Aga Khan · current implementation
Healthcare AI Workflows
Routing patient inquiries and appointment requests that were previously handled manually. An early-stage engagement: the routing layer is the focus, and clinical judgement stays with clinical staff.
- Client
- Aga Khan
- Sector
- Healthcare
- Status
- Early implementation — scope still being defined with clinical teams
The situation
Patient inquiries and appointment requests were arriving through channels that all funnelled into manual triage. The cost of that lands twice — patients wait for a response to a question that may be routine, and administrative staff spend their day sorting requests rather than resolving them.
This is an earlier-stage engagement than the automotive work, and the honest description of it is a routing problem rather than a resolution problem. The question being answered is "where should this go," not "what should the patient do."
What made it hard
Clinical judgement is out of scope
Anything touching diagnosis, triage severity, or medical advice belongs with clinical staff. The system classifies and routes; it does not advise. That boundary is the design, not a limitation to be engineered away later.
Sensitive data handling
Patient information carries handling requirements that shape what the system is allowed to store, log, and pass between steps.
Conservative handoff
In a healthcare setting the cost of wrongly keeping someone in an automated flow is much higher than the cost of handing off unnecessarily. The threshold is set accordingly.
How it works
Inquiry received
Patient calls or messages with a question or request.
Intent routed
Request is classified and sent to the right department.
Handoff when needed
Clinical or administrative staff step in for anything sensitive.
What the system does
A patient calls or messages with a question or request. The request is classified by intent and routed to the department that owns it, rather than sitting in a general queue waiting for someone to read it and forward it on.
Where a request is sensitive, ambiguous, or clinical in nature, it goes to administrative or clinical staff with the context already captured — so the person picking it up starts with the request understood rather than starting the conversation again.
What it deliberately does not do
It does not give medical advice, assess urgency, or make any clinical determination. Those decisions stay with qualified staff. Building an AI layer in a healthcare setting means being precise about where its authority ends, and in this case it ends at routing.
What is still being defined
This is an active engagement with scope still being worked out alongside clinical teams. There is no results section on this page yet because there is not yet a result worth publishing — the measures that matter are being defined with the people who will have to live with them.
Where this stands
Figures below are labelled by what they actually are. Nothing here is presented as a delivered result unless it is one.
Engagement stage
Scope and success measures are being defined with clinical teams rather than assumed up front.
Published outcomes
This page will carry real figures when there are real figures. Until then it describes what is being built, not what has been achieved.
Stack
Working on something similar? I'd be glad to talk through it.
Book a Discovery Call