What Ad Tech Taught Me About AI That CS Grads Don't Know

I spent more than one late night staring at a conversion report that was confidently wrong. The attribution model was crediting the wrong touchpoints, the “winning” channel was winning because of a measurement artifact, and a real budget was about to move based on a lie. Untangling it meant reasoning about confounders, selection bias, and what the data could and couldn’t support.

Years later I learned the name for what I’d been doing on those nights. It’s data science. It just had a different job title and a worse coffee budget.

That’s the argument of this whole piece, and it’s the most important thing I can tell another career changer coming from marketing: your background is a moat, not a deficit. The myth that you have to “start over” at zero is not just demoralizing — it’s wrong, and acting on it makes you worse at the pivot.

The “start over” myth

The pivot advice aimed at marketers usually assumes you’re arriving empty-handed: learn to code, learn statistics, learn ML, then you’ll be employable. It treats a decade of commercial experience as irrelevant prologue.

But the thing that makes someone valuable on a data or AI team is rarely raw algorithm knowledge — that’s increasingly commoditized and, frankly, googleable. What’s scarce is judgment about which problem is worth solving and whether the answer is real. Marketing and ad tech train exactly that, under fire, against money. I came up through the analytics-heavy side of marketing, and the more I learn about data work, the more of it I recognize.

Five ad-tech skills that map almost one-to-one

Here’s the translation table I wish someone had handed me on day one:

What I did in ad tech	What it’s called in data/AI
A/B testing campaigns	Experimental design, hypothesis testing
Attribution modeling	Causal inference under messy real-world data
Audience segmentation	Clustering and feature thinking
Campaign dashboards	BI, metric design, “will anyone act on this?”
ROI / budget defense	Modeling business value, cost-aware decisions

None of these are loose analogies. Designing a clean A/B test — worrying about sample size, run length, and what could confound the result — is the experimental method that underpins data science. Arguing for years about what truly drove a conversion is applied causal reasoning, the hardest and most valuable part of the field. Segmentation is the intuition behind a whole category of unsupervised learning.

And the last row is the one that makes hiring managers lean in. A lot of technically strong people cannot answer “what is this model worth to the business?” I can’t not ask it. That instinct is rare on technical teams and disproportionately prized.

Where the gaps genuinely are (no pretending)

A moat isn’t a force field. There are real gaps, and being honest about them is part of the credibility:

• Statistical rigor. I had working intuition for significance; I did not have the formal grounding. That gap is real and closeable.
• Code fundamentals. Writing SQL to answer a question is not the same as writing maintainable Python, using version control well, or reasoning about how code runs. This is the steepest part of my climb, and I’m not going to pretend otherwise — it’s part of why I think you should still learn to code in 2026.

Naming the gaps is what separates a confident career changer from a delusional one. Hiring managers can tell the difference instantly.

Reframe the résumé bullet (before → after)

Most marketers undersell themselves by describing the activity instead of the analytical work. The fix is to rewrite each bullet to surface the data thinking that was already there:

Marketing résumé bullet	Reframed as data work
”Ran A/B tests on email campaigns"	"Designed and analyzed controlled experiments; quantified lift with attention to sample size and significance"
"Managed audience segments for targeting"	"Built behavioral segments from first-party data to drive measurable downstream outcomes"
"Reported on campaign performance"	"Designed decision-oriented metrics and dashboards that changed budget allocation"
"Optimized spend across channels"	"Modeled marginal ROI to allocate a constrained budget under uncertainty”

Copy that pattern. You’re not inflating anything — you’re describing what you actually did in the language of the field you’re entering.

What honestly doesn’t transfer

Two things, so I’m not selling snake oil. First, deep ML engineering — the systems, the math, the production infrastructure — is genuinely new territory; experimentation intuition gives you a head start on thinking, not on building. Second, the pace and tooling of software work is its own culture you have to actually absorb, not just map onto marketing analogies. Respect that, and the transfer story stays credible.

The takeaway

If you’re coming from marketing or ad tech: stop apologizing for your background in cover letters. The experimentation reflex, the causal suspicion, the relentless “what’s this worth?” — those are the expensive, hard-to-teach parts of data and AI work, and you already have them. Learn the code and the stats to go with them, be honest about the gaps, and you’re not a late beginner. You’re a different, and often better, kind of candidate.

---

Next best read: How to Get Into AI Without a CS Degree — the objection-by-objection version of this argument — or the roles that reward this background directly.