TL;DR — Key Takeaways

  • AI moves at machine speed. Governance doesn’t: Developers and AI agents can generate thousands of lines of code in seconds, leaving platform engineers struggling to validate, secure and safely deploy the results.

  • More automation isn’t always the answer: Expanding infrastructure, compliance requirements and increasingly complex platforms risk turning platform engineering into an endless exercise in firefighting.

  • Stop being gatekeepers. Start being architects: By embedding security, compliance and operational guardrails into platforms, engineers can enable faster development without sacrificing control.

Just over 50 years ago, Fred Brooks, one of the most well-regarded figures in software development, published The Mythical Man-Month, a study of software management that in turn gave rise to Brook’s Law – “Adding manpower to a late software project makes it later“.

The premise is one familiar to anyone who has worked in approvals of any kind, let alone software development – in complex systems, delivery is rarely constrained by the ability to create, more so the need to validate, test, and ensure that a change won’t have unintended consequences elsewhere.

For platform engineers, the latter challenge is particularly relevant. Unlike many forms of engineering, they are often modifying systems that are already running, serving users and supporting critical business operations – think replacing a wing while the plane is flying.

This was a problematic task for an engineer in 1975, when Brooks also famously estimated that a professional software developer produced roughly ten lines of debugged, shipped code per day. Now a developer equipped with AI, or indeed an AI agent itself, can create thousands of lines of code in a matter of seconds.

Machine speed changed the equation. The faster we can create software, the greater the burden becomes on the people responsible for making sure it can safely make its way into production.

It was difficult enough for humans to manage the complexity created by other humans. Now they are managing a deluge of code produced at a scale that is quickly becoming unmanageable.

The Impossible Job

What makes a job difficult? What makes a job impossible?

For most of software’s history, the industry assumed the answer was productivity. Give developers better tools, automate repetitive tasks, reduce friction and the problem would solve itself.

Today, we’ve largely succeeded in that aspect. Tools are better and easier to use, AI can generate code in seconds, infrastructure can be provisioned automatically and entire application environments can be spun up with a handful of prompts. Yet despite all this progress, platform teams have never been more overloaded.

Every year the industry finds a way to add more. More software, more infrastructure and more services, whilst at the same time, the burden of understanding, governing and maintaining all of it continues to grow. Each new workload expands the surface area that organisations need to understand, secure and govern, and each technical decision brings knock-on consequences for everything from cost and compliance to sovereignty and sustainability.

The result is that platform engineers increasingly find themselves responsible for systems whose complexity grows faster than their ability to reason about them.

Moving Targets

In Greek mythology, Sisyphus was condemned to spend eternity pushing a boulder uphill, only for it to roll back down again. Platform teams increasingly know the feeling. One week they are standardising infrastructure, only to find a new cloud service has appeared. A governance process is finally automated, then fresh compliance requirements emerge. Costs are brought under control, only for another workload to arrive and begin consuming resources again.

It’s ironic, therefore, that some of the solutions proposed by our industry risk adding to the very complexity they are trying to solve. An emerging narrative around Platform Engineering 2.0 once again promises more. More personas, more workflows, more automation, autonomous agents and intelligence throughout the delivery lifecycle. The ambition makes sense, but there is a danger that in trying to accommodate everything a modern platform could do, we lose sight of what the people running it need.

The conversation is increasingly revolving around agents, pipelines, workloads and personas, but behind all those abstractions is still a person who gets pinged at three o’clock in the morning when something fails.

The platform engineer who inherits infrastructure they didn’t build and is trying to understand Terraform written by someone who left two years ago is often the same person attempting to explain an unexpected cloud bill to a chief financial officer or ensuring software can move quickly without exposing the organisation to unacceptable risk.

These aren’t edge cases for our industry – this is ‘the job,’ and increasingly, that job is becoming impossible.

From Gatekeeper to Architect

For too long, platform engineers have been positioned as the people standing between developers and production. Their expertise is used to answer questions that, in many cases, they have already answered before. Questions around security, cost, compliance or infrastructure are vital, but highly skilled engineers shouldn’t have to start from scratch every time they are asked.

The opportunity for platform engineering is to now move those decisions upstream. Give platform engineers the authority to define how software should move through an organisation and encode those decisions into the platform itself and governance becomes an integral part of the path to production.

This means that an engineer’s expertise is expressed through the environment they have designed, rather than through the number of requests they personally approve. Developers can do more for themselves because the safe route to production has already been defined, while platform teams retain control over the boundaries within which they operate.

For that engineer, that represents something of a promotion. They stop being the person operating every part of the platform and become the person shaping software delivery across the business – and defining the conditions under which those decisions can safely be made. No longer a gatekeeper, they become an architect.

This becomes particularly important as AI changes the economics of software development. If we can produce code at a scale that would have been unimaginable when Brooks was writing in 1975, we cannot expect the people responsible for safely putting that code into production to simply work faster to keep pace.

Platform engineers need to be able to establish the guardrails once and have the platform apply them consistently, whether the work originates with a developer or an AI agent. That gives developers greater autonomy without asking organisations to surrender control and frees platform teams to spend more of their time shaping how software is delivered rather than processing every individual request.

Brooks’ lesson was that more hands don’t necessarily make light work of software development. Fifty years later, perhaps the answer is to stop adding hands and give the people who understand the system the authority to design a better way through it.

Frequently Asked Questions

Why is AI making platform engineering more difficult?
AI dramatically increases the speed and volume of software development, but every new workload brings security, compliance, infrastructure and operational responsibilities. Platform teams must manage this growing complexity without becoming bottlenecks.
Why might Platform Engineering 2.0 make things worse?
Adding more agents, workflows, personas and automation can introduce additional complexity if these capabilities aren't designed around engineers' real operational needs. The goal should be to simplify delivery, not create more systems to manage.
How can platform engineers move from gatekeepers to architects?
By embedding governance, security policies and operational standards directly into development platforms, engineers can establish reusable guardrails. Developers and AI agents gain greater autonomy while platform teams maintain consistent control over production environments.

SHARE THIS STORY