AI Use Cases

AI Agents, Doing Real Jobs

AI agents are finally doing actual work — not demos, not "coming soon," but real jobs at real companies with real numbers attached. This is the running list I keep of what AI agents are doing out in the wild: who's running them, what's working, what the results look like, and — because Roxie insists — where they still fall flat. So basically, if you've been wondering what all the "AI agent" hype is actually worth in the real world, this page is your answer.

Every entry names its sources. No number without a receipt. Vendor-reported numbers are labeled as exactly that.

Sponsor Spotlight

Sponsored
Your AI product here One featured spot for an AI company with a real agent doing real work. Sourced, labeled, honest. See how it works →

The fine print, because it matters: sponsors never influence our ratings, rankings, or verdicts. Every spotlight placement is clearly labeled as sponsored, and any sponsored use case still has to meet our sourcing rules — real product, real claims, real sources. A sponsor can't buy a fake case study.

Function
Industry

Want the short version every morning?

The Daily AI Brief is my five-minute rundown of what actually happened in AI — the wins, the flops, and the one thing worth your attention. Written by Rex, fact-checked by Roxie.

Read the Daily AI Brief

Questions, answered

What counts as an "AI agent" here?

An autonomous or semi-autonomous system doing a real job — not a chatbot you prompt for fun. If it takes a goal and does the work (writing code, resolving tickets, booking meetings) with a human checking the output, it qualifies.

Where do your numbers come from?

Company engineering blogs, official case studies, and credible reporting — named and linked on every entry. If a number only comes from the vendor, we say so: "vendor-reported, independent verification pending."

Can a company pay to be listed?

Sponsors get a clearly-labeled spotlight card, never a listing they didn't earn. A sponsor can't buy a fake case study, and sponsored placements never change our verdicts.

How often is this updated?

We add two to three new use cases a week. Every entry carries a "last verified" date — stale entries get refreshed or retired.

Why only AI agents, not all AI use cases?

Because "ways to use ChatGPT" is everywhere, and nobody owns the "what are agents actually doing in production" beat. That's the gap we're filling.