AI Won’t Fix Your Marketing Function

Photo by Cookie the Pom on Unsplash

Before you hire another marketer, add another platform, or use ChatGPT to write 30 LinkedIn posts that’ll fade into the feed, it’s worth looking at the thing underneath.

A lot of founder-led businesses are looking at AI with tense optimism. There’s a hope that it might make marketing less demanding and expensive. Which it absolutely can (but there’s a caveat).

A decent AI workflow can speed up research, shape rough thinking into usable drafts, make reporting easier to interpret and reduce some of the dead manual labour that sits around your marketing operations.

But it only helps properly when the function it sits inside is clear enough to absorb it.

The caveat(s)

We get to feel busy: we’re producing a shedload of content, doing research in minutes that used to take hours, and generating visual content at the click of a mouse.

But.

If marketing is already spread across a few people (in house people, agencies, a founder, a couple of freelancers), AI tends to magnify the sprawl.

Because fundamentally, AI just magnifies and multiplies what you’re already doing. If your marketing strategy and operations are a rats’ nest of processes, AI will make it them more convoluted, not clearer.

So what?

At risk of repeating today’s common advice about AI, you need to start with clarity, then feed that to AI to increase your speed and efficiency. Ask:

• What are we actually trying to make happen?
• Who decides what matters?
• Why are we doing this channel again?
• Is any of this changing the quality of enquiries, or are we just producing better-formatted crap?

That is usually the point where the AI conversation gets less exciting and more useful.

Start with the job, not the tool

The worst place to start is usually: “What should we use AI for?”

It sounds practical, but it sends you straight into tool-fiddling and time-wasting — layering shit on top of shit. Before long, someone is comparing platforms, someone else is saving prompts in a Google Doc, and another person is suggesting an automation that nobody will maintain after the first fortnight.

A better question is:

What do we need, to make marketing easier and more effective?

That could mean better leads, more consistent content, clearer reporting, less founder involvement, better follow-up, faster repurposing, stronger sales material, or less time spent recreating the same work in slightly different outfits, for example.

Those are not the same problem. They should not all lead to the same tool or workflow.

A useful 90-day question is:

If AI genuinely helped our marketing, what would be different by the end of the next quarter?

Not “we’d be using AI more”. That’s self-congratulatory and essentially useless.

• What would change in the way work actually happens?
• Would the newsletter go out consistently?
• Would sales calls have better follow-up and ultimately conversions?
• Would reports make decisions clearer?
• Would founder thinking become usable content without the founder writing everything from scratch?

Define the job before choosing the machinery.

Look for the drag

AI is most useful where there is repeatable drag.

By drag, I mean the work that happens again and again, takes more attention than it deserves, and still needs enough judgement that you can’t simply automate it blindly.

Look at where marketing slows down. Is it ideas, briefing, drafting, approval, publishing, follow-up, reporting, or deciding what to do next?

Look at where you keep starting from scratch. Briefs, newsletters, case studies, campaign plans, sales follow-up and reporting are common culprits.

Look at where humans are adding judgement, and where they are just moving sludge around.

Keep the judgement, and reduce the sludge.

This is often where AI earns its keep. Not by inventing more marketing from nowhere, but by helping a business use what it already knows or owns.

Founder notes, sales calls, client questions, workshop material, proposals, delivery frameworks, old articles, testimonials. Most businesses have far more raw material than they realise. It just sits in odd corners, growing fluff.

Decide what stays human

There’s a reductive version of the AI conversation that treats humans as inconvenient meat obstacles between the prompt and the output.

This is a poor way to build trust.

Some work should stay human because it involves judgement, taste, ethics, relationship, commercial nuance or consequences.

AI can produce a first pass at messaging, but it shouldn’t decide what your business stands for. It can summarise customer feedback, but it cannot feel the small twitch that tells an experienced founder, “That’s technically correct, but not how our buyers think.”

Before you integrate AI properly, draw the boundary.

Ask:

• Where can AI reduce effort without lowering judgement?
• Where do we need a human to make the final call?
• What would be embarrassing, risky or damaging if it went out without proper review?
• What kind of work needs our actual voice, experience or commercial instinct?

This matters especially in founder-led businesses. The founder often carries the tone of the business before anyone has written it down properly. If AI sands that down too early, the marketing may become smoother, and more consistent. But also far less alive at the same time. Very tidy, and very dead.

Start with boring use cases

I would not start with the most dramatic AI use cases. I would start with the boring ones.

Boring is where businesses usually save time, reduce mistakes and stop squirreling time away on their own operations.

• The first place I’d look is turning raw thinking into usable content. Record a voice note; use a call transcript; aggregate themes from client questions. Feed AI real material, not a bland prompt asking it to “write thought leadership”. Then use a human to decide what’s worth keeping.
• The second is repurposing. A decent article can become an email, a few posts, a short video outline or a sales enablement note. The trick is not to flatten everything into beige content paste. Start with a real argument or a rough draft, then use AI to reshape it.
• The third is better briefs. Bad briefs create wasted work. AI can help turn a loose idea into a sharper instruction by identifying missing context, clarifying the audience, pulling out likely objections and making the purpose of the work explicit.
• The fourth is reporting. A lot of reporting is theatre: numbers arrive, everyone looks serious, a few things are up, a few things are down, someone says “interesting”, and nothing changes. AI can help summarise patterns and produce clearer commentary, but only if the report is attached to decisions. What should we keep doing? What should stop? What needs more evidence? What are we now better able to decide?
• The fifth is documenting workflows. Take the messy thing that currently lives in someone’s head and get it down. Use AI to turn the rough explanation into a process, refine it, then repeat. Deeply unsexy: pretty effective.

Before you add more, look underneath

AI is already changing marketing.

For founder-led businesses, the biggest gain may not come from doing everything faster. It may come from finally seeing the machinery: the half-decisions, the supplier sprawl, the reporting gaps, the founder approvals, the content bottlenecks, the places where marketing activity exists but the function underneath it is too fragile to compound.

Before you hire another marketer, add another platform, increase spend, brief another agency, or produce another hundred pieces of content, ask:

Is our marketing function actually designed to use AI well?

If the answer is yes, move faster with more confidence.

If the answer is no, start there.

What is marketing meant to achieve? Where is the work getting stuck? What needs human judgement? What can be made repeatable? What should stop? Who owns the rhythm? How does learning feed back into decisions?

Answer those questions, and AI becomes useful.

Skip them, and you may just build a faster mess.

I’m new to Medium, and I’ll be writing on marketing ops, workflows, AI…all that unsexy systems stuff (that profitable businesses can’t do without).

Do give me a follow if you might find that helpful.