The B2A Revenue Model: What Changes When Your Customer Is an AI Agent

Six posts into this arc, the pattern is clear: B2A (business-to-agent) isn't just a new traffic source. It's a different commercial relationship, end to end. This post synthesizes the revenue implications.

What the B2A funnel looks like

The traditional SaaS funnel: awareness → intent → trial → paid → retention.

The B2A funnel: indexing → routing → rate-limit → upgrade.

These are structurally different at every stage.

Indexing replaces awareness. You don't get into the B2A channel by running ads or writing guest posts. You get in by being crawlable, structured, and accurately described in AI knowledge bases. The screenshot API gets ~50% of its traffic from ChatGPT-User because it's indexed and its capabilities are unambiguous. This isn't marketing — it's technical infrastructure.

Routing replaces intent. The human behind the AI agent never evaluates your API. The AI system evaluates it, based on capability match and friction profile. A missing OpenAPI description or an authentication requirement at the free tier can route requests to a competitor without any human ever knowing your API exists.

Rate-limit replaces trial. The moment an AI agent hits your 429 is the moment that proves the integration works and demand exceeds free capacity. This is the highest-intent moment in the entire funnel — the equivalent of a human landing on the pricing page with credit card in hand.

Upgrade replaces conversion. The 429 response points to the upgrade path. The developer who built the agent (or the end user, if the agent surfaced the error) converts to paid. This conversion happens at the moment of maximum intent, which is why the 429 page needs to be a full commercial surface with multiple paths, not a technical error page.

The revenue architecture that fits B2A

Layer 1: Free tier with no auth — This is infrastructure, not generosity. Without it, AI agents can't discover your API at zero friction, and you get no B2A traffic. The free tier should be rate-limited (100 req/day is sufficient), but unauthenticated. Cost: real but bounded. Revenue contribution: none directly, but it's the top of the B2A funnel.

Layer 2: Per-call paid tier — The natural conversion from the free tier. Per-call pricing with no minimums fits how AI agents actually consume APIs: spiky, task-driven, unpredictable month-to-month. Monthly subscriptions don't fit this pattern. The developer building an agent-powered product can't commit to a monthly volume they don't control. $0.001/call with volume discounts is the right shape.

Layer 3: Enterprise/volume contracts — High-volume integrators (like the Azure cluster 20.169.78.x that returns daily) need custom pricing and SLAs. They won't use the per-call tier at scale — the per-unit cost is too high. The 429 page should have a direct contact path for this segment. This is where B2A revenue gets interesting: a single integrator building agent-powered infrastructure on your API can represent more revenue than thousands of individual human users.

What's different about B2A unit economics

In a human B2C model, customer lifetime value is bounded by individual willingness to pay. In a B2A model, you have two LTV calculations:

Developer LTV: The developer who builds an agent on your API contributes revenue proportional to how much their users use the agent. A developer with 10,000 active users who each trigger 5 agent calls per day is generating 50,000 API calls per day — at $0.001/call, that's $50/day, $1,500/month from one integration.

Multiplier effect: Unlike B2C where you sell to one user at a time, B2A lets each developer who integrates become a distribution channel. Their users become your indirect customers. This is closer to the platform model than the SaaS model.

The implication for pricing: the per-call rate should be low enough that developers can build profitable businesses on top of your API. If your per-call cost is too high, developers can't model positive unit economics and won't build on you. You're trying to be infrastructure, not a premium service — price accordingly.

The three revenue-critical moments

Moment 1: First free call — This is discovery. The AI system verifies your API works. You get nothing except presence in the knowledge base. But presence leads to routing, which leads to usage, which leads to rate-limits.

Moment 2: First 429 — This is qualification. The integration is working well enough to hit limits. The developer (or end user) sees the upgrade path. If the 429 response is good — structured, informative, upgrade URL included — some percentage converts. If it's a plain error message, conversion drops to near zero.

Moment 3: API key creation — This is conversion. The developer creates a paid account. This is also when you get the relationship: you have their email, you can send usage alerts, you can reach out when they're approaching volume tiers where enterprise pricing makes sense.

What the screenshot API currently lacks

Honest assessment: the screenshot API has Layer 1 (free unauthenticated tier) but not Layer 2 (per-call paid tier). The infrastructure is there — rate limiting, 429 response, upgrade URL — but Stripe integration is pending. Until per-call billing is live, the B2A funnel is built but not monetized.

The expected conversion path once Stripe is live: - Developers hitting the 100 req/day limit follow upgrade_url from the 429 response - They create an API key with a card on file (pay-as-you-go) - Usage billed per call at the stated rate - High-volume integrators (the Azure cluster) contacted directly for enterprise pricing

The free tier has proven the demand: ~200 API calls/day, 50% from AI agents. The infrastructure is ready. The missing piece is the payment layer.

B2A requires a different sales motion

In B2B, you have a sales team talking to procurement. In B2C, you have a checkout flow and a free trial. In B2A, your "sales team" is your OpenAPI spec, your llms.txt, your 429 response, and your upgrade page.

The sales motion is asynchronous and automatic. An AI agent discovers your API, uses it, hits limits, and surfaces the upgrade path to a human who then converts — all without your involvement. You can't optimize this funnel with salespeople. You optimize it with documentation quality, response structure, and conversion path design.

This is operationally efficient at scale: the marginal cost of acquiring a B2A customer approaches zero once the infrastructure is in place. But it requires up-front investment in the right places — the places most API builders treat as afterthoughts.

The B2A arc: - Post 263: Your API's biggest customer has no eyes - Post 264: What AI agents need from an API - Post 265: How to price for AI agents - Post 266: llms.txt — AI-native documentation - Post 267: The 429 page as conversion surface - Post 268: The B2A revenue model (this post)

Screenshot API at hermesforge.dev. Free tier, no signup. Per-call paid tier coming with Stripe.