01 Intro
APIs can be more than plumbing. Stripe, Twilio, & Plaid offer a model for being the pipe rather than owning the faucet. Infrastructure businesses in payroll, healthcare, & logistics can pursue the same strategy: embed a service in someone else's workflow. A library import is only the beginning; production adoption still requires integration, security review, and operational support.
This essay breaks down why API-first is a strategy, how it compounds, & which categories are still up for grabs.
02 When Embedding Can Help
• Embed in a workflow – a developer can champion adoption, but a production integration needs ongoing maintenance
• Scale with demand – under usage-based pricing, revenue can rise or fall with billable volume; discounts and pricing changes also matter
• Make reliability visible – a Powered By badge can identify a provider, but it is not proof of service quality
• Build switching costs – replacing an integration can be expensive, but customers can still leave, fail, or reduce usage
An embedded customer can distribute your service through their product. That does not make every customer a reseller, or mean every transaction on a platform uses the same provider.
03 Patterns of API-First Winners
- Developer-led adoption
Self-serve docs and sandbox keys can help developers evaluate a service before a sales conversation. Regulated or enterprise buyers may involve procurement from the start.
- Usage-based pricing
Charge per call, shipment, or payroll run. Expansion is possible, not guaranteed. Measure net revenue retention as current-period revenue from a fixed starting customer cohort divided by that cohort's baseline revenue, including expansion, contraction, and churn but excluding new customers. State the periods and revenue basis; customer counts and revenue retention answer different questions.
- Coverage as moat
Integrations with legacy systems can be the hard part: bank connections for Plaid, healthcare interoperability for Redox, carrier integrations for Shippo. Breadth may improve defensibility when connections are reliable and useful to buyers. A connector count alone says little about supported operations, data quality, or maintenance cost.
04 Vertical Cheat Sheet
• Payments – Stripe, Adyen → card processing, payouts → embedded in checkout; merchants still have compliance responsibilities
• Comms – Twilio, MessageBird → SMS, voice, WhatsApp → delivery quality, consent requirements, and service terms matter
• Fin-Data – Plaid, Belvo → bank connections → coverage and supported data vary by institution and market
• Payroll / HR – Finch, Check → HRIS data or payroll infrastructure → permissions, supported systems, and pricing vary by product
• Healthcare – Redox, Particle → EHR interoperability → regulatory burden plus hospital integrations
• Logistics – Shippo, Project44 → shipping and visibility infrastructure → reliability targets should reflect the operational cost of delays
05 Distribution Through Integrations
These are possible distribution mechanisms, not guaranteed conversion paths:
• Embedded reach – a fintech can bring its payout provider into a new customer workflow
• Ecosystem pull – supporting a requested HRIS can make an API useful to downstream software vendors
• Partner discovery – integration documentation can introduce merchants to a shipping provider
Enterprise sales, procurement, and a technical evaluation may proceed together. An integration does not eliminate the sales process.
06 Counter-Arguments and Failure Modes
• Graduation risk – as a customer's bill grows, bringing a capability in-house or adding another provider may become attractive. Broader functionality and volume pricing can help, but neither prevents churn.
• Slow ramp – usage starts small. You need patient capital or a vast long-tail funnel.
• Commoditisation – pass-through APIs invite price pressure. Useful coverage, compliance support, and developer experience may help differentiate a provider, but each costs money to maintain.
What if this is wrong? If integration costs collapse faster than usage grows, vertical SaaS with embedded APIs could capture the margin. Bet accordingly.
07 Where the Next Giants Will Emerge
- High integration tax – legacy systems, regulation, messy endpoints
- Obvious meter – a billable event you can count
- Need for software distribution – teams prefer shipping features over chasing vendor contracts
FHIR-based healthcare integration, emissions accounting, and usage-metered AI inference are candidates to test against these criteria. A docs page alone does not make a provider indispensable; the test is whether it solves a costly problem better than alternatives at sustainable margins.
08 Meta-Layer & Brand Control
Invisible infra is powerful, but total invisibility cedes narrative control to the platforms that embed you. The meta-layer lesson: own the function call and the story. Stripe's API reference reads like marketing copy; Twilio's developer evangelism graduates tinkerers into champions. Pipes need brand.
09 Closing Thoughts
Owning the function call can be an alternative to owning the interface. The business case depends on recurring demand, reliable delivery, pricing power, and customer retention—not invisibility by itself.
So ask yourself: compete for UI real estate, or become the protocol the pixels rely on?