Abhi Shimpi is a technology executive specializing in enterprise technology and AI transformation for Fortune 500 organizations.
Every few years, our industry finds a new way to help developers work faster. Cloud changed infrastructure. Containers changed software packaging. DevOps changed how teams build and operate software. Today, AI is changing how quickly engineers can turn an idea into code.
But after leading enterprise modernization across hundreds of applications and global teams, I’ve learned that writing code faster does not necessarily mean delivering software faster.
I learned this firsthand while leading a modernization involving more than 140 mission-critical applications, over a billion documents and digital assets and hundreds of terabytes of enterprise data, supporting tens of thousands of associates.
Our biggest challenge wasn’t writing migration code or developing features quickly enough. It was everything surrounding the code: inconsistent delivery pipelines and testing approaches, environment dependencies, repeated security and governance requirements and additional steps between completed development and production.
That experience shaped a principle I still use: The constraint in enterprise software delivery is often not how quickly an organization can create code. It is how efficiently it can turn that code into secure, reliable software running in production.
AI makes that distinction even more important.
You’ve been optimizing the wrong thing.
Organizations often ask how to help developers write code faster. I believe the better question is, “How do we make thousands of engineers consistently successful?”
The first improves an individual. The second improves an organization. At that scale, modernization had to become repeatable.
We standardized CI/CD and delivery patterns, automated more of the build, test and deployment life cycle, created reusable capabilities, strengthened observability and self-service and incorporated security and governance into delivery workflows.
We moved repeatable engineering work out of individual application teams and into shared capabilities.
Across the broader modernization effort, these changes helped increase migration throughput by approximately 80% while reducing modernization timelines by approximately 60%. Those results didn’t come from engineers writing code 80% faster. They came from removing friction around them.
And that is why platform engineering becomes more important as AI-assisted development grows.
AI can feed the bottleneck faster.
When AI increases development capacity, every downstream part of the software delivery system has to absorb more change. More code creates more reviews, tests, dependencies and operational events.
If an organization relies heavily on manual approvals, bespoke pipelines and application-specific practices, AI can accelerate work directly into those constraints.
AI doesn’t automatically remove the bottleneck. It can feed the bottleneck faster.
Recent DORA research describes a similar tension: Time saved during initial development can shift toward auditing and verification. DORA describes AI as an amplifier, making strong engineering systems more valuable while exposing weaknesses in the systems surrounding development.
So instead of asking only, “How much faster can AI make our developers?” leaders should also ask, “Can the rest of our engineering system keep up?”
Platform engineering is about the path to production.
Platform engineering is sometimes treated as another infrastructure function. I think that definition is too narrow.
Infrastructure provides resources. A strong engineering platform provides a repeatable path from idea to production through standardized delivery, testing, security, observability and self-service.
The objective isn’t to force every team to work identically. It’s to stop asking every team to independently solve problems that aren’t unique.
I saw this during modernization. Established patterns and reusable capabilities meant teams didn’t have to re-create the mechanics of production delivery for every application. That was valuable before AI. It becomes even more valuable when AI increases the volume and pace of implementation.
The most valuable product may be the one your customers never see.
I’ve come to think of an internal engineering platform as a product. Its immediate customers are developers, but its impact extends to everyone using what those developers build.
Our users didn’t care whether deployment pipelines were standardized, security controls were embedded or observability was built into the platform. But they experienced the consequences. Repeatable delivery reduced manual intervention, observability improved diagnosis, and reusable capabilities freed capacity for new features.
At enterprise scale, that leverage compounds. A reusable capability across dozens or hundreds of applications changes the economics of engineering.
AI can multiply engineering output. The platform determines whether the organization can safely absorb it.
This changes where engineers add the most value.
Removing repetitive delivery work doesn’t make engineers less important. It moves their attention to harder problems. I saw this as shared capabilities matured. Standardization gave engineers more room to focus on complex engineering decisions and new ways to solve business problems.
AI extends that shift. As AI takes on more implementation work and platforms handle more repeatable delivery activities, human judgment becomes increasingly valuable.
Here’s my advice to technology leaders.
Before investing heavily in another AI development tool, examine whether the engineering system surrounding your developers is ready for increased output. Ask these questions:
1. Where does engineering work actually wait? Look at code reviews, testing, environment provisioning, security processes, release queues and production readiness. Ask, “If AI doubled our development output tomorrow, which queue would get longer?” That’s a strong candidate for your next platform investment.
2. How often are teams solving the same problem? Identify problems teams repeatedly solve independently and consider turning them into reusable platform capabilities. Communities of Practice and Centers of Excellence can also surface opportunities for reuse across silos.
3. Can our controls scale without proportional human effort? If software changes doubled, would security reviews, testing, release approvals or operational support need to double too? If so, AI may reveal a scalability problem before it creates a productivity advantage.
AI is making software easier to create. But generating more software isn’t the same as delivering more value.
The organizations that benefit most won’t simply produce code faster. Their engineering systems will absorb increased output without increasing operational complexity at the same rate. AI can make an engineer faster. A great engineering platform can make the entire organization faster.
That is why I believe platform engineering becomes more important—not less—as AI changes how software gets built.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?