Why training your marketing team beats hiring a data scientist

Every commercial leader I speak to in Dubai is being asked the same question by their board: what are we doing about AI? And the instinct, almost every time, is to go and hire someone technical. A data scientist. An AI lead. Someone whose CV has the word "machine learning" on it.

I understand the instinct. It feels like the responsible answer. It is also, on the available evidence, the least effective thing you can do with the money.

What the study actually found

Researchers looked at S&P 500 companies over a ten-year period and matched roughly 25 million job postings against each firm's return on assets. The question was simple: which talent strategy actually produces a financial return? They separated the workforce into three groups.

• -Pure specialists. People hired for deep technical skills - data science, machine learning - without a commercial background.
• -Specialists who learned the business. Technical hires who also picked up domain and commercial knowledge.
• -Business people who added AI skills. Existing commercial staff - marketers, salespeople, operators - who learned to use the technology.

The gap between them is not subtle.

Read that again with a commercial hat on. The same investment, directed at the people who already understand your customers and your margins, returned nearly ten times what it returned when directed at technical hiring.

Why the gap is so wide

It isn't that technical people are less capable. It's that in a commercial function, the hard part was never the technology.

The hard part is knowing which problem is worth solving. Which campaign is underperforming and why. Which accounts are quietly being under-served. Where the margin is actually leaking. Which of the fourteen things on the plan will move the number and which eleven are theatre.

That knowledge lives in your marketing manager's head. It does not live in a job description. A data scientist arriving on Monday has to spend six months acquiring what your team already knows - and in most organisations they never quite get there, because they sit outside the commercial rhythm and are handed problems rather than finding them.

Meanwhile the marketer who has been taught to use these tools properly can move immediately, because they already know where to point them. They know the brief is the bottleneck. They know reporting eats the first three days of every month. They know which client asks for four variants of everything.

What this looks like in a marketing function

The work that compresses first is production, and production is where most marketing budget quietly goes. Brief to first draft. Variants and resizes. Localisation. Monthly reporting decks. The long tail of content that never justified a proper brief but got made anyway.

None of that requires a model to be trained. It requires someone who knows what good looks like, working with tools they have been shown how to use, inside workflows that have been rebuilt around the fact that a first draft now costs almost nothing.

Which is a very different project from hiring a data scientist. It is faster, it is cheaper, and it does not depend on one person staying.

The counter-argument, fairly put

There is a real case for technical hires, and I don't want to dismiss it. If you are building a proprietary model on your own data, if you have a genuine data-engineering problem, or if AI is going into the product itself rather than the commercial function, you need people who can build. The study measures the average across large firms - it does not say technical talent is worthless, it says that for most companies most of the time, the return sits elsewhere.

And there is a sequencing point too. Training a commercial team without leadership having decided what it is for produces enthusiasm and very little else. That is a large part of why 95% of organisations report no return at all on their generative AI investment despite tens of billions spent. The tools were bought. The direction was never set.

What I'd do instead

• 01Establish where the commercial function is slow, expensive or under-covered today - with numbers, before anyone is trained on anything.
• 02Get the leadership team to a decision about what is being funded and protected. Nothing survives without that.
• 03Train the people who do the work, on their own workflows, not on a generic prompt course.
• 04Rebuild the two or three highest-cost workflows properly, and measure them against where you started.

If after all that you still need a technical hire, you will at least know precisely what for - which is a much better brief than the one most companies are recruiting against right now.