Table of Contents
- Introduction
- The productivity myth
- Where AI actually adds value
- Develop critical thinking skills
- Conclusion
Introduction
After reading countless studies and observing real-world implementations, I've learned to use AI as an augmentation tool rather than something that replaces my job. It's clear as day that AI adoption is on the rise in our industry. We can use it in various ways like as a co-teacher or co-worker, but the gap between marketing promises and actual results should be top of mind for all of us.
I'm a very pragmatic person, so I don't like hearing the positive perspective of using GenAI without talking about the downsides. There is a lot hype and investment in this field and only some reap the benefits of GenAI.
The productivity myth
In my opinion, the biggest mistake organizations make is chasing the wrong metrics in terms of software development productivity. It's easier to understand (and do marketing) on simple numbers like: "55% faster than the developers who didn’t use GitHub Copilot", "more than a quarter of all new code at Google is generated by AI" or "developers using AI are 19% slower".
Productivity gains aren't about producing more code, especially when it's easy to create AI slop. They also shouldn't be measured on producing boilerplate or simple tasks. A TODO app is different from a real production system. We can only make sure we have such gains by choosing metrics that make sense for our team and context, then measuring and reflecting on the results. This is how we can become more effective and steer the ship in the right direction.
I'm much more skeptical of statements done by AI vendors, CEO's or content creators, and that helps me keep focus on my goal which is to continue improving and bringing value to my team. If AI can help with that great, if not, life goes on.
Where AI actually adds value
From my perspective, the real value of AI in software teams lies in three specific areas that traditional tools cannot address effectively.
AI as Strategic Thinking Partner: I believe the most undervalued application is using AI for architectural discussions and trade-off analysis. When an engineer can have a deep conversation about a technical problem, generate 10 possible solutions, and then filter out the bad ideas, that's really helpful. This isn't about getting perfect implementation details - it's about expanding the solution space before making critical decisions.
Having a Co-Teacher: It's hard to be a force multiplier that improves everyone around you, which is why this is a key differentiator on senior developers that have this skill. The challenge of onboarding junior developers, explaining business logic, and sharing design patterns has always been there. We always want our senior devs to share and help junior devs grow, and using AI as a co-teacher helps us with that. Anthropic mentioned in their article how Claude Code helps them:
At Anthropic, using Claude Code in this way has become our core onboarding workflow, significantly improving ramp-up time and reducing load on other engineers.
Practical Augmentation as a Co-Worker: I'm convinced AI augments my team on the mundane but time-consuming tasks. Initial code reviews, generating PR summaries, drafting Architecture Decision Records, creating unit tests for specific scenarios, and generating KQL queries for troubleshooting. Our team at CloudCockpit has also been creating reusable prompts and custom agents that help every dev develop new features and have architecture reviews on proposals.
So far, I noticed that using AI is helping me think better, but it has the potential of helping me work faster with the use of these agents. Still, I mostly use it for "deep research" into possible solutions, learning new technologies through analogies, finding relevant documentation and troubleshooting problems. The most important piece remains, which is keeping a high level of technical excellence and quality in our team.
Develop critical thinking skills
Here's what concerns me most: the tendency for developers to become over-reliant on AI outputs without developing judgment to evaluate them. On the recent DORA 2025 report, they found 65% of technology professionals report to relying on AI at least a "moderate amount". It's important to understand this behaviour in our teams. All software engineers need to exhibit critical thinking skills, in my opinion, seniors more than juniors. But still, this skill must be learned and developed. We can't have good professionals in our field without this skill. But I am seeing more software engineers delegate their thinking to a machine, a tool.
Teams that accept AI suggestions without the "push-back" that experienced practitioners recommend, usually are trading off speed for quality. Sure, there is nothing wrong with that in some scenarios like prototypes and demo apps. For products with millions of users that need to be robust? No, not a good trade-off.
You should think critically about the AI output and be the human in the loop. Are you confident it will behave well if you give it more tools and autonomy? Are you confident the output is based on facts and truth, instead of lies and hallucinations? Don't delegate your critical thinking to tools, and don't become over-reliant on them either without fact-checking. Always evaluate if what the AI is telling you very confidently is even true, and be mindful of its limitations.
Conclusion
From everything I've observed, learning to use GenAI tools is something I recommend. Learn its strengths and limitations. Organizations that approach AI adoption with healthy skepticism while investing in experimentation, innovation and learning, will build sustainable competitive advantages.
Ask yourself these hard questions:
- Are you measuring business outcomes, or just code output from AI tools?
- Are your teams getting augmented, or just more dependent on external intelligence?
- Are you blindly believing the AI hype or learning how to leverage this new tool?
