How engineering teams want to use AI in software development

With all the buzz around AI in software development, it’s easy to get lost in the hype. But what do engineering teams actually expect from these tools?

The truth is, at the core, every tech team wants the same thing: fewer boring tasks, more time to think. They don’t want AI to write an entire system on its own. They want it to take the load off so they can focus on what really matters.

A study from Microsoft and GitLab’s 2024 DevSecOps report show that expectations are high. But they also show teams are keeping it real. Here are a few insights:

1. Automate the boring, repetitive stuff

Let’s be honest: a big chunk of engineering work is repetitive. Unit tests, documentation, ops tasks. We do it because we have to, but no one wakes up excited for it.

96% of devs want AI to handle this kind of work, according to Microsoft. Makes sense: the more automation on standard tasks, the more room for creativity, problem-solving, and architecture improvements.

And timing matters. Teams that automate docs and tests right after writing code save hours over the sprint. That boosts not just productivity, but also team morale.

AI in software development shines here: it handles the operational so teams can focus on the strategic.

2. Better time use and task management

Backlog grooming, task management, constant context switching. All of it drains energy. AI can be a silent partner in this fight against chaos.

According to GitLab, 38% of leaders want AI to help design metrics and spot bottlenecks. This is not just about prettier dashboards. It is about surfacing signals that help teams make better decisions.

Is this task blocked? Why?
Who reviews this type of PR the fastest?
Which part of the codebase causes the most bugs?
Did the work that shipped actually lead to the expected outcome?

But there is an important caveat: isolated metrics do not tell the whole story.

A faster PR review, a shorter time to merge, or more AI-generated code can all look good on a dashboard. But none of that proves, on its own, that the team is shipping the right work.

To understand the real impact of AI, engineering teams need to connect the full chain: task intent → pull request → review → release → outcome. PR review is still an important signal, but it needs to be tied back to the task context, what was reviewed, what reached production, and what happened after release.

Otherwise, teams risk measuring activity instead of impact. As One Horizon argues, AI agent metrics need a work graph, not just another dashboard.

Having AI process these signals in real time helps teams react with more precision, instead of relying only on guesses or gut feeling.

3. Code quality with less friction

Code review is one of the most critical parts of the engineering flow — and one of the most prone to noise. Lack of context, PR overload, reviewers with no time.

37% of leaders want AI to explain vulnerabilities and how they can be exploited. That’s a game changer. It’s not just about pointing out what’s wrong — it’s about showing the impact and suggesting fixes. A reviewer that doesn’t just critique, but teaches.

Plus: AI can review with consistency, in seconds, and without cognitive fatigue. It doesn’t replace human eyes, but it sets the stage. Less noise, more clarity, faster cycles.

4. Smarter internal support

Every team has that tech support channel that turns into a FAQ for stuff that’s been answered a hundred times. “What’s the staging URL?”, “How do I run the project locally?”, “What’s this flag for again?”

36% of leaders want bots that reply with internal context. That frees up senior engineers to solve real problems and lowers friction for new people on the team.

Here, AI acts like a layer of shared memory. A copilot for onboarding, troubleshooting, and understanding complex systems.

5. The real limits and fears

Of course, the excitement comes with caution. Devs have very real concerns:

That AI ends up being more noise than actual help.
That it introduces bugs or vulnerabilities without oversight.
That it replaces human judgment with generic answers.

And those fears are healthy. They’re reminders that AI needs to be auditable, transparent, and only opinionated when it makes sense. Tools that don’t respect this get shut off, no matter how promising they seem on paper.

Using AI responsibly in software development means keeping this critical lens.

6. The future teams actually want

Engineering teams aren’t asking for magic. They want:

• Tools that fit into the real flow of the team.
• AI that understands context, not just code.
• Fewer clicks, less context switching.
• Real impact, not vague productivity promises.

Mature teams know AI isn’t a silver bullet. But they also know that when done right, it changes the game.

AI should make things easier, not harder

At the end of the day, engineering teams want AI that solves real problems. That clears repetitive work out of the way, helps keep code quality high, and supports the stuff that slows teams down.

They’re not expecting a magical revolution. They want tools that actually work, that fit into their flow and make building software better. If AI does that, it sticks around. If it gets in the way, it’s out.

The bar is clear: make the job easier, with less friction and more consistency. And that’s already a big win.