Improving Referral Submission Efficiency & Accuracy

Aug 2025 – Feb 2026

01 · THE PROBLEM

Manual work across portals was unsustainable

This work sits in health technology: a setting where referral volume, compliance expectations, and patient experience all matter. Referral coordinators were spending a significant amount of time submitting referrals across numerous portals. The process was highly manual, varied by coordinator, and prone to data entry errors, leading to cases of rework, delays and patient dissatisfaction.

At a glance

  • Highly manual steps spread across many systems and portals
  • Inconsistent process by coordinator, increasing error risk
  • Rework, delays, and frustration for coordinators and patients

02 · DISCOVERY & RESEARCH

Understanding the real workflow

To understand the core problem, I worked very closely with a large group of referral coordinators to observe and document their end-to-end workflow.

Interviews & shadowing

Conducting interviews and shadowing to learn the role and surface pain points and process variation.

SIPOC mapping

Creating SIPOC diagrams to define scope, inputs, outputs, and dependencies.

Workflow visualization

Mapping detailed As-Is workflows: handoffs, decision points, and ownership, plus common error sources around data entry and provider-location mismatch.

This work revealed that a significant portion of time was spent repeatedly manually entering information across different systems, sifting through unstructured data to answer complex questions, and learning different workflows. This increased both the cognitive load of the referral coordinators as well as error and rework rates.

03 · DEFINITION & REQUIREMENTS

From insight to measurable intent

Upon my process discovery findings, I helped define and evaluate problem statements and success criteria centered on:

I translated my workflow insights into clear functional requirements including things like a point of singular data entry to initiate the process, automatic navigation, as well as pre-population of referral information.

04 · THE SOLUTION

AI-assisted browser extension

AUTOMATION · AI

Browser extension aligned to real workflows

As a result, a browser extension-based automation utilizing AI was implemented to support the requirements. The solution allowed coordinators to enter a patient's MRN only once, launching the navigation to the correct referral portal the patient required, and submitting all required information without manual entry.

The approach aligned with existing workflows, reducing change friction while addressing the highest impact pain points.

Browser extension AI Healthcare workflow

05 · OUTCOME

Impact on coordinators and quality of care

The solution significantly reduced the time it required to submit a quality referral and lowered the occurrence of errors caused by manual entry. To continue, the extension also provided a solution regarding provider-location mismatch as I ensured to include Human-In-The-Loop and other reinforcements after integrating with AI APIs and address normalization.

Impact included

The workflow documentation created during the discovery phase also became a reference for onboarding and future process improvements surrounding referrals.

What I Learned

Process discovery

Deep process discovery is critical before proposing solutions.

Meaningful impact

Targeted improvements can drive meaningful impact.

Adoption

Aligning solutions to preexisting workflows increased adoption.

Documentation

Clear documentation enables delivery and scalability.

Product design

Thoughtful product design creates intuitive, user-centered workflows.