way to work. In practice it is the version that does not blow up.
Automate the junior work and you stop making marketers
This is the part of the plans I keep coming back to, and it is the one no marketing team is discussing out loud.
Katie Gallagher’s plan flags a sharp fall in young people entering the digital workforce, and is careful not to overclaim the cause. Its proposed answer is an industry-led Early Careers Jobs Alliance, built on the principle that firms should redesign early-career roles rather than retreat from them.
The evidence
In 2024, UK digital sector employment fell for the first time in a decade, and the number of 16 to 24-year-olds in computer programming dropped 44 per cent in a single year. The plan states directly that while this is suggestive, there is no conclusive evidence AI is the cause. Worth holding both halves of that. Source: AI adoption plan, digital and technologies, 2026.
Marketing has its own version of that first rung. The meta descriptions. The first draft of the case study. Pulling the monthly numbers together. Writing forty social captions. It is unglamorous work, and it is precisely what AI does well, so it is the first thing to go.
It is also how marketers are made. Nobody develops judgement about what makes a good case study by reading about case studies. They develop it by writing thirty bad ones and having someone tell them why. Take that away without replacing it and you have a team of people who can operate the tools and cannot tell when the output is wrong. Which is the exact shortage every one of these six plans complains about.
I am not arguing for keeping busywork alive out of sentiment. The plans do not either. The question is whether the junior role gets redesigned or quietly deleted, and that is a decision someone makes, or fails to make, this year.
The hire is not an AI person
The life sciences and digital plans keep circling the same shortage, people who combine real subject knowledge with enough understanding of the technology to put it to work. The professional services numbers say the same thing from the other direction. Roughly four times as many roles change shape as disappear. The reshaping is the story, and it is the harder management problem by some distance.
The evidence
Life sciences alone points to a need for more than 100,000 new and replacement workers, which reads more like a pressure valve than a threat. In professional and business services, around 13.7 per cent of roles are at risk of direct substitution while 52.8 per cent are likely to be significantly augmented. Sources: life sciences and professional and business services plans, 2026.
Applied to marketing, this is the hiring mistake I see most. Businesses go looking for an AI person. What they actually need is a marketer who knows the commercial context well enough to brief a model properly and, more importantly, to spot when what comes back is confidently wrong. A brilliant technologist with no feel for your customers will produce fast, plausible, useless work. Someone who knows the business and can hold a model to account is worth considerably more.
Where that combination does not exist internally, the pattern that works is a pair rather than a unicorn. Someone owning capacity and implementation, someone owning growth and performance, with the handoff between them named out loud rather than left to good intentions. I have seen that shape hold up across several partner relationships now, and it consistently beats trying to hire one person who can do both.
What to do with this in the next 90 days
None of this needs a restructure. It needs one honest hour and then one decision.
Weeks one to two. Put an hour in the diary with whoever owns marketing and go through the plan for the sector you sell into, together. Not to extract a strategy. To hear the language your buyers are about to start using.
Weeks three to four. Run the using-versus-integrating table above over your own marketing activity, and be unkind about it. For each row, ask whether it would still work if the person who currently holds it in their head left tomorrow. If the answer is no, you are using, not integrating.
Weeks five to twelve. Pick one workflow. One. Scan it properly, so you know which commercial number it moves. Pilot it small. Then scale it, which mostly means writing down how it works and giving it an owner. One workflow properly integrated is worth more than five in permanent pilot.
And somewhere in the first hour, ask the question most leadership teams avoid: who here is quietly worried about what this means for their job? Every one of the plans lands on the same point about culture, that where people feel the change is being done to them, the change slows down. More than half of professional services employees fear being replaced, 11 percentage points above the national average. That anxiety does not show up as resistance. It shows up as things mysteriously taking longer.
The businesses that pull ahead here will be the ones willing to have the awkward conversation and then change one process properly. The tools were always the easy part.
Questions I get asked about this
Do these plans matter if my business is not in one of the six sectors?
Yes, for two reasons. Your buyers are likely to sit in one of them, and the barriers described are near-identical across all six. Unclear use cases, a shortage of people who can judge AI output, data that is not ready, and projects that never leave pilot. If your business has AI plans that have stalled, the reasons are almost certainly in these documents.
Which plan should I read if I run a professional services firm?
Shaheen Sayed’s Professional and Business Services plan. It is the most directly useful for accountancy, legal, consultancy and advisory businesses, and it carries the clearest data on where the workforce is against where the systems are.
Is there a financial services plan?
Not yet. Six were published on 8 June 2026 and the financial services plan is still expected. Worth watching if you sell into that sector.
Does augmentation-first mean we should stop using AI for content?
No. It means separating the generative role from the verification role and being explicit about which is which. AI drafting a first pass that a named person then verifies and signs off is a well-run process. Publishing straight from the model is not a content strategy, it is an unmanaged risk.
How do I know whether we are integrating or just using?
The test I use is departure. If a workflow stops working the day the person holding it goes on holiday, it is not integrated, it is being carried. Integration means it is written down, owned, measured, and survives a change of personnel.
Where this leaves you
The plans are more useful than most government documents, mainly because independent people from the sectors wrote them. Reading one is the easy half. The value lands in whether you use it to open the conversation you have been putting off.
If content, authority and AI visibility are the part of this you want to get right first, the E-E-A-T Website Scorecard will tell you in a few minutes where your site currently stands on the signals that both search engines and AI tools use to decide whether to trust you. It is the same diagnostic I run at the start of most engagements.
About the author
I am Jonny Ross, a Fractional CMO working with scaling SMEs and PE-backed businesses, mostly across Yorkshire and the North of England. I sit on leadership teams as the marketing seat, usually two days a month, and I have spent 27 years in search and visibility work. I co-host the 90 Day Website Mastery podcast with Pascal Fintoni. More about me | Fractional CMO
Read next
