Marketers, stop using AI*

More than a decade ago, I started my career in marketing selling data for Experian. At the time we sold address validation data, Mosaic data (geo-demographic data that sorted every household into groups based on lifestyles, money and habits for those not familiar) and the then highly popular email validation data. Am I showing my age?

Back then I frequently wrote articles on how data would be democratised, on why we should protect what we share online, and on the importance of data quality.

There's no doubt we know the importance of data quality today. But we are still solving the same old problem more than a decade on. I can't think of a business I've walked into where there hasn't been some sort of data quality issue. Duplicate contacts, missing or disparate data, varying naming conventions. You name it, I've seen it. And I'm sure you have too.

But this isn't an article about data quality. It's actually an article about AI. And why most businesses just aren't ready for it.

These very issues, the issues marketing teams have been battling with for years, are the very reason we're not ready. They are the issues that prevent businesses, and marketers, getting value from AI.

We know more than anyone that rubbish in gets rubbish out. And AI is only going to amplify this.

And data quality isn't the only problem.

Here are some other common challenges I've seen:

The master pitching deck the company wants AI to automate so your sales team can send personalised decks to that new enterprise account… it's only half-baked. Slides 10 and 15 need updating by the product team, slide 16 needs something on our ISO 27001 certification, and slide 20 has the wrong contact details.

The brand guidelines the agency needs for AI to create LinkedIn ads… that's last year's version. The same guidelines you wanted to update but never had time to because you had to chase pipeline in Q1.

The tone of voice document you need for a nurture flow… it's never been written.

The automated Lead > SQL process sales want AI to handle… together you've never defined what good looks like, or the hand-off process, or who does what, or even what an SQL is.

This is a very real problem for most organisations. The information, documents or processes still live in someone else's head, or simply haven't been created yet.

And this, my friends, is a human problem. Sure, AI can help, but you still need to fix it.

Everyone is losing their heads trying to automate "stuff". Regardless of how willing you are to adopt AI, most companies simply aren't ready.

So my plea to you all today is something people wouldn't expect me to say.

Please stop using AI*.

*For now.

Which is probably a huge relief to hear as, honestly, you're busy enough. The marketing role has proliferated, you're expected to do more and more, companies are hiring less, and AI is just another task on your plate that you don't have time for.

But I'm afraid we're not going to get off that easily. There's still work to be done.

Because as marketers we're resilient, adaptable to change. Most of us have taken the bull by the horns. Our answer to the AI challenge has been to upskill, do a course, and then apply that knowledge and build.

But this is actually the wrong approach.

I don't think the businesses that get left behind are the ones who are actually IN these platforms. I think it's the ones that are so in the weeds building that they are doing it blind: without a plan, without the right foundations, documents and systems to make AI successful.

And I'm seeing this result in a lot of wasted time building, creating things that only kinda-work, that require a lot of time tweaking and directing, and I've even seen employees build the same thing. 

That's an expensive use of marketing time. Time that none of us have.

I believe in a centralised approach and, more than that, a plan!!

Marketers, we love a plan!

Most advice tells you to think about all the repeatable tasks you do and hand them to AI. I'm telling you: don't do that.

I want you to down tools and create an AI plan.

Your marketing team

Start with your vision for your marketing team, humans and agentic. Structure it the way you would if your CEO asked what you want the marketing team to look like next year.

What do the roles look like? What do you need them to do? Who works with who? And, like any hire, what do those "people" need to be successful? What tools do you need?

A graphic designer can't be successful without a clear brand and tone of voice. A marketing operations manager isn't successful without a CRM, and accurate data.

Personally I like to plan out a human team first and then work backwards, asking myself what could be done by AI. This forces you to think clearly about what you really need and how you would set them up to succeed.

Jobs of work that need to be done

Think about the tasks you need to do today that will let you use AI effectively in three months' time. 

Create your list of "jobs of work" that need to be done to support your marketing team (human or otherwise). The unglamorous tasks. Data cleansing and de-duplication. A document that clearly maps how you use every field in your CRM and what each one means. Shared definitions for sales and marketing data. Alignment on what you'll report on and when. Brand guidelines and tone of voice. Your value proposition. All of this will literally be the key that lets you unlock AI, and make it effective.

Ironically, AI can help you do almost all of these. But every one of them still needs your unique perspective, and that's exactly where we can offer the most value.

Your AI roadmap

Evaluate which of these jobs AI can actually help with and prioritise these by impact and importance. 

Then for each job map the exact process you want AI to follow, what tools it should use and when, how humans will be in the loop to provide feedback, approve items or push things live. 

And lastly, the bit most miss, how you will know if AI has been successful.

Map your AI roadmap to business outcomes

Like an employee, you need to map your AI roadmap to business outcomes, not time saved. This could be outcomes like pipeline generated, conversion rates or revenue.

For example, if the AI agent is automating LinkedIn advertising then your success metric could be ROI, conversion and revenue generated. If it's building a nurture flow and persona-based emails, it's click-through rate.

It's also okay to have output-based outcomes. For example, if the project is data cleansing, then the metric of success is whether that project has been completed. 

Similarly, you can also measure an AI project based on velocity: has it enabled us to deliver more (although be careful of this one, more isn't always good!).

What matters is that you're treating your AI plan like you would a human employee. And their output is tied to SMART objectives and tangible KPIs, not just time saved.

None of this is as exciting as building. I know. But this is the work that actually makes AI work.

So, if I can encourage you to do one thing this week, it's this. Stop using AI, get out of Claude, Gemini, GPT, Copilot, whatever you're using. Building more "stuff" is not the answer. Get your house in order first.