AI can now produce a marketing plan before the kettle has boiled.

This is impressive. It is also a small menace.

Ask for a plan and you will get one. Usually with audience segments, campaign ideas, channel recommendations, content pillars, a quarterly roadmap and a few phrases like "drive engagement" thrown in for the sake of sounding more marketing.

It will look like a plan. It may even feel like progress, but if the inputs are vague, the output will be vague. Faster, shinier and more confident for sure, but still vague.

That is where a lot of AI-assisted marketing planning starts to go wrong. Not because AI is useless. It is not. Used well, it can be a brilliant planning partner, challenger, organiser and accelerant. The problem is that many teams are asking it to solve a problem they have not clearly explained to themselves.

The prompt is rarely the real issue

There is plenty of advice about better prompts. Some of it is useful. Some of it reads like someone trying to summon a demon from marketing hell.

"Act as a world-class growth strategist..."

"Create a comprehensive omnichannel plan..."

"Use a tone that is friendly, authoritative, disruptive and human..."

A better prompt will only get you so far if the thinking underneath is thin. The real work is not learning how to flatter the machine. The real work is deciding what you need it to think with. Before you ask AI for a marketing plan, you need to be clear on five things.

1. What are you actually trying to change?

"Build awareness" is not enough.

Awareness of what? With whom? For what commercial purpose?

Are you trying to increase qualified enquiries? Improve conversion from existing demand? Reposition a tired service? Launch into a new market? Reduce dependency on paid leads? Help sales teams explain the proposition properly? Stop good enquiries disappearing into a CRM-shaped cupboard?

These are all different jobs. AI will not know which job matters unless you tell it. A useful marketing plan starts with a commercial problem, not a content calendar.

Try this instead:

We need to increase qualified enquiries for [service] from [audience] because [commercial issue]. The current problem is not total demand, but [conversion / clarity / follow-up / trust / differentiation / channel dependency]. Build the plan around that.

Already, the output will improve. Not magically, but meaningfully.

2. Who is the customer, in human language?

Most AI-generated plans are full of audience descriptions that sound plausible and help nobody.

"Busy professionals seeking convenience."

"Decision-makers looking for trusted solutions."

"Homeowners interested in expert guidance."

These are not audiences. They are LinkedIn mist.

You need sharper customer truth. What situation are they in? What are they worried about? What have they already tried? What would make them hesitate? What language do they actually use before someone from marketing gets hold of it and buffs it into beige?

For service-led businesses, this matters because the customer is often not just buying a product. They are trying to move through uncertainty.

A landlord may not be looking for "property management solutions". They are probably thinking, "I cannot keep dealing with this tenant issue myself."

A seller may not be looking for "award-winning estate agency expertise". They are probably thinking, "I need to know whether moving is realistic before I turn my life upside down."

A clinic enquiry may not be looking for "patient-centred care pathways". They are probably wondering whether someone will actually listen without making them feel daft.

A proptech buyer may not be looking for an "AI-powered property platform". They are probably thinking, "I need one place to see what is happening across the business without chasing five different spreadsheets."

Feed AI the real situation, not the polished version.

3. What do you really do?

This is where many marketing plans start wobbling. The business describes itself by category.

Estate agency. Proptech. Consultancy. Clinic. Multi-site service provider. Managed service. Customer experience platform.

Categories are useful. They help people place you, but they are not your proposition.

A proposition needs to explain the useful change you create.

Try this:

We help [specific customer] move from [messy/problem state] to [better/useful outcome] by fixing [specific system, decision, journey or capability].

For Break.Beat, that might sound like:

We help service-led businesses stop losing growth in messy marketing, CRM and customer journeys by turning scattered activity into clearer plans, useful systems and measurable rhythm.

That is much more useful than asking AI to "write a marketing plan for a consultancy". It has a direction; it knows what must be made clearer; it knows the plan cannot just be content and campaign; it must connect proposition, journey, follow-up, CRM and measurement.

Which is where the real commercial work usually lives.

4. Where does the customer journey leak?

This is the bit many AI marketing plans politely skip. They assume the job of marketing is to create more attention. Sometimes it is. Often, though, the more expensive problem is what happens after attention has already been earned.

Someone visits the website. Reads reviews. Checks the team. Looks at social proof. Compares three options. Fills in a form. Calls a branch. Books a consultation. Asks a question.

Then the system either builds trust or quietly drops the ball.

A marketing plan that ignores this is not really a plan. It is a traffic-generation document wearing a smarter jacket.

Before you ask AI for recommendations, give it the shape of the journey:

  • Where do enquiries come from?
  • What does the first response look like?
  • How quickly does the right person respond?
  • What is captured in the CRM?
  • What happens after the first attempt?
  • Where do people stall, drift or disappear?
  • What does the team need to say, send or record?

This does not mean AI should replace human follow-up. Absolutely not. It means your plan should include the points where marketing, sales, service and systems meet. That is often where the money leaks.

5. What rhythm will make the plan usable?

AI is very good at producing a lot of activity. Thirty content ideas. Twelve campaign themes. Six audience segments. Four channel plans. A majestic spreadsheet of doom. The issue is not whether the ideas are clever. The issue is whether your team can use them.

A practical plan needs rhythm:

  • What are we doing this month?
  • What are we testing?
  • What are we pausing?
  • What will we review weekly?
  • What decision will we make after 30 days?
  • What should managers, sales teams or branch teams do differently?
  • What evidence will tell us whether the plan is working?

This is where AI can be genuinely helpful. Ask it to turn the plan into a working rhythm, not just a document.

For example:

Turn this into a 90-day plan with weekly priorities, clear owners, review points, decision gates and measures. Include a "stop doing" list so we do not simply add more activity to an already noisy system.

That last line matters. Most businesses do not need more marketing noise; they need better signal.

A better way to brief AI

Here is a more useful structure.

Do not start with:

Write me a marketing plan.

Start with:

I want you to help build a practical marketing plan. Do not write the plan yet. First, review the information below, identify the gaps, challenge weak assumptions and ask the questions needed to make the plan useful.

Then give it:

Business context

What the business does, where it operates, who it serves, what is changing.

Commercial problem

What needs to improve: lead quality, conversion, retention, proposition clarity, response, trust, margin, channel dependency.

Customer situation

Who the customer is, what triggers a need, what they worry about, what they compare, what language they use.

Proposition

What you help them achieve, why that matters, what makes you credible, where you are meaningfully different.

Proof

Evidence, case studies, performance data, reviews, customer stories, experience.

Current journey

Website, forms, calls, CRM, first touch, follow-up, sales handoff, customer onboarding.

Current activity

Channels, campaigns, content, paid media, email, events, partnerships, referral activity.

Constraints

Budget, team capacity, systems, geography, compliance, seasonality, internal politics if you are feeling brave.

Measurement

What success looks like, what can be tracked, what is currently invisible.

Then ask AI to produce four things:

  1. The assumptions it thinks you are making.
  2. The gaps that need answering before planning.
  3. A focused 90-day plan.
  4. A weekly review rhythm.

That is a much better use of AI. Less magic trick. More planning partner.

Where human judgement still matters

AI can organise thinking. It can spot gaps. It can generate options. It can pressure-test a structure. It can help turn a messy conversation into a workable plan.

Good. Use it.

But do not outsource judgement.

AI does not know which trade-off your business is ready to make. It does not know which proposition will survive contact with your sales team. It does not know whether your branch managers will actually use the CRM field it has just recommended. It does not know which customer concern gets raised on every call but never appears in the marketing copy.

That is still your job. The useful role for AI is not to replace marketing judgement. It is to make weak assumptions easier to spot.

A simple test

Before you use AI to build your next marketing plan, score your inputs from 1 to 5:

  1. We know the commercial problem we are trying to solve.
  2. We can describe the customer's situation in their language.
  3. We have a clear proposition, not just a list of services.
  4. We understand the enquiry journey and where leads leak.
  5. We know what we will measure and how often we will review it.

If those scores are low, AI will not rescue the plan. It will just make the uncertainty more presentable. Sort the inputs first. Then use AI to sharpen, challenge, organise and accelerate the work.

That is where it becomes genuinely useful. Not as a substitute for thinking. As a way to stop useful thinking getting trapped in someone's head, someone else's spreadsheet and a strategy deck nobody opens after the launch meeting.

Break patterns. Build rhythm.