API-Led Architecture

Organizing integration into system, process and experience APIs to decouple consumers from systems of record, with governance and reuse metrics.

Intent

Popularized by MuleSoft, API-led connectivity layers integration into reusable APIs:

Layer Purpose Example
System APIs Unlock systems of record with stable interfaces customer-sapi over SAP, policy-sapi over a mainframe
Process APIs Orchestrate and combine data for business processes onboard-customer-papi
Experience APIs Shape data for a specific channel mobile-account-eapi
flowchart TB
  M[Mobile app] --> E1[Mobile experience API]
  W[Partner portal] --> E2[Partner experience API]
  E1 --> P1[Customer onboarding process API]
  E2 --> P1
  P1 --> S1[CRM system API] --> CRM[(Salesforce)]
  P1 --> S2[Core banking system API] --> CORE[(Mainframe)]
  P1 --> S3[KYC system API] --> KYC[(KYC vendor)]

Real-world example

An insurer exposes its policy administration mainframe through a system API once. Web quoting, the broker portal and the claims app reuse it through process APIs, so mainframe changes are absorbed in one place.

Pros and cons

Pros Cons
Reuse and faster delivery of new channels Layer overhead and latency for simple cases
Shields consumers from legacy changes Risk of a central integration team bottleneck
Clear ownership and governance Can become an ESB in disguise if logic accumulates centrally

Interview questions

How is API-led different from an ESB?

API-led promotes decentralized, product-owned APIs with clear layers and self-service consumption via an API catalog. An ESB typically centralizes routing, transformation and business logic in one middleware team and runtime.

How do you measure the success of API-led integration?

Reuse rate of system APIs, time to deliver new channel integrations, number of point-to-point integrations retired, API consumer satisfaction and incident rates.