Healthcare AdminRepresentative engagement

3 hours of inbox triage down to 20 minutes

Freeing front-desk capacity at a busy clinic with AI-powered email routing

Client:Primary care clinic
Size:4 providers, 6 admin staff
Timeline:2 weeks from kickoff to live

-90%

Triage time

3 hrs → 20 min/day

95%+

Routing accuracy

<4 hrs

Response time

Was 1-2 days

2 weeks

Setup time

The problem

What they were dealing with

The front desk was getting 100+ patient emails a day across appointment requests, prescription refill questions, billing inquiries, and records requests. The morning lead spent the first 3 hours of every day reading each email and either responding, forwarding to the right person, or queuing for follow-up. Patients waited 1-2 days for non-urgent responses; urgent things sometimes got buried under the volume.

The build

How we built it

  1. 1

    Pulled 4 weeks of historical inbox data

    Categorized every email into the actual buckets the team already used: appointments, refills, billing, records, urgent, other. The buckets matched the team's workflow, not some abstract taxonomy.

  2. 2

    Built the AI classifier

    A Claude-powered classifier that reads each incoming email, identifies the category and urgency, and labels it accordingly. Trained on the clinic's actual email patterns, not generic data.

  3. 3

    Routed to existing queues

    No new tools. The AI tags get translated into Gmail labels and the team's existing Practice Management filters route from there. Front desk sees a pre-sorted inbox; everyone else sees their own filtered view.

  4. 4

    Built escalation paths for the unclear cases

    Anything the classifier flags as ambiguous (low confidence, mixed intent, or the word "urgent") goes to a human review queue first. About 5-10% of emails — way more manageable than 100% of them.

Tools and integrations used

Google Workspace (Gmail)Claude APIPractice Management filtersCustom classification webhook

The outcome

What changed

Triage time dropped from ~3 hours/day to ~20 minutes (just reviewing the ambiguous ones). Patient response time on routine inquiries dropped from 1-2 days to under 4 hours. The morning lead reclaimed time for actual patient-facing work. Compliance was preserved — no PHI ever leaves the clinic's Google Workspace, since classification runs through their existing Workspace tenant via API.

Drowning in inbound emails?

High inbound volume, repetitive triage, time-sensitive responses — this pattern shows up in dental practices, property management, legal intake, and small ops teams everywhere. If it sounds familiar, book a free 30-minute consultation and we'll scope what your version would look like.