Working With AI: What a Year in the Trenches Taught Me

A year ago, I thought I understood AI because I had helped deploy it.

In my last executive role, I led the rollout of Intercom's Fin AI agent and cut support costs by 60%. I walked away from that project sure I had a handle on what the technology could do.

I was wrong.

Over the past year I've spent hundreds of hours using AI for real work with real stakes, not the keynote version, the daily version. What I found is that leading a corporate deployment and using the tool yourself are two different skills entirely. One is a business case. The other changes how you think.

I tested that by building a few things myself. An ATS Optimizer to match job descriptions against candidate resumes. A Supplier Relationship Management prototype I'd wanted for years but never had the engineering team to build, the kind of project that used to take years and a dedicated build team. This one took weeks.

Three things stand out from that year, and none of them are what I expected going in.

Expertise is the ceiling, not the tool

I assumed leading a deployment meant I understood the technology's limits. It doesn't. Daily use is a different skill, and the thing that determines how far AI can take you isn't the tool itself. It's how much domain expertise you bring to the conversation.

That cuts in your favor. AI doesn't replace judgment built over a career, it makes that judgment move faster and reach further than it could before. My SRM prototype didn't take weeks because the tool is magic. It took weeks because the expertise was already mine. AI was just the acceleration.

Context beats the perfect prompt

I thought the secret was wording the question just right. It isn't. Context is the lever, not the wording: the background, the constraints, what success looks like in this specific situation.

Treat AI like a sharp consultant who just joined your company. Hand it the full history of a vendor relationship, including where you have leverage and where you don't, and you'll get back thinking that would normally take days to develop alone. But that only works if your question is precise. Vague instructions get vague answers, the same way a vague brief gets a vague result from a capable new hire.

Stay in the conversation

What I expected going in was speed, more output, faster. That's not where the value turned out to be. The bigger value is a thinking partner with no politics and no ego, one that will hold the full weight of a hard decision and push back on your assumptions if you let it.

It only works as a partner if you stay in the dialogue. Hand it a problem and walk away, and you'll get an answer that's technically correct and practically off. AI doesn't know when the tone is wrong for a specific audience, or when a strategy is defensible on paper but misses the point that counts in the room. That judgment still has to come from you. Staying in it is how your judgment ends up in the final output.

There's a gap opening between leaders building daily fluency with AI and leaders waiting for it to get easier. That gap compounds. The people working through hard problems with it now are building a speed and clarity of thought the rest of the field hasn't caught up to yet.

AI won't replace the judgment it takes to run a complex organization. What it does is put that judgment to work faster, in more places, on what matters most.

After a year in it, here's what I keep coming back to: it works anywhere you already know what good looks like.

---

Tracey Gatlin is the founder of Pivotal Move Advisors, a fractional executive consulting practice focused on operations, procurement, and AI-enabled growth. She works at the intersection of experienced leadership and applied AI, helping organizations turn deep operational experience into faster, sharper execution.