An intake assistant that clears the overnight queue by 8am.
IntakeIQ is an AI intake agent that answers patient enquiries around the clock across phone, web form and message channels, triages each one against a rules-based urgency model, and books or escalates automatically. Anything flagged urgent pages an on-call nurse in real time instead of waiting in the same queue as routine questions.
The risk wasn't a slow reply. It was an urgent message sitting unread behind routine ones until morning.
The client is a healthcare practice that receives patient enquiries by phone, web form and messaging — at every hour of the day. Staff are excellent at triage, but they only start reading the queue when they log in. Overnight, every enquiry waited in one undifferentiated pile, urgent or not.
Patient enquiries came in at all hours through phone, web forms and messages, but nothing was actually triaged until a staff member logged in the next morning. That's fine when the overnight queue is three messages long. It breaks down once it's thirty, because an urgent case sitting at message twelve looks identical to a routine one at message two until someone reads both — and by 8am, someone should already have been called.
An AI agent framework handles conversation and understanding, a rules engine owns the urgency decision, and the practice's existing scheduling and messaging systems stay the source of truth for bookings.
The agent receives enquiries from phone, web form and messaging channels in one place, around the clock.
The language model's job is narrow: work out what the patient is actually asking and extract the details.
An explicit rules layer — not the model — classifies urgency, so every decision has a reason a clinician can read.
Routine requests are booked or answered; anything time-sensitive pages the on-call nurse in real time.
We built an intake layer that never stops reading the queue:
Let the language model decide what counts as urgent.
An auditable rules layer owns urgency; the model only interprets.
The easy way to build this is to hand the whole triage decision to a language model and trust its judgment on what counts as urgent — it's fast to prototype and feels impressive in a demo. In a clinical setting, an unexplainable model call on urgency is a liability nobody can sign off on if it's ever wrong. We built the urgency decision as an explicit, auditable rules layer instead, and used the model only for understanding what the patient is actually asking — so every escalation has a traceable reason a human can review, not a black-box judgment call.
Working software early, then the full build — with the client using it at every step.
We worked through real (anonymised) enquiries with staff to define exactly what makes a case urgent.
The agent went live on the web form first, with every triage decision reviewed by staff.
Phone and messaging added, real-time paging to the on-call nurse, and automatic booking for routine cases.
Triage rules documented and editable by the practice, with a full log of every decision.
Understanding language is where models shine. Making safety-critical calls is where you want rules a human can read and sign off.
Starting with the web form and reviewing every decision let staff shape the rules before the agent handled phones.
Next is summarising each overnight conversation into the patient record, so the morning team starts with context rather than a transcript.
Does intake need an AI agent, or would fixed rules do the job?
Our free tool answers it with your own figures — in your browser, no sign-up.
More on automation for healthcare & clinicsA regional courier planned every route by hand each morning, delaying the first pickups.
Every return needed three back-and-forth emails before a label was issued.
Reconciliation pulled numbers from five systems into a fragile spreadsheet.
We'll tell you whether it's worth automating, and what it would take. A founder or senior engineer replies within one working day.
