AI Marketing: Why Everyone Adopted It and Almost No One Gained

AI marketing is the use of AI systems to do the work of marketing: researching a market, deciding who to talk to, producing the content, personalising the message, and measuring what any of it did. That definition is deliberately broad, because the interesting question is no longer which tasks AI can do. It is why a capability now held by almost everyone has produced an advantage for almost no one.

The adoption numbers are settled. 87% of marketers use generative AI in at least one recurring workflow, up from 51% in early 2024 (Salesforce, State of Marketing 2026). Among those using it, 77% use it for creative development, the single most common application (Gartner, CMO Spend Survey 2025). The average marketer saves around six hours a week (HubSpot, AI Trends 2026). Those are real gains, and they are also the problem: a gain your competitor also banked is not an advantage. It is the new floor.

Adoption is not advantage

There is a pattern that repeats with every general-purpose technology, and marketing is living through it now. Early on, access is the advantage: having the thing others lack. Then access becomes universal, and the advantage moves to whatever access cannot buy.

Marketing crossed that line somewhere in 2025. When 87% of your market can generate a competent blog post, a competent blog post is worth roughly what a competent blog post costs, which is now close to nothing. The six hours a week you save are real, but so are your competitor’s six hours. The question that decides the outcome is what you do with the capacity, and most firms answer it by producing more of the same thing faster.

This is why “we use AI in marketing” has stopped being a meaningful statement. Everyone does. What separates firms now is whether the AI runs inside a system that compounds or sits on top of one that does not.

The four places AI actually pays in marketing

Underneath the noise, AI does four genuinely different jobs in a marketing function, and they have very different returns.

Research and targeting. Finding which accounts are showing real buying signals, what a market actually complains about, and which segment is worth the spend. This is the highest-leverage and least glamorous use, because it decides where every downstream euro lands. Getting the target wrong makes every other efficiency irrelevant.

Creation. Drafting, editing, resizing, repurposing. This is where almost all the adoption went, and where the returns are most exposed to commoditisation, because it is the layer your competitors automated too.

Personalisation. Making a message fit a specific reader at a scale a human team cannot reach by hand. This is where AI does something humans genuinely could not do before, rather than doing faster what they always did.

Measurement. Working out which demand actually produced pipeline, including the sources no analytics package tracks. Unglamorous, and the reason the other three can be improved at all.

Notice that the layer with the most adoption, creation, is the one with the least durable advantage, and the layer with the most durable advantage, targeting, is the one most often skipped. That inversion explains a great deal of disappointing AI marketing spend.

What AI does not do, and why it decides the result

A model can produce the artifact. It cannot supply the four things that make the artifact worth reading.

It cannot supply proprietary data: what your clients actually said, what your last forty engagements taught you, which objection kills deals in your niche. It cannot supply lived experience, the specificity that comes from having done the work. It cannot hold a position, because it is trained to produce the consensus and a position is by definition a departure from it. And it cannot be accountable, which matters more than it sounds: a claim nobody is answerable for is a claim a serious buyer discounts.

Those four are exactly what a buyer is scanning for, which is why the market can be saturated with AI-assisted content and still short of content worth reading. The scarcity moved. It did not disappear.

Where the leverage actually compounds

The firms getting a real return are not the ones with the most AI. They are the ones whose AI sits on a foundation that gets better with use.

The foundation is context: your customer data, your positioning, the record of what has worked, the language your market actually uses. A model with your context produces something only you could have produced. The same model without it produces what anyone could have produced, which is the sludge everyone is complaining about. This is the same argument that runs underneath an owned AI operating model rather than a drawer of subscriptions, and it applies with particular force to marketing, because marketing is where the output is public and the sameness is visible.

Nothing about this is unique to content. It is the same reason precise targeting lowers acquisition cost more than any bidding change, and the same reason demand generation decides who gets shortlisted long before a sales conversation starts. AI is leverage on the system. It is not a substitute for having one.

How to sequence it without wasting a year

Most firms sequence this backwards: buy the tools, generate the content, then wonder why the numbers did not move.

Start with the target, because it multiplies or wastes everything downstream. Then the context layer, so the output is yours rather than generic. Then one workflow end to end, not eleven tools in parallel, which is the pattern that separates the firms getting a return from the ones producing volume. Then measurement, so you can tell which part worked. Creation comes last, not because it does not matter, but because creation on top of a wrong target and no context is exactly how a firm generates a great deal of content and no pipeline.

If that order sounds slow, compare it against the alternative that most of the market is running: a year of faster output, a larger tool bill, and a market that cannot tell you apart from the firm down the road.

The useful diagnostic is uncomfortable and quick. Take the last ten things your marketing published and ask whether a competitor with the same tools could have published them. If the honest answer is yes, the constraint was never the AI.

If marketing still runs on tools nobody connected, tell Daniel what you use today, what you have already built, and what marketing needs to do for the business. Book 30 minutes with Daniel.

Frequently Asked Questions

What is AI marketing?

AI marketing is the use of AI systems across the marketing function: researching a market, targeting the right accounts, producing content, personalising messages at scale, and measuring what produced pipeline. It is not a single tool or channel. In practice the term covers four distinct jobs, research and targeting, creation, personalisation, and measurement, which have very different returns.

Does AI actually improve marketing results?

It improves efficiency reliably and results conditionally. 87% of marketers now use generative AI in at least one recurring workflow (Salesforce, State of Marketing 2026) and the average marketer saves around six hours a week (HubSpot, AI Trends 2026). But efficiency your competitors also gained is not an advantage. Results move when AI runs on precise targeting and your own context, rather than producing generic output faster.

Where should a company start with AI in marketing?

Start with targeting, then the context layer, then one workflow end to end, then measurement, and leave content creation until last. Most firms invert this, buying creation tools first, which produces more output aimed at the wrong people with no way to tell what worked. Targeting decides where every downstream euro lands, so getting it right multiplies everything after it.

Why does AI-assisted marketing content all sound the same?

Because a model without your context produces the consensus, and the consensus is what every competitor’s model produces too. Models cannot supply proprietary data, lived experience, a genuine position, or accountability, and those are precisely what a serious buyer scans for. The sameness is not a flaw in the tools; it is what you get when the tools run without the inputs only you have.

Will AI replace marketers?

It is replacing specific tasks, not the function. The work that commoditises fastest is drafting and production, which is where most adoption went. The work that does not commoditise is deciding what the company stands for, which market to pursue, and which trade-off to make, along with accountability for the result. Teams that shift their hours from production to judgment gain; teams that only produce faster do not.

How much of marketing can realistically be automated?

More of the volume than most firms expect, and less of the judgment than vendors imply. Research, personalisation at scale, first drafts, repurposing, follow-up, and reporting all automate well. Positioning, the decision about which market to serve, the genuine point of view, and the relationships that close large deals do not. The practical answer is to automate the repetitive layer and spend the recovered hours on the layer that decides the outcome.