Last week, I asked AI to read 186 documents for me.

It did it in an afternoon. I drank coffee.

The project was a competitive landscape analysis for a strategy initiative. I needed to understand who 186 different licensed operators were, what they did, where they operated, and how a specific piece of regulation affected each of them. The traditional version of this job is one of two things. Hire a research consultant, pay six figures, wait three weeks. Or assign it to a junior analyst, brace for the timeline, and accept that the depth will be uneven.

I did neither.

I wrote a prompt that described what I needed. I told the AI which categories I wanted for each operator. I gave it examples of the kind of output I was looking for. I let it loose on the dataset, reviewed the results, asked it to fix what looked off, and ended up with a clean, structured analysis I could take into the next meeting.

Total time: under four hours. Total cost: the AI subscription I was already paying for.

I'm not a developer. I'm not a data scientist. I don't write code for a living. What I am is reasonably organized and reasonably clear about what I want. That, it turns out, is enough.

What this actually looked like

There's a phrase floating around tech circles right now: vibe coding. The idea is that you describe what you need in natural language, and AI helps you build the tool. You don't need to know how the AI is writing the code or how the analysis is structured under the hood. You need to know what good looks like, and you need to be able to tell the AI when it's wrong.

For the 186-document project, the workflow looked like this.

I started with a clear sentence: I have a list of operators. I need to know who they are, where they're licensed, and how a specific regulation affects them. I gave the AI a sample of the source data so it could see what it was working with. I asked it to draft an approach. I pushed back when the approach was too generic (I need state-level detail, not regional) and refined until the output structure made sense.

Then I let it run. When the first pass came back, about 70 percent of it was right and 30 percent needed correction. I worked through the corrections in a single sitting. The next pass was 90 percent right. By the third iteration, I had something I would have paid a consultant for.

The whole project sat on my laptop. No engineering team. No new software. No CIO conversation. Just clarity about what I wanted and willingness to push the AI when it missed.

Why this matters if you're not in tech

The companies I work with are sitting on top of the same kinds of analytical questions I tackled with the 186-document project. Competitive landscape. Regulatory exposure. Vendor comparisons. Contract reviews. The work that used to require an outside consultant or a long internal cycle.

Most of those questions can be answered by someone reasonably organized, using AI tools that already exist, in an afternoon.

You don't need a data team. You need a clear question and a few hours.

The honest part

It isn't always this clean. Some projects don't work this way. AI struggles when the data is messy in ways it can't see, when the question requires judgment it doesn't have, or when the source material is locked behind systems that won't talk to it. Knowing when not to use this approach is half the skill.

But the times it does work, it works the way I just described. Hours instead of weeks. A laptop instead of a consultancy. And the leader who knows how to ask the question keeps the analytical authority that used to flow to outside experts.

That's the shift. It's quiet, it's already happening, and it's the most underrated productivity story in business right now.