 |
VOOZH | about |
|
VOOZH | about |
We’re so glad you’re here. You can expect all the best TNS content to arrive Monday through Friday to keep you on top of the news and at the top of your game.
Check your inbox for a confirmation email where you can adjust your preferences and even join additional groups.
Follow TNS on your favorite social media networks.
Become a TNS follower on LinkedIn.
Check out the latest featured and trending stories while you wait for your first TNS newsletter.
Do AI tools make developers 56% more productive, 26% more productive or — as a recent METR study showed — 19% less productive?
The answer is, as always, it depends. Throwing AI tools at developers won’t magically solve all of their problems. On the other hand, I regularly talk to developer productivity leaders from companies like Netflix, Slack, Shopify or LinkedIn at Hangar DX podcast, and they all agree that AI is a developer productivity force multiplier.
Sridhar Ramakrishnan, who leads the Developer Experience team at Slack and supports thousands of developers, says AI will improve developer experience and velocity by 10 times. Ryan Cooke, engineering director at Pinterest, sees the biggest opportunity to use AI in “projects that required a lot of people, were a lot of pain and nobody wanted to work on, such as code migrations.”
Here’s some practical advice I’ve learned from talking to experts on using AI tools to meaningfully improve developer experience and impact.
AI tools bring the most impact when applied to large-scale engineering problems that, as Shopify’s Daniel Doubrovkine put it, are hard for humans alone to solve. Shopify’s developer productivity challenges were flaky tests and low test coverage, so they tackled those using AI agents. It created and open sourced Roast, a tool designed to structure AI workflows effectively. (The name Roast comes from its initial use case: roasting your tests to find areas for improvement.)
Slack turned to developer surveys to find existing pain points worth solving with AI. This led to experimenting with AI using a large language model (LLM) to convert 15,000 unit and integration tests from Enzyme to React Testing Library (RTL). By combining abstract syntax tree (AST) transformations and AI-powered automation, it achieved an 80% success rate in code conversion, greatly minimizing manual work and showing how AI tools can be used in the software development life cycle (SDLC) beyond just coding assistants.
Max Kanat-Alexander, the author of the book “Code Simplicity,” has a simple piece of advice: If you don’t know where the frictions are, ask your developers. Ask them where they would like improvement and how AI tools are helping them or not. Developer feedback should be the primary signal to observe, combined with quantitative data. Developer experience is not just about making developers more productive; it’s about pleasing developers:
“Happy people are more productive, and productive people are happier. It’s hard to imagine someone being as productive as they could be if they’re consistently unhappy at work.”
Slack doesn’t expect every developer to be on board with using AI tools. They take extra care in communicating when introducing such tools into workflows so developers don’t get a negative first impression.
AI can help developers produce more code, but that doesn’t automatically mean they’re being more productive. Even if the metrics you’ve decided to track show an increase in productivity, Kanat-Alexander reminds us that metrics are only meaningful if they’re actionable and have a business impact:
“Productivity is about being both efficient and effective. If you’re generating code that doesn’t deliver value, you’re just moving fast in the wrong direction.”
Developer time saved is a useful metric to consider, but not blindly. It has to be applied across the whole SDLC, not just code generation. What do developers do with the time they save by typing faster?
Waiting three days for pull requests to be reviewed and tested just to bounce back for fixes means that inefficiencies across other steps of the SDLC will eat up any productivity gained in faster code creation.
There is no AI tool that you can switch on and instantly solve your engineering organization’s problems. None of the companies interviewed claimed they had “solved” AI in development. Instead, they approached it like any new engineering initiative: experiment, fail, iterate, repeat.
In fact, Ramakrishnan says that all of their successful AI projects, like the code migration one, started as failures.
“We didn’t just plug the code into AI tools and everything worked right away. None of our AI experimentations was like that. So, the main failures were the ones when we tried doing something once and stopped.”
Poor documentation, unclear ownership and fragmented tooling don’t magically disappear with AI. If your workflows are messy, AI will only amplify the chaos, almost all of the experts stress.
Titus Winters, senior principal scientist at Adobe, focused on psychological safety in engineering teams, reminds us that it all boils down to culture:
“The rise of generative AI [GenAI] is a great opportunity to reflect on how we work, whether it involves AI or not.”
Kanat-Alexander is excited about the potential of AI to transform software engineering, but advises taking everything with a grain of salt:
“Find a sweet spot between hype and skepticism. Be realistic about what AI can deliver now for software engineering and optimistic about its future potential. The technology is improving at an unbelievably rapid rate.”
AI is not a magic solution for developer productivity. It’s a powerful tool that, when applied thoughtfully, can address real pain points and unlock meaningful gains.
Top engineering teams succeed not by blindly adopting AI but by experimenting, listening to their developers and aligning tools with actual needs.
