Unnat Patel

Work · 8 of 17 · Deep dive · 5 minutes

On this page
  1. Launch the console first for the single worst pain point, opened from the legacy tool, and make it primary only at about half feature parity.
  2. What I learned
  3. What I owned
  4. The situation

Service experience / Case study

One console for pharmacy customer care: about three minutes saved per contact in the pilot

Amazon Pharmacy customer care worked across a legacy agent tool plus chat and voice systems, while pharmacists and technicians juggled many separate tools. I was the product manager for a unified care console: I flipped the launch sequence to protect continuity of care, piloted it, and then shipped the capabilities that removed the biggest sources of transfers and hold time. The pilot saved about 3 minutes per contact across more than 6,000 contacts a month. AI-assisted routing saved 21 minutes per routed contact and eliminated about 1,000 transfers a month.

Amazon Pharmacy · 2025–2026 · from my work, written for public reading

01

1 / 9 · Executive brief & flow

Launch the console first for the single worst pain point, opened from the legacy tool, and make it primary only at about half feature parity.

  • Pain
    Prescription-status questions required a separate technician conversation
  • Sequence
    Console first for prescription status, deep-linked from the legacy tool
  • Pilot
    10 to 40 to 100 representatives to general availability
  • Capabilities
    Duplicate search, orders, medications, packets, chat, replacements
  • Routing
    Clinical-ticket context routes the contact to the appropriate team

Product workflow, shown through the decisions and handoffs.

02

What I learned

Sequence a platform migration around the worst customer moment, not the cleanest architecture diagram. Parity is a threshold you earn, and continuity of care is the constraint you never trade.

My reflection

State reporting automation

I also launched the customer-care console and personally led duplicate-account search, embedded chat, packet workflows and clinical-ticket routing.

Customer-care contacts often arrive with relevant history that the representative must reconstruct before helping. I envisioned assistance that brought useful context into view sooner and worked with customer-experience and contact-center partners to bring it into the product. Source-linked context mattered: a representative needed to inspect the basis for a suggestion, not take polished text on faith.

Around that core, I worked on self-service assistance, representative support and conversation review. Retrieval from approved knowledge can help with routine questions; suggested answers must remain distinguishable from actions; a human needs sufficient context when a conversation exceeds the assistant’s scope. Product evaluation included quality, latency, cost and correction effort. Clinical authority remained with qualified people.

Other work involved assistance with billing exceptions and prescription data entry. I separated proposing a correction from confirming its effect, and paired automation with quality evaluation and operator feedback. A higher automation rate without its error rate would have been an incomplete product claim.

A later account describes a voice handoff failure, transfer monitoring and a callback fallback. The operating lesson is to verify the handoff outcome, rather than rely only on component health.

Tools and methods: customer-care workflow design, product requirements, launch readiness and operator feedback. 11 Automation with visible boundaries

03

What I owned

I launched the customer-care console and personally led duplicate-account search, embedded chat, packet workflows and clinical-ticket routing, alongside the requirements, design and launch coordination.

04

The situation

At Amazon Pharmacy, generalists handled about 83% of contacts with the legacy care tool, the contact-center platform and a chat channel to technicians. To answer why a prescription was in follow-up, a generalist messaged a technician while the customer waited three to four minutes on hold. Pharmacists and technicians juggled many separate tools. The console’s goal was to handle a much larger share of clinical contacts.

The decision

I prioritized a narrower prescription-status workflow first, opened from the existing care tool, before expanding the console. Launch sequencing protected continuity of care.

What shipped after the pilot

CapabilityWhat it addressedEffect
Duplicate account search with match confidence13% of monthly contacts involved duplicate or partial profilesThe duplicate-search account reports less technician chat and about 2 minutes saved per affected contact; the affected contacts are a subset of overall volume.
Orders view and cancellationCapabilities the legacy tool never hadNet-new for care
Medications entity23% of contact volumeFull prescription-level detail in one view
Packet schedules, recovery and modificationsPacket contacts ran 18% longer for lack of visibilityFirst time care could act for packet customers since the acquisition
Chat in the console30% of contacts had no console accessThe chat-integration account reports about 60 seconds saved per affected chat contact
Clinical-ticket routingGeneralist-to-specialist transfersThe routing account reports about 1,000 avoided transfers per month and 21 minutes saved per affected contact

Results

The prescription-status pilot account reports about three minutes saved per affected contact across more than 6,000 monthly contacts. The roadmap called for making the console primary and retiring the legacy tool; that plan is separate from the pilot result.