Platform Engineering’s Second Act: From Vending Machine to Passport Control

Three years ago on the QCon SF stage, I made the case for “Acceleration, Autonomy, and Accountability” as the pillars of a successful platform. Those pillars haven't moved. AI has just rewritten what each one requires, and the platform team's job along with it.

Every organization I talk to is somewhere on the same curve, from a few engineers quietly running experiments to a mandate that "AI-native" show up on the roadmap. Few have honestly answered the question that actually matters: did any of it move the needle for the business?

In this talk, we’ll cover:

• Where your platform team sits today on the curve from AI experimentation to AI-native, and what changes in its mandate, org architecture, and skillset at each stage
• Why "vending machine" infrastructure — compliant, fleet-managed, disposable — was never the hard part, and what the real missing layer looks like: the capability API and lifecycle contracts that sit between infrastructure and autonomous agents
• The unsolved problem hiding inside that layer: data access and agent trust, and the architecture we're using to give agents scoped, revocable access instead of all-or-nothing keys
• A concrete case study from Apollo's platform team — joined by Alex Mann, who leads it — on the point where this architecture had to get real, and what broke first
• An honest scorecard from our own organization's transformation: what got faster, what got better, and where business impact is still genuinely an open question

For platform leaders navigating the same curve, this talk is a primer on what to build next, what to postpone, and what to answer honestly before you call your org AI-native.