With the advent of AI, a running joke is making the rounds in software circles. Say it traditionally takes two days for a software developer to write code and one day for a quality assurance engineer to test it. Now with AI as a sidekick, it takes just one day to write the code, but two days to test the sloppy results.
That might prompt a laugh among developers, but it’s a serious challenge for software engineers and their companies. AI-driven productivity gains are shifting the balance of labor from writing code—traditionally the lion’s share of development—to everything else that follows before a piece of software goes out the door.
Where what developers call the Inner Loop has long been the center of the action, now the Outer Loop is growing in importance. I know, it sounds like the plot of a sci-fi movie—complete with highly capable robot assistants.
It’s early days for AI and software development, so this dramatic change hasn’t swept through the ranks just yet. But as the founder of three software firms, I see it coming sooner rather than later. To avoid getting caught flat-footed, companies and teams should start planning now.
THE INNER VS. OUTER LOOP
While the process isn’t always formalized, many software teams apply a similar division of labor when it comes to making their products.
The so-called Inner Loop encompasses the high-value creative tasks involved in developing software. They include designing, writing, building, and debugging code—work performed by individual developers before they share it with fellow team members.
By contrast, the Outer Loop is home to the repetitive tasks that drain attention away from the high-value work taking place in the Inner Loop. That includes testing the code—doing security, reliability, and quality assurance so it’s ready to use.
So how does AI shift the balance between the two? Intelligent agents help developers create code more easily. The mental heavy lifting of translating a concept into strings of code—work that can consume days or even weeks for humans—is accomplished in seconds by AI. But the results are more akin to stream of consciousness than a polished novel. All that raw material needs an editor to prime it for publication.
For example, developers using GitHub’s Copilot AI assistant are 55% more productive. The catch? Copilot yields code with security bugs and design flaws 40% of the time. Combined with the increase in code volume, those vulnerabilities turn the Outer Loop into a bottleneck.
How will that situation change development teams? A common ratio of developers to testers is three to one. At a big bank with 40,000 software engineers, 10,000 might do security, reliability, and quality control. But the AI effect is like squeezing a balloon so it expands on the other side. The coding productivity jump is offset by a dramatic increase in cycles spent on testing.
HOW DEVELOPMENT TEAMS CAN GET AHEAD
For software teams, the pressure is on to adapt. Companies that want to stay ahead of the game should first get a handle on a long-time adversary: toil.
Toil refers to the tedious, repetitive tasks that already consume too much of developers’ time and sanity. Whether they’re testing code manually, waiting around for builds to finish, or seeking approval to move things along, software engineers find themselves frustrated, slowed down, and pulled away from the creative work that drives innovation.
As AI automation makes the Inner Loop ever faster, the Outer Loop threatens to get bogged down in more toil than ever. The answer: automated, scalable systems for security, reliability, and quality. Many people regard all this as a boring afterthought. Indeed, I’ve seen plenty of engineering teams who are happy enough to hack together their own DIY fixes.
But professional tools are essential to eliminating the toil that still plagues the software development life cycle. Continuous integration/continuous delivery (CI/CD) is a crucial one. High-performing software teams use it to automate much of the labor needed to push new code through production. Building, testing, deployment—CI/CD takes care of what has historically been a tedious series of manual tasks.
A big part of that is security testing, one of the most time-consuming jobs. Besides giving a clear picture of vulnerabilities, CI/CD can prioritize the most urgent ones and recommend quick, efficient fixes.
Cloud cost management is another Outer Loop task ripe for automation. The right tools can track cloud usage in a high level of detail, shutting down idle resources to cut costs. In my experience, the savings can be as much as 70%.
Even something as simple as internal developer platforms (IDPs) can be game-changing here. These self-service portals help speed up production by gathering together tools, services, and information, including an inventory of software components.
All of these tools are themselves being enhanced by generative AI, making them even more capable of minimizing toil and accelerating deployment.
START DEVELOPING TALENT NOW
Automation is just half the battle. As the Outer Loop expands, it will also fuel a talent shortage. I’m not kidding when I say we’ll soon be in a world where QA engineers are more sought-after than coders in some circles. Folks who specialize in testing might also start commanding bigger paychecks.
Companies should plan accordingly by recruiting and building the talent they’ll need. Central to that effort is an attractive developer experience. Pay is just a starting point; developers also need to understand the company’s mission, feel challenged in their work, and have the right tools.
What roles will be most in demand as AI reconfigures the software development life cycle? Expect the QA manager as well as the engineer to become hot tickets, along with the reliability engineer.
AI won’t bring this transformation overnight, but software teams that prepare now can be in a better position to get their products to market faster. Ultimately, shoring up the Outer Loop could make the difference between success and failure. That’s no joke.