AI helps strong DEV teams and hurts weak, according to Google’s 2025 Dora Report

Follow ZDNet: Add us like a preferred source On Google.

ZDnet’s most important take -away restaurants

• Almost all developers now rely on AI tools.
• AI strengthens the strengths and increases dysfunction.
• High quality platforms are a must for AI success.

Google has been released 2025 Dora Software Development Report. Dora (DevOps Research & Assessment) is a research program at Google (part of the Google Cloud Organization). Dora investigates the possibilities and factors that stimulate software delivery and operational performance.

This year, the Dora project has investigated 5,000 software development professionals in various industries and followed more than 100 hours of interviews. It is perhaps one of the most extensive studies of the changing role of AI in software development, especially at company level.

Also: 10 chatgpt codex secrets that I only learned after 60 hours of programming with it

This year’s results are particularly relevant because AI has infiltrated in a fairly extreme software development. The report shows some encouraging notes, but also shows some areas of real challenges.

When writing this article I went through the report of 142 pages and pulled five important observations that cut the hype to reveal, which really changes into software development.

1. AI is now widely used in development

According to the respondents of the survey, somewhere between 90 and 95% depends on software development for work. The report states 95% in the intro and 90% later in a details section, but regardless of which number you choose, almost all codingers now use AI. According to the report, this is a jump of 14% compared to last year.

The median time spent interaction with an AI was two hours a day. However, there is more nuance in it. For example, only 7% of the respondents “Always” always report the use of AI when they are confronted with a problem to solve. The largest group, 39%, “sometimes” reports for help to AI. But what struck me is that a full 60% AI uses “about half the time” or more when you try to solve a problem or complete a task.

Eighty percent of the programmers reported a general increase in productivity, but only 59% reported that their code quality improved. Another important benchmark is this: 70% of the respondents relate to the quality of the AI, while 30% don’t.

Also: I received 4 years of product development in 4 days for $ 200, and I am still stunned

Let me share a personal thought about this. I just finished with a huge coding sprint made possible by the AI. The code that came out was almost never good on the first run. I had to spend a lot of time cauning the AI ​​to get it right. Even when the work was done, I went back to do a full QA wheep, where I found more mistakes.

My conclusion is that I cannot be received in any way near the amount of work I just did without AI. But there is no way to trust in Ach, I am going to trust a code that the AI ​​writes without doing much assessment, validation and testing. That of course does not differ much from how I felt when I was a manager and was encrypted to employees or contractors.

2. Think of AI as an amplifier

This was one of the more fascinating results that came from the study. The Dora team claims that AI has become an amplifier. In essence, AI “increases the strengths of good performing organizations and the dysfunctions of struggling.”

That makes sense. If you read my most recent article about “10 chatgpt codex secrets that I only learned after 60 hours of a few programming”, I pointed out that AIS quickly makes big mistakes. One deformed promptly can disable an AI to create a large destruction. I had the experience where Codex decided to remove a large part of one of my files and then immediately check in those changes in Github.

Also: I did 24 days in 12 hours with an AI tool of $ 20 – but there is one big pitfall

Fortunately I was able to roll back those changes, but I saw a huge amount of work disappear faster than I could take a sip of coffee.

In essence, the more effective and organized a team is, the more AI will help. The more spread or random a team is, the more AI will hurt. In my case I really have a good overhaul control, so when the AI ​​ate my homework, I could get it back because of the controls I had set before I once gave the AI ​​the first access to my codebase.

3. Seven team archetypes in the AI ​​era

So who wins and who loses? The Dora team identified eight factors that determined the general performance of a team.

1. Team performance: Effectiveness and cooperation strength of a team
2. Product performance: Quality and success of products that are produced
3. Software delivery throughput: Speed ​​and efficiency of the delivery process
4. Software -Sleveringsinstability: Quality and reliability of the delivery process
5. Individual effectiveness: Effectiveness and sense of satisfaction for individual team members
6. Valuable work: Degree in which individual team members think their work is valuable
7. Friction: How much does people trying to get their job done
8. Burnout: Feelings of exhaustion and cynicism among team members

They then measured these factors against respondents and their teams. This helped in identifying seven team archetypes.

1. Fundamental challenges: Survival mode, everywhere holes
2. Legacy Bottleneck: Constant fire fighting, unstable systems
3. Limited: Stable but bogged down by bureaucracy
4. High impact, low cadence: Strong output, unstable delivery
5. Stable and methodically: Deliberate pace, consistent quality
6. Pragmatic artists: Reliable, fast, moderately involved
7. Harmonious highlights: Sustainable, stable, top performance

Ai, the report says, is a mirror of organizations. The use of AI makes the strengths and weaknesses of teams clearer. But what I found particularly interesting is the idea that the assessment of the “speed versus stability” is a myth.

This is the idea that you can be fast or you can produce good code, but not both. It appears that the top 30% of the respondents fall into the harmonious high -performanceers or pragmatic artists archetypes, and those people quickly produce output, and the quality of that output is high.

4. Seven important practices

The report emphasizes: “Successful AI acceptance is a system problem, no problem with tools.” The Dora people seem to like the number seven. They say that the next seven important practices stimulate the impact of AI (for good or bad).

1. You have polycy: The clear, communicated AI attitude of an organization.
2. Data Ecosystems: General quality of the internal data of an organization.
3. Accessible data: AI tools connected to internal data sources.
4. Version control: Systematic way to manage changes in code.
5. Small batches: Breaking changes in small, manageable units.
6. User focus: Teams that prioritize the experience of the end users.
7. Quality platforms: Shared options available throughout the organization.

As you could imagine, the successful teams use more of these practices. Although the non -successful teams may have very productive individual programmers, it is the lack of these foundations that seem to lower them.

They recommend: “Treat your AI adoption as an organizational transformation. The greatest efficiency will result from investing in the fundamental systems that strengthen the benefits of AI: your internal platform, your data ecosystem and the nuclear -technical disciplines of your teams. These elements are the essentials for the changes of the changes of changes”.

5. Two factors that influence AI success

Last year it became quite big news when the previous Dora report showed that AI actually reduced the productivity of software development instead of it. This year the opposite is true. The Dora Explorers were able to identify two important factors that reversed those results.

Development organizations are more familiar with AI and know how to work more effectively than a year ago. The study shows that 90% of the developers organizations have adopted platform technique. This is the practice of building strong internal development platforms that collect the tools, automation and shared services for a development team.

Also: the best AI for coding in 2025 (and what cannot be used)

According to Dora, when the internal platform works well, developers spend less time fighting the system and more time creating value. If you consider AI as an amplifier, you can see how good systems can really improve the results. Interestingly, if platforms are weak, AI does not seem to improve the productivity of the organization. Good internal platforms are a very clear condition for effective AI use.

The next factor seems like a fashion word from a sitcom in the workplace, but is really very important. It is VSM (or Value Stream Management). The idea is that managers make a card of how work from idea to delivery moves. It is actually a current chart for operations instead of just bits.

By seeing each step, teams can identify problem areas, such as very long code reviews or releases that get stuck in different phases. The report states that the positive impact of AI acceptance is “dramatically strengthened” in organizations with a strong VSM practice. For the record, the word “dramatic” appears four times in the report.

The report states: “VSM acts as a power-to-have for AI investments. By offering a display at a system level, it ensures that AI is applied to the right problems, thereby converting localized productivity gain into important organizational benefits instead of simply creating electric chaos.”

What it all means for software development

There are a few clear conclusions from the report. Firstly, AI has moved from hype to mainstream in the world of the development of Enterprise software. Secondly, the real benefit is not about the tools (or even the AI ​​that you use). It is about building solid organizational systems. Without those systems, AI has little advantage. And third is AI a mirror. It reflects and increases how well (or bad) you already work.

What do you think? Has your organization used AI tools in software development? Do you see AI as a real productivity boost or as something that adds more instability? Which of the Seven Team archetypes feels closest to your own experience? And do you think that practices such as Platform Engineering or VSM really make the difference? Share your thoughts in the comments below.

You can follow my daily project updates on social media. Make sure you subscribe to My weekly update newsletterand follow me on Twitter/X @Davidgwirtzon Facebook on Facebook.com/davidgwirtzon Instagram on Instagram.com/davidgwirtzon bluesky @Davidgwirtz.comand on YouTube Youtube.com/davidgwirtztv.