Patient Portal

Conversational Scheduling: Chat-orchestrated appointment booking with guardrails

  • Role -  Senior Product Designer
  • Team - Product, Engineering, Design, Clinical operations stakeholders
  • Timeline - Shipped + Iterated
  • Platform - Web patient portal (chat widget experience)

Astrana’s patient portal supported core actions, but scheduling could still feel like navigation work: find the right place, choose the right provider, pick a time, and confirm details. I designed a conversational scheduling assistant that turns scheduling into a guided flow, using chat to orchestrate the journey while structured UI handles the high-precision steps.

At a glance

  • Problem: Scheduling is a high-intent task, but patients often face uncertainty and friction: who to see, where to go, and what time is available.
  • Solution: A hybrid conversational flow that guides decisions and progressively reveals structured choices.
  • Result: A faster path from intent to confirmation with fewer mis-bookings and fewer dead ends.

Scheduling is not a “read an FAQ” task. It is a transactional workflow where small mistakes create real consequences: wrong provider, wrong location, wrong time, or uncertainty about whether the booking succeeded. We needed to make scheduling feel:

  • Discoverable and low effort
  • Precise and reviewable
  • Safe, with clear boundaries and escalation cues

What success meant

  • Patients can start scheduling from a single entry point without hunting through navigation
  • The experience reduces cognitive load while keeping decisions explicit
  • The final booking is reviewable before confirmation
  • The system provides clear closure and next steps after booking

The hard constraints

  • Patient-facing, high-trust context with low tolerance for ambiguity
  • Scheduling requires precision, so freeform chat alone is not sufficient
  • The flow must support common paths and safe recovery (change date/time, reconsider provider)
  • Content must be accessible and readable, especially for high-stress users

The Flow

Chat orchestrates, UI executes

The design uses a hybrid pattern:

  • Chat drives the flow, confirms intent, and keeps momentum
  • Structured UI handles selection (provider, slots) and confirmation to reduce mistakes

This prevents two common failure modes:

  • A chat experience that feels vague or unreliable for transactional actions
  • A scheduling UI that feels like a maze for patients who just want to book quickly
1) Entry and orientation
The assistant introduces its purpose and offers suggested questions to reduce typing and ambiguity. It also sets clear boundaries for emergencies.
  • Sets scope and trust
  • Suggested prompts reduce effort
  • Clear invitation to start scheduling
2) Provider selection
Rather than asking the patient to navigate provider directories, the assistant presents a structured provider list with availability cues.
  • Guided disambiguation
  • Availability communicated without overwhelming detail
  • Familiar list pattern reduces friction
3) Mid-flow Correction Handling
Patients often change their mind after starting scheduling. The assistant supports “in-flow corrections” without forcing a restart, preserving context while re-opening the right decision step (provider selection).
  • Free-text correction recognized (“different provider”)
  • Returns to the correct step without losing progress
  • Keeps the experience calm and recoverable
4) Time slots and date control
The assistant presents available time slots in a high-scan list. Patients can choose another date without restarting the flow.
  • Quick scanning, low cognitive load
  • Date change without restarting
  • Clear tap targets for selection
5) Review before confirmation
Rather than asking the patient to navigate provider directories, the assistant presents a structured provider list with availability cues.
  • Pre-confirmation review prevents wrong bookings
  • Human-readable summary
  • Commitment is clear and intentional
6) Confirmation and recovery path
After booking, the assistant provides explicit confirmation and tells the patient what happens next, including how to change the booking.
  • Outcome is explicit
  • Sets expectation for confirmation email
  • Safe path to modify date/time

Design decisions that mattered

1) Hybrid pattern over pure chat
We intentionally avoided forcing precision actions into freeform chat. Selection and confirmation are handled with structured UI to reduce errors.
2) Progressive disclosure
The flow asks for only what is needed at each step. This reduces the feeling of a long form and keeps the patient moving.
3) Guardrails for a high-trust context
The experience includes:
  • Clear emergency disclaimer
  • Mandatory review step
  • Explicit confirmation with recovery options
4) Designed for correction, not perfection
Patients often change their mind mid-flow. The interface supports alternate date selection and makes rescheduling feel like a normal continuation, not a failure.

Outcomes

  • Reduced navigation burden by turning scheduling into a guided flow
  • Increased confidence through a mandatory review step
  • Reduced friction for common scheduling scenarios (provider, date, slot selection)

What I owned

  • Conversation flow and decision structure for scheduling
  • Hybrid UI patterns (lists, review cards, confirmation messaging)
  • Content and microcopy for clarity and trust
  • Safety guardrails and error prevention patterns
  • Usability validation inputs and iteration with engineering

Takeaways

Patient-facing conversational experiences work best when they do not pretend chat is magic. The assistant earns trust by combining a guided conversational flow with structured UI for precision, plus guardrails that prevent errors and create clear closure.