MEDICALCITY · The AI Factory for Healthcare

From everyday needs to a multi-agent system.

Healthcare is our starting point. The AI Factory turns healthcare teams’ tasks into support to design, develop and test. Health, food, housing and mobility guide how we scope needs. We build the digital infrastructure and develop physical execution interfaces within separate robotics pilots.

For healthcare teams

Care staff · Medical secretaries · Administrative teams · Healthcare facilities · Physicians

Four pillars of life

From an everyday need to a scope for testing.

For each need, the AI Factory turns a task into a project to scope with the team. Uses related to food, housing and mobility remain to be defined and developed.

  1. Health

    Choose a file or summary to prepare, specify authorised sources and identify the professionals who will check the result.

  2. Food

    Scope a draft meal-request summary: sources, constraints and review by the team before taking the next step.

  3. Housing

    Define one step in a residential admission request or support in a person’s living environment, its reference documents and approval responsibilities.

  4. Mobility

    Scope a draft checklist of travel-related documents: necessary information, people to contact and review by the team.

Teams specify sources, constraints and approvals. Agents prepare; the people responsible check, correct and decide what happens next.

From digital work to physical execution

Two infrastructures. Human responsibility.

MEDICALCITY develops the digital infrastructure for multi-agent systems and the physical execution infrastructure for robots. The Factory defines their interfaces, scope and approval points around the work of healthcare teams.

Digital infrastructure for multi-agent AI

Connect specialised agents to AI models, reference knowledge, memory and business software. Workflows organise document preparation, administrative processes and follow-up, with access and approvals defined for each project.

Research direction: “Multi-Agent AGI”, towards artificial general intelligence. A development ambition; the systems presented rely on specialised agents.

Infrastructure for robotic execution

Prepare robot connections, operator supervision and execution feedback. Each pilot must confirm hardware capabilities, permitted tasks, site conditions and ways to stop or take over the work.

Robotics: pilots in preparation. Drones: a longer-term development prospect. Scope, timing and service conditions remain to be confirmed for each project.

From a request to a supervised robotic task.

Proposed pilot scenario, to be validated in a trial: a team requests that small supplies be moved to a defined receiving point. This is a material-handling task, without a care procedure or physical contact with a person.

  1. The agents prepare

    Using the team’s request, they propose a task sheet specifying the items, route, recipient and missing information.

  2. The operator authorises

    The operator checks the information, equipment, route, execution conditions and stopping arrangements before deciding whether to start.

  3. The robot executes under supervision

    Within the pilot, the robot receives only the authorised task. The operator supervises and takes over when necessary.

  4. The team confirms the result

    The recipient checks receipt. Agents can prepare a report; the designated person checks the facts and approves follow-up.

An AI draft never triggers a physical action. If information is missing or a condition is not met, the task waits for clarification and a human decision.

Design. Develop. Test. Deploy.

The support described is for healthcare teams in France. Functions, connections and scope are confirmed before setup. Robotic integration is a separate pilot, with the operator deciding when to start a task.

  1. Define the need

    Choose a recurring task linked to health, food, housing or mobility and identify the people involved. With the team, specify authorised sources, constraints, the result to test and approvals. Distinguish digital preparation from any proposed physical pilot.

  2. Design the system

    Define agent roles, reference knowledge, workflows and human approvals. Identify the interfaces required.

  3. Develop and connect

    Configure agents and tools, develop the agreed connections and organise access. For a robotics pilot, prepare hardware and operator interfaces.

  4. Test with the team

    Check results, errors, access and human takeover within an agreed scope. Physical trials depend on the hardware and site conditions.

  5. Deploy under supervision

    Once the trial is approved, agree on service conditions, responsibilities and follow-up. The operator retains authority to approve each robotic task.

  6. Support and refine

    Gather feedback, track corrections and develop the scope with the team. Check any new capability or connection before use.

A verifiable scope, from software to pilot.

Roles and deliverables

Agent tasks, expected documents or task sheets, usable reference sources and the person responsible for final approval.

Data and interfaces

Necessary data, access, hosting, confidentiality and business connections. Robotic interfaces are defined and tested within the pilot.

Approval and taking over

Steps requiring human agreement, handling missing information, stopping arrangements and takeover by a professional or operator.

Evaluate quality, errors and total preparation, review and supervision time against usual practice. Agree on arrangements and any commercial terms before the trial. Medical decisions and signatures remain with authorised professionals.

Which need should become your first project?

Describe a task related to health, food, housing or mobility, your tools and the people involved. Together, we will define a scope for testing, sources and human approvals. Any proposed robotics pilot has its own scope.

Let’s discuss your daily work