Automate · AI Agents

AI that understands
patients
— and knows
what to do next.

Agents that understand a question, use approved healthcare knowledge, collect what is needed, perform configured tasks, and hand the conversation to the right team when a person should take over.

W
WisemelonPatient Concierge
Online
I have an appointment tomorrow. What should I bring? AI agentYour appointment is tomorrow at 10:30 AM. Please bring your previous reports, your current prescriptions and a valid ID. Can I change my appointment time? AI agentYes, I can do that. These are the times still open tomorrow.
Tue10:30 Tue12:00 Tue16:30
Appointment updated · 4:30 PM Message

What makes it an agent

More than answers. AI that can take action.

A chatbot replies. An agent is given a job, the knowledge to do it, the actions it is allowed to take, and somebody to hand it to when it cannot finish.

Understand

AI agent Given a job, not a prompt

Reason

  • ActPerform the supported actions instead of stopping at an answer.
  • EscalateBring the right person in when the configured process needs one.

The jobs

One platform. Agents for different jobs.

An agent that books appointments and an agent that qualifies enquiries are not the same worker with a different greeting. They have different knowledge, different actions and different people to escalate to.

  • Patient Concierge

    • Appointments
    • Preparation
    • Patient requests

    “What should I bring tomorrow?”

    Active

  • Care Assistant

    • Follow-up
    • Education
    • Reminders

    “How are you getting on with the instructions?”

    Active

  • Front Desk Agent

    • Requests
    • Information
    • Routing

    “Which department handles this?”

    Active

  • Support Agent

    • Service requests
    • Tickets
    • Status

    “Any update on my report?”

    Active

  • Lead Qualification

    • Enquiries
    • Qualifying
    • Routing

    “Do you do corporate health checks?”

    Active

  • Internal Assistant

    • Staff questions
    • Processes
    • Configured tasks

    “Who is the on-call cardiologist tonight?”

    Active

Routing

One agent doesn’t have to do everything.

A patient should not need to know which department owns their request. They write one sentence; Wisemelon routes the task to the configured agent, workflow or human team behind it.

Patient I was charged twice for my scan last week.

Patient agent Classified as a billing question
  • Scheduling agentNot this one — nothing about the appointment changed.
  • Billing agentChecks the charge, raises a refund request, tells the patient what happens next.
  • Support teamStanding by if the billing agent cannot close it.

The patient sees one conversation. Wisemelon sees three possible owners and picks the one the request belongs to.

For your teams

AI isn’t only for patients.

The people running the hospital ask the same question a hundred times a week. An internal agent can answer it — and start the thing they were going to do next.

Hospital front-desk staff at their workstations, working through requests on screen while patients wait at the counter
The question nobody should have to walk down a corridor to ask.
W
WisemelonInternal assistant
Who is the on-call cardiologist tonight? AssistantDr. Mehta, until 8 AM. Extension 4821. Which department handles a duplicate charge? AssistantBilling. I can raise the request for you now if you like. Request #2211 created

Inside the process

Give your AI a workflow to work within.

Workflow Builder defines the process. AI Agents handle the interaction and the configured tasks inside it. Neither one is useful on its own.

Workflow BuilderThe process

AI agentThe interaction

ActionThe thing that happens

Human teamWhen a person is needed

What the agent does with one request

  1. 01 · ReceiveA patient or a staff member starts a request.

    A message, a form, a workflow step — whatever the configured trigger is.

  2. 02 · UnderstandWork out what they are trying to accomplish.

    Not what they typed. What they want to have happen.

  3. 03 · CheckUse the configured knowledge, data and rules.

    Approved content and real patient context, not whatever the model remembers.

  4. 04 · ActPerform the supported action, or start the next step.

    Book, raise, update, route — within what it is allowed to do.

  5. 05 · EscalateHand over when the process says a person is needed.

    With everything it has already gathered attached.

In practice

What that actually looks like.

“I need to see a cardiologist.”

The agent works out the specialty, checks who is available, and offers the times that exist — rather than telling the patient to call the desk.

It moves the request toward completion, not toward an answer.

W
WisemelonPatient Concierge
I need to see a cardiologist. AI agentI found these cardiology appointments open this week.
Tomorrow16:30 Friday11:00 Friday16:00
Booked · Dr. Ananya Rao, Friday 11:00

Collect what the team needs before they have to ask.

A short configured set of questions, asked conversationally, turning into structured information the care team can read before the visit.

This is configured information collection, not diagnosis. The agent asks what it was told to ask, and records the answers.

  • Reason for visitChest discomfort
  • DurationAbout three weeks
  • Current medicationTwo, listed
  • Previous reportsNot uploaded yet
W
WisemelonPre-visit intake
AI agentWhat brings you in today? Some chest discomfort when I walk upstairs. AI agentHow long has that been going on? About three weeks. AI agentAre you taking any medication related to this? Yes, two.

Don’t stop at the first response.

“Any update?” is a question about a process. The agent can check where that process actually is, and move it on.

  • Request createdYesterday, 4:12 PM
  • With Medical RecordsSince 9:04 AM
  • Report ready11:10 AM
  • DeliveredSending now

The agent can read the state of a configured process and take the next step in it.

W
WisemelonSupport Agent
I requested my report yesterday. Any update? AI agentIt was ready at 11:10 this morning. I’m sending it to you now. Report sent · 2 pages

After the visit

Keep the journey moving after they leave.

Discharge is the point at which most follow-up quietly stops. A configured agent can keep going — checking in, sending what was promised, and escalating anything that sounds wrong.

  1. Discharge
  2. Follow-up
  3. Check-in
  4. Education
  5. Next step

Each stage is configured — what is sent, when, and what happens if the answer is concerning. An agent that hears something it was told to escalate hands over immediately. Patient Education supplies what it sends.

W
WisemelonCare Assistant
AI agentHow are you getting on with the care instructions from your visit? Mostly fine. The dressing keeps coming loose. AI agentHere is the dressing guide your team approved. I’ve also flagged this for the nursing desk. Flagged to Nursing · Ward B

Control

Define the job, the tools and the boundaries.

An agent is only as safe as what it is allowed to reach. Its knowledge, its actions, its escalation paths and its permissions are all things your team sets.

  • Agent
  • Knowledge
  • Actions
  • Escalation
  • Permissions
Patient Access Agent Goal — help patients with configured appointment and access-related tasks.

Knowledge

  • Hospital informationApproved
  • Appointment guidelinesApproved
  • Department informationApproved
  • Clinical recordsNot granted

Actions

  • Check configured availabilityAllowed
  • Start the appointment workflowAllowed
  • Create a requestAllowed
  • Cancel an admissionNot allowed

Escalation

  • Front DeskFirst
  • Scheduling teamThen

Powerful AI, bounded by your team. Staff & Access decides who can change any of this.

Connected

AI Agents don’t work alone.

An agent with no context is a chatbot. What makes it useful is everything around it — the patient record, the process, the ticket queue, the people.

What it draws on

  • Patient OSWho this patient is and what has already happened.
  • Approved knowledgeThe organisation’s own content, not the open web.
  • Workflow BuilderThe process the agent is working inside.
AI agentUnderstands, checks, acts

What it reaches

  • TicketingAnything it cannot close becomes somebody’s work.
  • Internal ViewsWhere the team sees what the agent did.
  • IntegrationsThe systems outside Wisemelon that have to know.

AI becomes useful when it is connected to the systems, processes and people around it.

A doctor talking with a patient at her bedside while nurses and reception staff work around them

AI handles the routine. People handle the human moments.

The goal was never to take people out of the patient journey. It is to give them back the hours currently spent answering the same question, so there is time for the conversations that genuinely need a person in the room.

An agent that hands over well is worth more than one that never hands over at all.

FAQ

Frequently asked questions

What are AI Agents in Wisemelon?

Healthcare AI agents are configurable AI systems that can understand requests, use approved information, perform supported tasks, interact with workflows and route work to people or systems. Wisemelon AI Agents handle defined patient, staff and operational tasks using approved knowledge, configured actions and workflow logic.

How are AI Agents different from a chatbot?

A chatbot generally focuses on responding to messages. An AI Agent can be configured to understand a request, use available context and knowledge, perform supported actions and continue a configured process.

Can AI Agents work with Workflow Builder?

Yes. Workflow Builder can define the process while an AI Agent handles configured intelligent or conversational steps inside that process.

Can AI Agents help hospital staff?

Yes. Internal agents can be configured to help staff find information, understand configured processes and initiate supported operational tasks.

Can an agent route a request to another team?

Yes. Configured routing and escalation paths can direct work to appropriate departments, agents or human teams.

Can AI Agents collect patient information?

They can be configured to collect information through conversational interactions and pass the resulting information into supported workflows or systems. This is configured information collection, not diagnosis.

Can I control what an AI Agent is allowed to do?

Yes. Agents can be configured around defined knowledge, actions, workflows, permissions and escalation paths.

Can AI Agents perform tasks instead of only answering questions?

Yes. Where supported and configured, agents can participate in actions such as appointment-related processes, requests, routing and other operational workflows.

Give every patient journey an intelligent layer.

Build AI Agents that understand the request, work through the configured process, take the supported actions and bring your team in when they are needed.

Back to Patient OS