Your Next Power User Isn't Human. Developers Are Rewriting Apps for Agents.

What happened

Bloomberg's newsletter coverage reports that software developers are systematically rethinking their applications for a new class of user: AI agents. The redesign spans everything from API surface area to pricing, as vendors move away from per-seat subscriptions toward usage and outcome-based models better matched to how agents actually consume software.

"Artificial intelligence agents are poised to reshape the application-software market, lowering barriers to software creation"

— Bloomberg Intelligence

Why it matters for SaaS buyers/founders

The implication for product roadmaps is bigger than it sounds. When your app's primary consumer is an agent, the traditional web UI becomes decoration. What matters is a well-documented, deterministic, machine-consumable API surface with fine-grained auth and predictable per-call pricing. The screens humans still use become a thin shell over the same primitives.

For buyers, this changes the vendor-evaluation checklist. It's no longer enough to demo a slick dashboard — you need to know whether the vendor exposes the same functionality via API, how they price agent-driven traffic, and what happens to your bill when a single autonomous workflow triggers a thousand actions in a minute. Vendors who can't answer those questions cleanly are going to lose the accounts where an agent-first buyer sits in the room.

Our take

The pricing implication is where this gets uncomfortable for incumbents. Every legacy SaaS company built on per-seat pricing has to answer an ugly question: if their most valuable customer is an agent that never buys a seat, what's the SKU? Some are trying credit-based systems. Some are trying outcome-based pricing. Most are trying both and confusing their customers in the process.

The founders best positioned right now are the ones building AI-native from day one — no per-seat legacy to defend, no sales team compensated on seat count, no CFO panicking about the transition. They can price cleanly for consumption because that's all they've ever done. The incumbents are going to spend the next two years untangling their pricing models while the AI-natives quietly take share in every category where agent-driven workflows are viable. That gap will not close quickly.