AI-Assisted Prior Authorization Workflow
Designing Human-Centered AI for Clinical Operations
Artificial intelligence has the potential to transform healthcare operations, but successful implementation requires much more than introducing a new model or automating an existing task. It requires thoughtful workflow design, clear product boundaries, stakeholder trust, and a deep understanding of how technology fits within existing clinical and operational processes.
This project explores how AI can be responsibly integrated into the administrative intake phase of a prior authorization workflow. Rather than attempting to automate medical necessity decisions, the prototype demonstrates how AI can organize information, identify missing documentation, request targeted clarification, and prepare a structured intake packet that enables qualified reviewers to make faster, better-informed decisions.
Beyond the working prototype, the broader effort demonstrates the product artifacts that transform an AI concept into a solution stakeholders can understand, evaluate, and refine. It illustrates how executive alignment, product definition, and interactive prototyping work together to validate an AI-enabled workflow before implementation begins. The project includes an executive presentation describing the business case and adoption strategy, a detailed product definition documenting the workflow and implementation approach, and an interactive prototype that allows stakeholders to experience the proposed workflow firsthand. Together, these artifacts demonstrate not only what could be built, but how product leaders communicate a vision, build organizational alignment, validate product concepts, and support responsible AI adoption.
The Problem
Prior authorization requests frequently require multiple rounds of communication before they are ready for clinical review. Missing documentation, fragmented clinical information, and inconsistent submissions create unnecessary back-and-forth between provider offices, intake teams, and clinical reviewers, delaying care while increasing administrative burden across the healthcare system.
Artificial intelligence is often discussed as a way to automate these processes. In practice, however, the greatest opportunity frequently exists before any clinical decision begins. Many requests simply need to be organized more effectively, checked for completeness, and routed appropriately before they ever reach a clinician.
This project intentionally focuses on that administrative preparation rather than clinical decision-making. By improving information readiness, AI can reduce avoidable friction while preserving the expertise, judgment, and accountability of qualified healthcare professionals. The workflow demonstrates how AI prepares work for people rather than replacing the people doing the work.
Product Goals
The objective extends beyond demonstrating AI capabilities. It was to explore how product leaders might redesign an existing healthcare workflow to improve efficiency while maintaining trust, transparency, and human oversight.
The workflow was designed to:
Improve administrative readiness before clinical review begins.
Extract meaningful clinical and administrative information from unstructured submissions.
Identify missing documentation and information gaps early in the process.
Generate targeted clarification questions rather than generic requests for additional information.
Route requests through clearly defined administrative pathways.
Prepare a structured intake packet that supports qualified clinical reviewers.
Preserve clinical judgment by maintaining clear boundaries between administrative workflow and medical decision-making.
Beyond the workflow itself, the project also explores the broader organizational challenges of AI adoption by demonstrating how executive communication, stakeholder alignment, workflow design, implementation planning, and interactive prototyping work together to support responsible digital transformation.
What This Project Reinforced
Developing this prototype reinforced another principle that has consistently shaped my approach to healthcare product development: artificial intelligence creates the greatest value when it improves workflows rather than attempting to replace expertise.
As implementation becomes easier, the responsibility of product leaders expands rather than shrinks. Building AI-enabled capabilities is no longer the primary challenge. The more difficult work lies in understanding existing workflows, identifying where AI genuinely reduces friction, establishing appropriate guardrails, preserving human accountability, and designing experiences that clinicians, operational teams, providers, and leaders are willing to trust.
This project also reinforced that successful AI adoption extends well beyond technical implementation. Organizations must build confidence among stakeholders by clearly communicating what the technology does, where it fits within the workflow, where human oversight remains essential, and how the solution improves rather than disrupts existing ways of working.
The most valuable AI products are rarely those that automate the greatest number of tasks. They are the ones that allow people to spend less time coordinating information and more time applying their expertise where it has the greatest impact.
AI Product Design
This project was intentionally designed to demonstrate more than a working AI prototype. It illustrates the sequence of product artifacts that help transform an initial concept into a solution stakeholders can understand, evaluate, and refine.
Rather than beginning with implementation, the process starts by building organizational alignment, defining the workflow, and validating the product concept before development begins.
The work is presented through three complementary artifacts:
Executive Presentation
Purpose: Communicate the business problem, product vision, AI strategy, stakeholder concerns, adoption approach, and expected organizational value.
This presentation is intended for executive sponsors, clinical leadership, and operational stakeholders. It demonstrates how product leaders communicate AI initiatives, establish appropriate expectations, and build organizational support before implementation begins.
Prototype Definition
Purpose: Define the scope, workflow, user experience, and implementation approach for a functional prototype that demonstrates the proposed AI-assisted workflow and enables stakeholder evaluation before production development begins.
This document represents the bridge between product strategy and prototype implementation. It translates the product vision into a working concept by defining the workflow, AI and human responsibilities, guardrails, conversation design, conditional routing, acceptance criteria, and implementation approach. Rather than specifying a production-ready system, it provides the direction needed to build a realistic prototype that stakeholders can evaluate, refine, and validate before further product investment.
Interactive Workflow Prototype
Purpose: Allow stakeholders to experience the proposed workflow through a working demonstration.
Using synthetic healthcare scenarios, the prototype illustrates how AI can organize information, identify documentation gaps, request targeted clarification, and prepare requests for human clinical review while preserving appropriate human oversight.
The prototype is intended to facilitate discussion, gather feedback, and validate workflow design rather than demonstrate production-ready automation.
Launch Interactive Prototype (Coming Soon!)
Technologies Used
Google Cloud Platform
Dialogflow CX
Google Gemini
Prompt Engineering
AI Workflow Orchestration
Conversation Design
Session Parameters & Conditional Routing
Human-in-the-Loop AI
Healthcare Workflow Modeling
Product Workflow Design
Responsible AI & Guardrail Design
Executive Product Strategy
Digital Transformation