AI Web Agent for a Multi-Location Clinic

A static brochure website that could not answer patient questions or take bookings — every enquiry went to a phone line that was busy after hours, so for roughly a third of the day the clinic's only public-facing channel was effectively closed, and the site itself functioned as little more than a digital business card with an address and a phone number on it.

Patients asked the same handful of questions repeatedly — clinic timings, whether a specific insurance provider was accepted, Saturday availability, parking, whether walk-ins were possible — in a mix of English and Tamil, and the front desk fielded every single one of them manually, call after call, with no way to deflect the routine ones to free up time for genuinely complex requests.

No structured record of who enquired or why existed anywhere in the business: a missed call at 9pm simply vanished into a voicemail box that may or may not get checked the next morning. There was no way to know how many patients tried to reach the clinic outside working hours, no way to measure it, and therefore no way to argue for fixing it.

New patients frequently abandoned the booking process entirely after failing to reach anyone by phone on the first or second attempt, and had no self-serve way to check basic eligibility information — like whether their specific insurance plan was in-network — before deciding whether it was even worth calling in the first place.

Front-desk staff spent a significant, measurable share of each morning working through manual call-backs to voicemails left overnight, which delayed same-day booking confirmations for patients who had called that morning and created a visible backlog before the clinic doors had even opened for the day.

Multiple satellite locations across the city meant patients frequently called the wrong branch by mistake, adding an extra transfer step to an already slow process and meaningfully increasing the chance that the caller simply gave up rather than wait to be redirected to the correct front desk.

Clinical staff were regularly pulled away from patient care to answer front-desk phone calls during peak hours when reception was overwhelmed, which both slowed down in-person consultations and left phone enquiries answered by whoever happened to be free rather than someone trained to handle them consistently.

There was no mechanism at all for capturing interest generated by word-of-mouth referrals or the clinic's existing Google Business listing — a referred patient searching the clinic name late at night found the same static page as everyone else, with no way to act on their interest until business hours resumed. Management had no reliable way to quantify how much of this word-of-mouth interest was being lost simply due to the timing mismatch between when people searched and when the clinic could actually respond.

Built an AI Web Agent on the clinic's own domain, trained only on approved service descriptions, per-location opening hours, the specific list of accepted insurance providers, and internal clinic policies — nothing invented, nothing assumed, and nothing pulled from outside that approved knowledge base.

Wired live Google Calendar availability directly into the chat interface so patients could see genuinely open slots in real time and confirm a booking without ever needing to pick up the phone, receiving an instant SMS confirmation with the date, time, and location the moment they did.

Enabled automatic language detection at the message level so each conversation runs naturally in the patient's own language — English or Tamil — switching mid-conversation if needed, with zero separate pages, zero translated content to maintain, and zero additional configuration required from clinic staff.

Configured location-aware routing so that a patient asking about the nearest branch, its specific hours that day, or its parking situation receives an answer scoped precisely to that location rather than a generic, unhelpful clinic-wide response that might apply to a different site entirely.

Every single conversation is automatically scored and stored as a structured lead with a full transcript attached, tagged by urgency level and preferred location, so that front-desk staff open each morning to an organized, prioritized list instead of an unsorted stack of overnight voicemails.

Complex, sensitive, or clinical cases — anything the agent cannot answer confidently from within its approved knowledge boundaries — escalate immediately and automatically to on-duty staff with a concise one-line summary of exactly what the patient needs, so no context is lost in handoff.

Rolled out a lightweight monthly review process with clinic management, walking through the most common new questions patients asked that month, so the agent's knowledge base keeps pace continuously with new services being added, price changes, and predictable seasonal shifts in patient demand.

Integrated the agent with the clinic's existing Google Business listing and referral flows, so late-night searches and word-of-mouth interest convert into a captured, scored lead immediately rather than waiting for the clinic to reopen the next business day.