What DORA 2025 Really Tells Us About AI and Software Teams - Part I of II

Key Takeaways

• AI is everywhere, but trust isn’t. 90% of developers now use AI tools, yet one-third don’t fully trust what AI produces.
• AI amplifies your culture. That’s right, good or bad! Great teams use AI to move faster; struggling ones just create chaos faster.
• Foundations still rule. Platform engineering, small batch delivery, and clear processes matter more than ever.
• Speed ≠ stability. AI boosts throughput but can hurt reliability if guardrails aren’t in place.
• Your best move? Strengthen fundamentals before scaling AI. Otherwise, you’re just accelerating dysfunction.

🤖 The AI Era Is Here (and Slightly Confused)

Let’s face it, 2025 is the year AI officially joined every standup meeting. According to the DORA report, over 90% of software professionals now use AI in some capacity, from autocompleting code to generating test cases.

The productivity gains are real: 80% of developers say AI makes them faster. But there’s a twist. 30% of developers admit, they don’t trust AI’s output. I can relate. I’ve accepted AI’s “helpful” code suggestions, only to spend my afternoon untangling its creative logic. It’s like pair programming with a genius intern who occasionally forgets how loops work or just occasionally doesn’t listen at all..

The takeaway: AI isn’t replacing engineers, it’s stress-testing our processes. It makes the good teams great and the messy ones… well, memorable.

⚡ AI Is an Amplifier, Not a Savior

DORA’s 2025 data drops a humbling truth: AI doesn’t fix broken systems, it amplifies them.

If your organization already practices good DevOps fundamentals (like version control, automation, and strong feedback loops), AI will make you faster and happier. But if your delivery process feels like a Rube Goldberg machine held together by duct tape and Slack messages, AI just speeds up the chaos.

Think of AI as caffeine for your engineering team: it boosts whatever energy you already have. If your culture’s great, you’ll build rockets. If it’s toxic, you’ll just burn out faster.

🏗️ The Real MVP: Platform Engineering

Here’s one of the DORA report’s less flashy but crucial findings: 90% of organizations now practice platform engineering, and those that do see the best AI outcomes.

A well-designed internal developer platform (IDP) is the foundation that makes AI useful. It ensures your environments are consistent, your pipelines automated, and your teams focused on delivery instead of firefighting. Without this backbone, all the AI in the world can’t save you.

Or as DORA puts it more politely: “AI provides the most benefit when combined with platform maturity.” Translation: stop chasing AI magic; fix your plumbing first.

🧩 Value Stream Management: Making AI Wins Stick

Even when AI works beautifully, there’s still the “last mile” problem — turning individual gains into team-wide improvement. Enter Value Stream Management (VSM).

Teams using VSM effectively can translate AI-driven boosts in productivity into measurable business impact. Without it, you just end up with pockets of efficiency: one dev cranks out features faster, but QA is still drowning in rework.

VSM isn’t glamorous, but it’s the connective tissue between fast code and real value. AI can make your developers sprint; VSM ensures the baton actually reaches the finish line.

🚦 Speed Without Safety Nets = Disaster

AI makes delivery faster. The DORA 2025 data shows clear gains in throughput i.e. the ability to ship features more quickly. But there’s a catch: stability often drops when AI is introduced too fast.

Teams that didn’t strengthen automated testing, CI/CD, and incident response found themselves fixing more post-release bugs. Essentially, they turned up the volume without checking if the speakers could handle it.

As one of my colleagues put it: “AI doesn’t break production, people using AI without guardrails break production.” A strong safety culture, automated checks, and version control discipline are what keep AI from turning progress into pandemonium.

🧠 What You Should Actually Do

DORA’s recommendations (and my own hard-earned lessons) boil down to six steps — in order of priority:

1. Set an AI policy. Define how and when AI can be used in coding, data, and documentation. Make it explicit.
2. Invest in your platform. Automate, standardize, and de-silo your developer environments.
3. Prioritize version control and testing. CI/CD is your parachute — pack it before you jump.
4. Train your teams on responsible AI use. AI literacy is the new DevOps literacy.
5. Map your value stream. See where work actually flows — or doesn’t — so AI gains compound.
6. Measure everything. Without metrics, AI success is just a good story.

💬 Your Turn

Here’s my challenge: Pick one foundational improvement your team can make this quarter, not another AI experiment, but something that will make AI worth using. Maybe it’s automating a test suite, or cleaning up your pipelines.

Post it. Share it. Inspire others.

While AI can accelerate things dramatically, it can’t eliminate humans… yet?

Sources:

• DORA 2025: State of AI-Assisted Software Development
• Google Cloud Blog: Announcing the 2025 DORA Report

Hashtags: #AI #SoftwareEngineering #DevOps