Choosing AI Tools for Your Work: The PROVE Framework | Course Preview

The other day, I sat in on a 90-minute contextual inquiry session with someone who was, by any normal measure, an extremely heavy AI user. He had Claude Code panels open in VS Code, between five and 10 sub-agents running at any given time, a stack of markdown files orchestrating workflows, and an automation chain that

touched almost every part of his job and his team. He'd been building this system for months. But at no point during that hour and a half did he ever say, "This is saving me five hours a week."

He never said, "This output is definitely sharper than what I would've made on my own." He didn't offer, in concrete terms, whether the quality of his work was actually better, only that he had AI in it now.

And that's a little scary, don't you think? Here's what happens. People sit down to evaluate an AI tool, run it once, read the output, and make a verdict in isolation, off-the-cuff, based on vibes.

"Wow, it wrote a whole summary in 10 seconds," or, "Huh, this output is so generic." Both feel like assessments, but they're not. They're impressions.

The problem with vibes is that they're missing a comparison. Remember, you already have a way of doing this task.

AI doesn't exist in a vacuum. AI is competing against you, your current process, your current quality, your current speed.

That's the benchmark. Not can it do it at all, but is it better than what I already do? So next time you sit down to evaluate an AI tool, do the task yourself first. That's the version to compare against. Without it, you're just looking at the tool's output and guessing.

Thanks for watching this clip from my self-paced course, Choosing AI Tools for Your UX Work. The full course walks through the entire framework, five steps for evaluating any AI tool against the work you actually do.

You'll learn how to put a number on output quality, speed, and day-to-day friction, and turn the whole thing into a recommendation that you can defend to your manager, including a defensible no.

It's one of many self-paced courses at Nielsen Norman Group, designed so you can move at your own pace and revisit sections whenever you want. Head to nngroup.com to find this course and the rest of our library.