API or screen automation: when to use each

Use a supported API whenever it covers the whole task: it skips the screen and has a defined contract. Use screen automation such as OpenAdapt only for work that's still only reachable through the screen, and check the result through the system's own data, which can be a read-only API even when the write isn't available.

Updated 2026-10-18

At a glance

A supported APIOpenAdapt
Best fitAny task the API exposes in fullTasks only reachable through the screen
When the screen changesNot affectedFinds the field again or stops; a person approves a fix
Proof the record savedThe response, plus a read-back if you add oneA second read of the record, where available
Cost per runSet by the service$0 in model fees in our runs; it needs a computer signed in
AccessOften gated by vendor agreements and per-site enablementYour staff's own session

Where an API is stronger

  • It skips screen rendering, focus and clicks entirely.
  • Requests and responses have a defined structure, with explicit status codes.
  • Authentication and conditional requests have standard meanings you can test.
  • The vendor supports the contract as the screens change.

When screen automation still makes sense

  • The API exists but doesn't cover the step, such as a referral entry or an attachment.
  • Write access needs a vendor agreement, fees or per-clinic enablement you don't have yet.
  • The system is an installed program or reached by remote desktop, with no API at all.

Even then, a read-only API, a report or a database role can check the save. That's the second read OpenAdapt uses. See How do you get data into an EMR without a usable API?.

Questions

Will OpenAdapt replace our integration?

No. Where an approved API covers a task, use it. OpenAdapt covers the remainder, and one program can mix the two: a step set up by hand can call an API directly while the rest goes through the screen.

Related: UiPath · Claims intake without re-keying · How it works · All comparisons

Sources

  1. IETF RFC 9110, HTTP Semantics (rfc-editor.org/rfc/rfc9110), messages, status codes and conditional requests.

See it on your process

Bring one task your team types every day. We'll tell you whether it fits, and what a test on your own system would cover.