Skip to content

Selected work / 02 / Workflow automation

When professional verification becomes an automation problem

AEMI · Independent engagement · 2024–2025

The decisionAny check that failed or was uncertain went to a person; the automation never approved or rejected one.

A verification workflow that automated the checks it could do confidently and routed the rest to a person, so automation never made the final call on an uncertain case.

DeliveredWorkflow designed, built and handed over. Performance figures are not published because their source data and definitions have not been recovered.

CHECKVERIFIED?
Fig. 3 Cards the check can confirm go to the verified stack. A card it cannot vouch for goes to a person.
  1. Registration intakeProfessional identification and registration details through an online form.
  2. Automated verificationChecks run in Microsoft Power Automate.
  3. If confident: verifiedThe result is recorded automatically.
  4. If failed or uncertain: manual review (person)Routed to a person with an email notification. Never approved or rejected automatically.
  5. RegistrationProcessing continues once the case is decided.
Verification workflow with a human-review route for uncertain cases. Reconstruction of the delivered workflow. Tools are generalised and no client artefacts are shown.

Context

An independent engagement for AEMI, in healthcare professional registration and verification, with multilingual registration and practitioners across several professions, including doctors, pharmacists and physiotherapists.

The constraint

Verification decides who may proceed, so a wrong automated answer is worse than a slow manual one. Registration details arrived in different forms across professions and languages, and some verifications could not be completed with confidence.

My role

Workflow design, requirements engineering and automation implementation: analysing the verification process, designing the workflow and its exception paths, building it in Microsoft Power Automate (cloud and desktop), and documenting it for handover.

The consequential decision

Because a wrong automated answer is worse than a slow manual one, the workflow was designed for uncertainty from the start: a verification that failed, or could not be completed confidently, went to manual review with an email notification.

What changed

  • Registration intake through an online form, collecting professional identification and registration details.
  • Automated verification with Microsoft Power Automate.
  • Successful verifications recorded automatically.
  • Uncertain or failed verifications routed to manual review, with notification.
  • Downstream registration processing on the organisation’s website.

Evidence

Workflow design, automation flows and handover documentation, September 2024 to January 2025.

What the work taught me

Most of the design work went into the exception path: who is told, and who decides, when a check is unsure.