A Former Microsoft Exec Says This Is How to AI-Proof Your Career

Rob Collie spent thirteen years inside Microsoft. Product leader on Excel, and one of the founding engineers on Power BI — the tool that turned raw company data into the dashboards half the corporate world now runs on.

He left in 2010. In 2013 he founded a consulting firm called P3 Adaptive, built entirely on the thing he already knew how to do. Today it's about fifty people, a Microsoft Solutions Partner for Data and AI, serving mid-market and Fortune 1000 clients across oil and gas, manufacturing, and chemicals.

He also has a test for whether AI is coming for your job. It's one sentence long, it requires no technical knowledge to run, and it produces an uncomfortably clear answer.

The two-week test

If a brand-new hire could become productive in your role within about two weeks, that role is first in line for automation.

Collie's example is call center work. Documented policies. Decision trees. Predictable responses to predictable situations. A new person can be handed the binder on Monday and be taking live calls by the middle of the following week.

That speed is exactly the problem. If the entire job can be transferred to a stranger in two weeks, then the entire job has been written down somewhere. And anything written down can be handed to a machine.

He's already watching the effect in hiring patterns. As he puts it, "the zero-years-experience job is now turning into, like, a three-years-experience job." The bottom rung isn't being automated out of existence so much as being quietly raised out of reach.

Now run the test in reverse.

What the test protects

Jobs built on years of accumulated judgment sit at the opposite end. Not because they're more technical — often they're less technical. Because they can't be documented.

Knowing which of the four fields labeled "customer ID" the underwriting team actually trusts. Knowing why the Thursday report gets built manually even though there's an automated version. Knowing that a particular process is technically obsolete and still load-bearing, because a compliance decision from 2019 quietly depends on it and the person who made that decision retired.

None of that is in a manual. It exists in the heads of people who have been somewhere long enough to have absorbed it by accident.

Collie's explanation for this is technical, and it's the sharpest part of his argument. That kind of knowledge doesn't fit in a context window — the limited amount of information a model can hold and reason over at one time. You cannot hand a system twenty years of understanding how a specific business actually behaves, because nobody ever wrote those twenty years down.

He ran the test on his own company

Here's where the story stops being commentary and starts being evidence.

P3 Adaptive was built on Power BI dashboard development. That was the product for more than a decade. It's what the firm sold, what it hired for, and what its reputation rested on.

Dashboard development fails the two-week test.

It's structured. It's documentable. It follows patterns. It is, in other words, precisely the category of skilled technical work that current AI systems have gotten genuinely good at — and Collie, who helped build the underlying platform at Microsoft, is better positioned than almost anyone to see it coming.

So he did something most founders in his position would not do. He froze hiring. And he began retraining his own staff out of the dashboard-development roles his own company was built on.

That's not a consultant predicting a wave for an audience. That's an owner who ran the test on his own business, didn't like the answer, and moved before he had to.

He's also candid that his first reaction wasn't strategic. It was fear. "I wasn't immune to that," he says — a man who spent thirteen years building this category of software at Microsoft. What changed his mind wasn't reassurance. It was working out, concretely, what companies actually have to do to make AI useful.

What he did instead of hiring AI people

This is the part worth copying.

He did not go out and recruit machine learning engineers. He took people who already understood the business deeply and taught them the new tool.

The logic follows directly from the test. If the durable asset is undocumented business context, then the fastest path to a working AI capability is to start with people who already carry that context and add the tooling — not to hire tooling expertise and spend two years teaching it the business.

His framing is blunt: "Don't treat AI as a technology problem. It's a business problem. It's about teaching it about your business."

The practical method is less exotic than it sounds. Collie describes writing what amounts to a handbook for the AI — a written brief covering what its role is, which information it should rely on, what standards apply, how it should communicate. "Every time it wakes up, it needs to be handed that manual before it starts working with you."

Read that again as a job description. The person best equipped to write that handbook is not the most technical person in the building. It's the person who knows how the place actually works. Which is why Collie can say, without much exaggeration, that you can become the AI expert at your company in relatively short order.

If you're in transition right now

Most people read a story like this one and hear a warning. I hear close to the opposite.

If the work that disappears first is the work a new hire could absorb in two weeks, then long experience just changed categories. For most of the last decade, senior professionals have been told their experience makes them expensive. Overqualified. Hard to place. A hiring manager's risk.

The two-week test says the reverse. The years are the part that survives — not as sentiment, but for a specific mechanical reason. They don't compress. They don't fit in the window. They can't be handed over in a binder.

And notice what Collie's own career actually is, twice over. He took deep expertise in one domain, walked it out of Microsoft, and built a business on it. Thirteen years later, facing a wave aimed directly at that business, he's doing the identical move again — pointing the same accumulated understanding at a new tool.

Neither move required becoming a different person. Both required being honest about which parts of the work were durable and which were about to be commoditized, then getting in front of it.

That move works whether you make it inside a company or on your own. The hard part isn't the tool. It's being specific about what you actually know that nobody wrote down.

Most people have never made that inventory. It's worth making before the market makes it for you.

See Where You Fit

I'll build you a free, personalized Opportunity Map + Executive SWOT — the specific companies hiring for a background like yours, and an honest read on how you show up right now. Back in your inbox within 24 hours, no cost: