Three Mistakes Agency Owners Make When They “Adopt” AI

In 2023, most agency owners I know made a quiet decision about AI. They chose to wait.

Not to ignore it. Not to dismiss it. Just to wait a little longer. See what shakes out. Get through this quarter first.

I made that decision for about a month before I noticed what I was doing. The waiting felt like strategy. It was avoidance. And that distinction — moving with intention versus moving when forced — is where the AI story separates into two very different outcomes.

Here are the three places I watch it happen.

Mistake 1: Starting with tools instead of a diagnosis.

The sequence most agencies follow: hear about a tool, try the tool, maybe stick with it. The sequence that actually works is different. Map where the time and money are. Then find the tool that addresses the specific problem.

We spend the first two weeks of every agency engagement doing exactly this. Almost every time, we find that senior people — strategists, directors, account leads — are spending a meaningful slice of their week on work that doesn’t require their judgment. Assembling reports. Pulling data. Drafting first versions of things.

The tools that solve that problem are different from the tools that get demoed on social media. Starting with the demo leads to adoption without a diagnosis. That’s why usage plateaus at most agencies.

Mistake 2: Underestimating the change management problem.

About eight months into rebuilding Newfangled, I sat in a client meeting where one of our account managers presented the account’s history like she’d been there from the beginning.

She had been on the account three weeks.

She referenced a strategic decision the client made in 2022 and why. She talked about the client’s preferences and history the way you do when you actually know someone. She’d spent twenty minutes with our Client Knowledge Agent before the meeting.

That was the moment the change management problem became real to me — not as a process problem, but as a relationship quality problem. Getting there required more conversation than I budgeted for. The senior people who’d been running things well for years needed permission to be uncertain without it feeling like failure.

If you’re planning an AI implementation and haven’t budgeted for the conversations that will need to happen twice, you’re going to end up short.

Mistake 3: Treating AI as an add-on instead of a rebuild.

This is the framing error underneath the first two. Most agencies I talk to are asking “how do we add AI to what we do?” The question that produced results at Newfangled was different: “What would we do if we were starting today?”

Those two questions produce completely different answers.

The first produces a layer on top of your current process. The second produces a new process that uses AI where it actually helps. The first is faster to start and slower to compound. The second is harder to start and faster to compound.

Three years in, our team does more than a larger team did before. The work is better. The margins are better.

That’s what the second question produces.