From drowning in messages to delighting patients

The situation when we first met

When this imaging center first reached out, their challenge was straightforward: volume. Over 600 emails landing in their inbox every day. A website chat that went unanswered after hours. The same questions coming in repeatedly—appointment times, prep instructions, records requests.

Their support staff were spending a good chunk of their day on routine tasks, which meant less time for the cases that genuinely needed human attention. Response times had stretched longer than they wanted, and they knew there was a better way to handle the workload.

The solution: two products, one unified approach

We tackled the problem in two phases, deploying a pair of AI systems that work together to cover both real-time and asynchronous patient communication.

An AI assistant that sounds like your best staff member

We started with the website. Patients were landing on their site at all hours—early mornings, late nights, weekends—and finding nobody to help them. The old chat widget was basically a contact form with extra steps.

We deployed an AI chat assistant trained on their specific services, locations, prep instructions, and policies. It doesn't just spit out generic answers. It knows the difference between their Glendale and Burbank locations. It knows which scans require fasting and which don't. It sounds like someone who actually works there—because we trained it on how their best staff members communicate.

Voice that actually works

Not everyone wants to type. Some patients—especially older ones—prefer to just talk. So we added voice. Real-time speech recognition, natural responses, and the same knowledge base powering the text chat. Patients can ask questions out loud and get spoken answers back, just like talking to a receptionist.

The voice system handles the back-and-forth naturally. It understands accents, handles interruptions, and knows when someone's finished speaking. It's not the robotic voice assistant people dread—it's conversational.

Secure access to real patient data

Here's where it gets interesting. Patients don't just want generic information—they want to know about their appointment, their results, their balance. But healthcare means HIPAA, and HIPAA means you can't just hand out patient data to anyone who asks.

We built a secure verification flow. When a patient asks about their specific appointment or records, the system sends a one-time code to their phone via SMS. Once verified, they can access their information—appointment times, prep instructions, even past visit history. All logged, all auditable, all compliant.

• SMS verification through Twilio before any PHI is shared
• Real-time appointment lookups from their scheduling system
• Complete audit trail of every verification and data access

Seamless handoff when humans are needed

Not everything can be automated, and that's the point. When a conversation needs human judgment—a complicated scheduling conflict, a sensitive medical question, an upset patient—the AI knows to step back. It hands off the conversation to staff through FrontApp, complete with the full chat history and context.

Staff don't have to ask “what were you calling about?” They already know. The patient doesn't have to repeat themselves. It's a warm handoff, not a cold transfer.

Taming the email flood

The chatbot was working. Patient satisfaction on the website was up, and after-hours inquiries were finally getting answered. But the email inbox? Still a disaster. 600+ messages a day, and staff were spending hours just reading, sorting, and routing them before they could even start responding.

Phase 2 tackled email. We deployed EmailTag—our AI-powered email triage system—directly into their FrontApp inbox. Now, every incoming email gets read, understood, tagged, and routed automatically. Routine requests get drafted responses. Complex cases get flagged for staff with full context attached.

Smart document handling

Medical imaging means paperwork. Lien documents from attorneys. Records requests from other providers. Faxes that somehow still exist in 2024. Most of these come as attachments, and reading them manually was eating up hours every day.

EmailTag doesn't just see an attachment—it reads it. OCR extracts the text. Barcodes get scanned for patient IDs and lien types. Signatures are detected and validated. If a document has been modified (strikethroughs, whiteout, alterations), the system flags it for review. Fraudulent or non-standard liens get caught before they cause problems.

Responses that sound right

Automated doesn't have to mean robotic. We worked with their team to build response templates for every common scenario—appointment confirmations, records release instructions, billing inquiries, lien acknowledgments. Each template uses their language, their tone, their specific policies.

When an email comes in asking about prep instructions for an MRI, the system doesn't just recognize the intent—it drafts a response with the right instructions for that specific scan type, personalized with the patient's name and appointment details. Staff can review and send with one click, or edit if needed.

Better together

The real magic happened when both systems started working together. The chatbot handles website visitors and after-hours inquiries. Email automation handles the inbox during business hours. Both feed into the same CRM, so staff have complete context regardless of how a patient reached out.

A patient might start a conversation on the website chat at 10pm, get verified and check their appointment, then send a follow-up email the next morning asking to reschedule. The email system sees the chat history. It knows who this patient is. It can draft an appropriate response with the right scheduling options. No context lost, no patient repeating themselves.

What the staff actually think

Here's what surprised us most: the staff love it. We expected some resistance—“the robots are taking our jobs”—but instead we heard relief. The scheduling coordinator who used to spend her mornings just sorting emails now actually has time to call patients with complex cases. The billing team can focus on negotiations instead of answering the same balance questions fifty times a day.

One staff member put it simply: “I finally feel like I'm doing the job I was hired to do.” That's the goal. Not replacing humans, but freeing them to do human work.

What's next

This isn't a “set it and forget it” deployment. We're continuously improving the system based on real conversations and feedback. The knowledge base gets updated as policies change. New response templates get added as patterns emerge. The AI gets smarter because it's learning from how staff handle edge cases.

Next on the roadmap: deeper scheduling automation (patients booking and rescheduling directly through chat), expanded SMS support for appointment reminders and follow-ups, and integration with their patient portal for even more self-service options.