Build or buy? Buy, and here's why.

Marketers don't need to be AI engineers. They need to be AI-willing — and let a team designed for nothing else do the building.

October 02

4 min read

Every build-or-buy conversation eventually produces the same hero fantasy: the scrappy in-house team that reverse-engineers its own AI stack, saves a fortune, and outmaneuvers every vendor in the category. It's a great story.

It is also, for almost every marketing team that tries it, not what actually happens.

The question isn't "can we build it?" It's "should we have to?"

Ask a marketer to write a campaign brief, and they'll hand you something sharp in twenty minutes. Ask them to architect an agentic workflow, evaluate model outputs against a governance framework, and keep the whole thing running after the one engineer who understood it takes another job — and you've just changed their job description without telling them.

Marketers aren't engineers. They shouldn't have to become one to do great marketing.

The math doesn't work in your favor

Even if your team wanted to build, the numbers argue against it. A realistic year-one cost for an in-house AI build — senior AI engineer, full-stack support, DevOps — runs $700,000 to $900,000 before it produces a single campaign.¹ Compare that to a specialist platform: one modeled ROI comparison put buying at a 93% year-one return, against a build option that started underwater and didn't turn a profit until year three.²

Buying wins year one: modeled ROI for a marketing AI workflow, buy vs. build. Buying reaches 93% year-one ROI; building starts underwater and turns profitable only after year three. Source: JustThink, Build vs Buy for Enterprise AI in 2026: A Decision Framework.

And that's before you try to hire for it. The AI talent market has roughly 1.6 million open roles chasing 518,000 qualified candidates globally — about three open jobs for every qualified person — and AI engineers now command a wage premium north of 56% over comparable non-AI roles.³

ManpowerGroup's 2026 survey of more than 39,000 employers across 41 countries found AI skills are now the hardest in the world to hire for — harder than engineering, harder than IT, harder than skilled trades.³ You're not just competing with other marketing teams for that hire. You're competing with every company on earth.

Build it yourself anyway, and the odds still aren't friendly: Gartner predicts more than 40% of agentic AI projects will be canceled by the end of 2027, largely by teams that took on complexity they weren't structured to support.⁴

AI-willing beats AI-expert

"Marketers don't need to build models or algorithms. They need to architect how AI systems operate."

That's marketing researcher Pierre DeBois, writing in CMSWire about the actual skill 2026 marketers need.⁵ Not model-building, but orchestration: defining the goal, the data, the guardrails, and the judgment call at the end. That's a fundamentally different, far more learnable skill than becoming an AI engineer, and it's the one marketers should actually invest in.

Call it what it is: you don't need to be AI-expert. You need to be AI-willing — curious enough to run the tool, disciplined enough to review its output, and clear enough about your goals to give it a real brief. Everything past that belongs to somebody whose entire job is the engineering underneath it.

Where the magic actually happens

This is the part the "build everything" instinct misses. A dedicated team — an entire org whose only job is making one category of AI marketing tooling excellent — will out-build your six-person in-house effort every time, not because they're smarter, but because they're structured for it.

They're solving your exact problem for hundreds of customers instead of one, which means every edge case your team would discover the hard way, they've already fixed. Compounding expertise is the actual moat. You can’t build it in-house on a marketing budget. But thankfully, you can buy access to it.