It became a meme on the internet: “You can just do things.” This was never more true in IT than it is today.

Great programming languages, frameworks, cloud computing, containerization, and Kubernetes—paired with Agile methodologies—let us prototype and deploy MVPs in a matter of weeks or months. That’s for projects that once took years. I won’t even get into the old costs of buying hardware, finding a colocation facility, or setting up your own data center.

We’re far ahead of where we were 10, 15, or 20 years ago. New products and features roll out faster than ever.

This holds true in general. But at larger enterprise companies, things get more complicated. Individual teams can move quickly. The problem comes from dependencies between teams. Those create constraints from clashing priorities. Reconciling them all into a company-wide goal often slows everything down.

Imagine a team that manages an internal service tracking pizza availability in a certain area. They depend on another service for raw ingredients data.

The company is pushing to increase sales. The salad team has the best margins, so their backlog fills up with items on lettuce and kale. No room left for flour.

The pizza team is working toward the same sales goal to meet the company target. But they’re blocked on flour updates from the ingredients service. At the company level, we’re all aiming the same way. Up close, though, the pieces don’t always fit.

And this is where AI steps in as a real game-changer.

In a company where everyone sees themselves as one team—no jealousy over codebases—you can pull a repository, use an agentic AI to grasp the code, spot the relevant files and lines, and make small to medium changes in just a few hours.

With strong test suites across repositories, clear documentation, shared standards, and some pre-commit hooks, cross-team work becomes straightforward. You can go straight to an upstream dependency, add your feature, and send a PR.

The pizza team jumps into the ingredients codebase. The AI finds the right place for the flour update. They submit the PR. Code review takes some engineering time, of course. But it’s much shorter than the full feature cycle.

We can take it even further. Individuals can build prototypes in hours to test a concept or set up a short-lived feature—things that were impractical before.

The other night, I needed an API to support a proof of concept. It wasn’t the POC itself, just the missing piece to make it work. An idea woke me up at 3 a.m., so I got out of bed, opened Claude, and had the API running in about two hours.

If you have an engineering mindset and a solid grasp of system architecture, an AI assistant lets you literally just do things.