In the last twelve months, “we need to be doing more with AI” became the default question in most marketing leadership conversations.
Most businesses have AI-literate marketing departments. Their teams are using AI across content, research, reporting, maybe custom GPTs/Projects for specific tasks.
The problem is often that all of this usage happens in pockets. There’s individual experimentation without a coherent approach to what gets used where, by whom, with what guardrails around data and quality.
I run a marketing consultancy that works with scaling businesses across the UK, and I use AI heavily in my own practice (including building an AI-powered product!).
I’m a big advocate for using AI in the right way. But I am also honest about where the gap sits between what vendors claim and what actually holds up when you’re trying to get a marketing function to perform under pressure.
The value of more is diminishing
Most leadership teams still see AI through the lens of volume: more content, more campaigns, more outreach, more activity. If AI reduces the cost of production, the logic goes, you should scale everything.
The problem is that everyone else has access to the same tools. AI has massively democratised the “doing”. A first draft that would have taken half a day now takes twenty minutes, but that is true for every competitor in your market, too.
The channels that depend on the recipient being able to distinguish signal from noise are also being degraded by the sheer abundance of AI-generated slop.
More can not be the answer, we need to ask how we can make marketing better – i.e. more closely aligned to your specific audience, grounded in what actually matters to them and distinct from everything else they’re seeing.
That is what wins when generic becomes the default.
Where AI adds leverage versus where it wastes time
Before getting into specifics, this is the principle that should guide every decision about where to apply AI in your marketing.
AI is excellent at doing things faster. It is not yet very reliable at deciding what to do.
| AI genuinely reduces time or improves quality | AI produces plausible output that still needs human judgement |
| Structuring briefs from raw inputs | Deciding which audience segment to prioritise |
| First drafts and content repurposing | Positioning and competitive strategy |
| Summarising reports and competitor content | Interpreting what research means for your specific situation |
| Spotting patterns in campaign data | Understanding the deeper why something is or isn’t working, especially if that reason is multi-faceted |
| Meeting summaries and agency briefs | Creative direction and brand voice |
| Adapting copy across formats and channels | Evaluating whether output is distinctive |
The left column is where a lean team should start. The right column is where over-reliance costs businesses, often without them realising it until the campaign has run and the money is spent.
Where AI genuinely helps a lean team right now
Briefing and structuring thinking
Converting a 90-minute sales call transcript into a structured creative brief. Pulling a customer interview into a clean set of notes. AI is genuinely good at converting unstructured input into structured output.
First drafts and repurposing
Repurposing a 2,000-word case study into five LinkedIn posts. Adapting one piece of pillar content into a newsletter, social posts and a sales-enablement one-pager. The time-to-first-draft compression is real and significant. But AI-generated copy without a human editing layer is identifiable and underwhelming. Editing is now the actual skill.
Research synthesis
Feeding ten industry reports into an AI tool and asking for the three themes that contradict each other. Summarising what twelve competitor blog posts have in common so you can deliberately sound nothing like them. AI is faster than a human at reading and summarising large volumes of text. It cannot tell you what the information means for your specific strategic situation, it will flatten nuance and revert to the statistically average interpretation.
Performance analysis
Pulling SEO or campaign data into an AI tool and asking “what changed in the last three months?” as a starting point, then bringing your own context to interpret it. AI surfaces patterns and anomalies faster than manual review. It cannot tell you why something happened. It also doesn’t know you migrated your site three months ago, unless you tell it.
Where AI does not yet replace human judgement
This is the section that matters most, and where the gap between vendor claims and practical reality is widest.
Strategy and positioning
AI can generate frameworks and options. It cannot tell you which positioning is actually true for your business, which customer segment is genuinely underserved, or which competitive move is defensible. The human brain can still hold a larger, more nuanced context window than any model – the accumulated understanding of your market, your customers, your competitive dynamics, your resources.
Audience insight
When you ask a model to simulate or describe your audience, it draws from a general corpus of training data, resulting in a flattening and averaging of information that carries structural bias from its sources. If your audience is specific, niche or represents a community underrepresented in those training sources, the model’s approximation of them is likely wrong. Genuine audience understanding comes from qualitative and qual research: talking to customers, sitting in sales calls, reading between the lines of how they describe their own problems. That understanding is now more valuable, relatively, than it was eighteen months ago, because a machine can not create net new information.
Creative direction
AI produces competent, average creative work at speed. For a business trying to differentiate in a crowded market, average is the worst possible outcome.
Creative distinctiveness requires judgement about what is surprising, relevant and true to a specific brand. Models regress to the mean; brands and businesses have to do the opposite.
What I’m ignoring right now
Anyone claiming they have fully automated their entire marketing workflow with AI agents is either overstating their case or running something remarkably fragile. These systems hallucinate, make mistakes, and are not production-ready for end-to-end autonomous operation.
There is also a cost argument. The price you pay today for AI compute is not the price you will pay tomorrow. Token costs are rising, platforms are restricting usage, and some companies have already begun rehiring positions they eliminated in the initial wave of AI enthusiasm. Building a fully automated marketing system on current pricing assumptions is a bet that may not pay off.
A 30-day starting point
If your team is not yet using AI systematically, three actions in the next month will build a coherent foundation rather than scattered experimentation.
Audit the jobs to be done, not the tools
Map your team’s recurring tasks against two axes: how much judgement they require, and how much time they consume. Your targets are the high-time, low-judgement tasks.
Pick one workflow and test drive it together
One. Content repurposing, or research synthesis, or meeting-to-action summaries. Track time saved and quality of output side by side. Resist the urge to add more tools or processes until one is genuinely working.
Set an editing standard/AI policy before you get started
Decide what the human-judgement layer looks like and who owns it before the volume of AI-assisted output grows.
The reinvestment question
Five in ten UK scaleups are currently leveraging AI, and nine in ten plan to adopt it. The risk now is building a two-tier business where the tools are moving faster than the approach or capability to leverage them correctly.
The most useful thing a marketing leader can do with AI right now is not adopt every new tool with reckless abandon. They should identify the highest-friction, lowest-judgement tasks in their team’s week and automate those specifically.
Then reinvest the time saved into the work AI cannot do: getting closer to your customer, sharpening your positioning, and making the decisions that actually determine whether your marketing works.
Everything else can wait.
If the gap between your marketing activity and your strategic clarity feels familiar, book a conversation with one of our senior partners or explore our AI Process Integration service.
