PlatformCon 2026: AI Is Raising the Stakes for Platform Engineering
- Daniel Bryant
- 27 minutes ago
- 5 min read
PlatformCon has always been a useful signal for where platform engineering is heading. And this year’s London Live event didn’t disappoint. The Syntasso team were out and about, watching talks, chatting with folks, and sharing recordings of their virtual talks (with more links coming soon!).
AI dominated almost every conversation, presentation, and panel. But despite the excitement around agents, copilots, and autonomous software delivery, the most important takeaway wasn't that AI is changing software engineering. Instead, it was AI that was making platform engineering more important than ever.
Across talks from practitioners at Booking.com, Thoughtworks, and elsewhere, a consistent theme emerged: the organisations that succeed with AI won't be those with the best models. They'll be the ones with the best platforms.
AI Needs Platforms More Than Platforms Need AI
Kaspar von Grünberg and Luca Galante set the scene early in the welcoming address:
"I'm 100% convinced that AI-based software engineering will not succeed without effective platforms."
That statement resonated because it captures a reality that is sometimes lost in discussions about agentic development. AI systems are probabilistic. Platforms must be deterministic.
An AI agent can propose architecture, generate code, create pull requests, or recommend infrastructure changes. But someone still needs to provide the secure environments, deployment workflows, operational controls, governance mechanisms, and feedback loops that enable those actions to occur safely and repeatedly.
The more autonomy we give software systems, the more important those foundations become. Shane argued exactly the same point in our webinar last week, “Lessons from Putting AI in Front of a Platform.”
AI Amplifies Existing Organisational Problems
Gregor Hohpe delivered what may have been the quote of the conference:
"GenAI does not make your problems go away. It amplifies your dysfunctions."
This idea appeared repeatedly throughout the day. Many organisations are approaching AI adoption primarily as a tooling challenge. In reality, most of the obstacles are organisational.
Poor developer experience.
Fragmented tooling.
Slow governance.
Unclear ownership.
Inconsistent environments.
These problems already slow teams down today. AI simply exposes them faster. If developers can generate code in seconds but still wait days for environments, approvals, security reviews, or operational support, the bottleneck becomes painfully obvious.
The lesson is familiar to platform teams: reducing friction matters. In an AI-enabled world, it matters even more.
The Role of the Platform Is Expanding
For years, platform engineering has focused on creating paved roads for application developers. Increasingly, those roads must support AI agents as well.
Patrick Debois explored this challenge through the idea of the "dark factory" and the systems required to safely enable autonomous software delivery.
The recommendations sounded remarkably familiar to anyone building internal developer platforms:
Sandboxed environments
Strong observability
Clear permissions and boundaries
Human override mechanisms
Governance and policy controls
Fast feedback loops
What's changing is the consumer. Platforms are no longer serving only human developers. They are increasingly serving a combination of humans, copilots, and autonomous agents. The responsibilities remain largely the same. The scale and speed are changing dramatically.
Security Must Be Built In, Not Bolted On
The security panel, featuring Nigel Douglas, Liz Rice, and Joe Baguley, highlighted another important trend. As AI accelerates software creation and potentially lowers the barrier to discovering vulnerabilities, traditional security approaches become increasingly difficult to sustain.
One panellist made a memorable observation:
"With the speed of developing exploits rapidly increasing with AI, patching is doomed."
The point wasn't that patching no longer matters. It was that organisations need additional layers of protection.
Runtime security.
Policy enforcement.
Platform guardrails.
Continuous verification.
Security can no longer rely solely on individual teams consistently doing the right thing. The platform itself must provide secure defaults. This aligns closely with a core principle of platform engineering: move expertise into the platform so that teams automatically benefit from it.
Guardrails Are Necessary. Lane Assist Is Better.
Another insight from Gregor Hohpe stood out:
"No one drives their car by bumping along the guardrails. You need lane assist as well as guardrails."
Many platform teams invest heavily in controls, policies, and restrictions. Those capabilities are important, but developers do not adopt platforms because they enjoy governance. They adopt platforms because they help them achieve outcomes faster and more safely.
The best platforms don't simply prevent mistakes. They actively guide users toward success. The emergence of AI makes this even more important. Developers and agents alike need systems that provide guidance, context, and automation, not just constraints.
Cognitive Load Remains the Core Challenge
Kasper Borg Nissen posed an important question in his talk, "Observability as a foundation for AI-enabled platforms": Is AI reducing or increasing cognitive load? The answer appears to be both.
AI can remove effort from many tasks, but it also introduces new tools, workflows, governance concerns, and operational complexity.
Platform engineering has always been fundamentally about managing cognitive load. That mission does not disappear in the age of AI. If anything, it becomes more critical.
Platform teams now have an additional responsibility: deciding which aspects of AI-enabled delivery should be exposed to users and which should be abstracted away. Success in supporting cross-functional requirements, such as observability, security, and reliability, will depend less on providing access to AI and more on providing the right level of abstraction around it.
Building Platforms That Float
Gregor Hohpe also revisited ideas inspired by Wardley Mapping and the need to continuously commoditise platform capabilities.
The warning was clear. If your platform simply duplicates capabilities already provided by cloud vendors, you risk creating unnecessary complexity and cost. Successful platforms create value by combining commodity capabilities with organisation-specific workflows, policies, and operational knowledge.
This mirrors a principle we frequently discuss at Syntasso: Build a platform that is unique to your organisation but common to your teams.
Platform teams should avoid reinventing what already exists. Their focus should be on capturing the knowledge, constraints, and practices that make their organisation successful. That is where differentiation lives.
AI Adoption Is an Organisational Capability
One of the most practical sessions came from Mansi Mittal and Bruno Passos at Booking.com, who shared lessons learned from adopting AI at scale.
The presentation reinforced something many organisations are discovering. Successful adoption is not primarily about model selection. It requires:
Education
Enablement
Measurement
Feedback
Community learning
Organisational alignment
The organisations seeing the most value from AI are approaching it as a platform capability rather than an individual productivity tool. That distinction matters because the goal is not simply to help individual developers write code faster; it is to improve the effectiveness of the entire software delivery system.
The Future of Platform Engineering
The closing panel with Gus Shaw Stewart, Cortney Nickerson, and Nicki Watt brought together many of the themes that surfaced throughout the day: AI, security, governance, developer experience, and the future of software delivery.
What stood out was how often platform engineering appeared as the connecting layer between them. Platforms are becoming the mechanism through which organisations operationalise AI. They provide the guardrails, lane assist, governance, security, context, and feedback loops that enable faster software delivery.
AI may be transforming software engineering. But PlatformCon's clearest message was that effective platforms will determine whether that transformation succeeds.
The future of software delivery may be increasingly autonomous. The future of platform engineering looks more essential than ever.
A big thanks to all the organisers, speakers, attendees, volunteers, and sponsors of the event from all at Syntasso! We’ll see you next year.
You can find our PlatformCon 2026 virtual talks recordings in our playlist (with more talks being added over the week).
Comments