Efficiency Isn't Growth: What AI Transformation for Marketing Teams Actually Requires.

I've been making some version of this argument for a while now.

"AI gives you efficiency. Your data gives you direction. Real growth requires both."

It's been the thesis behind a few pieces I've written, and something I said almost word for word when The AI Journal featured my perspective on this.

This is the most concrete example I've had of it in practice.

Most AI transformation content for marketing teams sells the wrong finish line. Faster reporting. Faster content production. Fewer manual hand-offs. All real, all valuable, and none of it, on its own, is growth.

Here's what that distinction actually looked like in a recent engagement, and why I think skipping past it is the most common mistake in how AI transformation gets sold right now.

The Work Started With Documentation, Not Tools

Before recommending anything, I mapped how work was actually moving through the marketing team, plus the cross-functional teams and external vendors it touched along the way, where things stalled, where the same question got asked three times, where a deliverable sat waiting on someone. I talked through feasibility and difficulty with the people who'd actually have to build anything. Recommendations came out of that, not the other way around.

That conversation split into two kinds of findings. Some fixes were genuinely low-hanging fruit, clear, buildable, no real unknowns. Others weren't, not because they were bad ideas, but because we didn't actually know yet what platforms were in use, whether we'd have the access we'd need, or whether the tools involved could even talk to each other. Those got flagged as needing more investigation before anyone could commit to them, rather than getting quietly dropped or, worse, promised anyway.

What most marketing teams try to fix first is the workload itself, the everyday friction of getting things done. Email is a good example, and it's a familiar one almost anywhere. Producing a single campaign usually means starting from scratch: a creative brief, a round of reviews before anything gets built, then production, then QA, then figuring out targeting and segmentation before it ever ships. Multiply that by every campaign running at once, and most of a team's time goes into process, not strategy.

There's often a second, quieter friction point sitting right next to it. In this case, those emails were meant to drive people toward more content elsewhere, or into a direct conversation with someone on the team. But the process for getting that other content ready was its own manual, disconnected pipeline, one the email team had little visibility into until it was already live. Earlier visibility into what was coming would have let them plan around it instead of reacting to it.

That's usually where AI gets introduced first, and understandably so. It's the most visible pain.

What I Proposed, Specifically

A measurement framework, not a report. The team asked for something that would show what was actually working, not just what happened. Faster, clearer reporting was the eventual goal once that framework existed, but it wasn't the recommendation itself. You can't speed up a report that doesn't yet know what it's supposed to be measuring.

A custom AI assistant to review creative asset hand-offs before they went out, checking for missing files, wrong formats, inconsistent labeling, the exact things that were causing repeated back-and-forth and stalling projects for a day at a time.

A way to systematically pull and repurpose existing site content into email campaigns, and just as importantly, earlier visibility into what content was coming down that pipeline in the first place, so the email team could plan around it instead of scrambling once it went live.

Each one was scoped, feasibility-checked, and proposed as unglamorous, specific fixes to a named bottleneck. Not because AI was the point. Because the bottleneck was real.

Here's the part I think gets skipped, and it's the part that actually matters.

Every one of those recommendations saves time or cost. None of them, by themselves, produces growth. That's not a small distinction; it's the whole argument.

Here's the order that actually works, and why the last step isn't optional.

Step 1: Document the Real Friction

Map how work actually moves through the marketing team. Where it stalls, where the same question gets asked twice, where a deliverable sits waiting. Not what you assume is slow. What's actually slow.

Step 2: Fix Specific Bottlenecks With AI

Fewer hand-off delays. Repurposed content instead of rewritten content. This is what buys back time, not for more execution, but for actual strategy. Real, useful, and exactly where most AI transformation efforts stop.

Step 3: Build the Data Foundation and Measurement Framework

Agreement on what growth actually means. Clean enough data to trust what a pattern is telling you. A framework that tells you which metrics matter and which don't.

This is what actually makes the data useful, not just having it. Good data tells you what's working and what isn't. It shows you what customers are actually doing, not what you assumed they'd do. And it's the only thing that lets you form a real hypothesis about what to try next, then actually test whether it moved anything. Without it, you're not testing. You're just changing things and hoping.

Step 3 is not a nice-to-have that happens after the real work is done. It's the step that decides whether steps 1 and 2 were worth doing at all.

That's the real promise of Step 2: not content at scale, room to think again. But strategy without Step 3 is just a guess with better production value. In this case, nobody was actually measuring whether those emails were driving growth, meaning real sales, not just opens and clicks. So I was asked to design an ideal measurement framework for it, regardless of feasibility. I built one, end to end, the version that would actually answer the question if nothing else got in the way.

I remember presenting it and watching people nod, genuinely agree it was the right approach, and still not quite believe it would happen. That reaction is familiar to me now. Foundational work gets agreed with easily and funded reluctantly. The last update I had was that it was moving forward anyway.

Cutting reporting time in half means nothing if the team still can't agree on what number actually indicates growth, or if the underlying data isn't clean enough to trust the pattern AI just spotted. Efficiency creates capacity. On its own, it doesn't tell you where to point it.

AI doesn't replace Step 3. It raises the cost of skipping it.

A marketing team that runs Steps 1 and 2 without Step 3 just gets faster at not knowing what's working. The speed is real. The direction is still missing.

Skip Step 3, or treat it as optional, and you've automated your way to motion, not progress.