Multi-agent decision support

Multidisciplinary Board Platform

Every specialty’s view, prepared before the room convenes.

The Multidisciplinary Board Platform is designed to prepare case review for multidisciplinary teams. Specialist-perspective agents organise the evidence each discipline needs, show where views agree or diverge, and list what is still unknown. Board time goes to the decision, which stays with the clinicians.

HealthcareDescriptive working title

Demonstration

Demo with synthetic data

A synthetic colorectal case is read by one AI agent per department, from gastroenterology to genetics. Agreement, divergence, and missing items are made explicit, and a second view shows the meetings and their analytics.

All names, values, and records in this demonstration are invented. It illustrates the workflow and is not a screenshot of a deployed system.

Multidisciplinary Board Platform

Conceptual demo, synthetic data

Case prepared for the board

Patient summary

64-year-old, rectal adenocarcinoma, first MDT review

Synthetic patient, colorectal tumour board

Presentation
Rectal bleeding and change in bowel habit for 3 months
Performance status
ECOG 1
Comorbidities
Type 2 diabetes (metformin), hypertension
Tumour
Mid rectum, about 8 cm from the anal verge

Documents

  • D1Colonoscopy 21/08
  • D2Biopsy 28/08
  • D3MRI pelvis 03/09
  • D4CT chest and abdomen 21/07
  • D5Laboratory 02/09

Department agents

Each agent reads the case documents and prepares its department’s considerations.

  • Gastroenterology agent

    Ready
    Tool calls (2)
    • Read colonoscopy report D1
    • Read laboratory results D5
    • Circumferential mid-rectal lesion, biopsied.
    • Proximal colon not fully assessed because of poor preparation. Completion colonoscopy or CT colonography should be considered before surgery.

    D1

  • Pathology agent

    Ready
    Tool calls (1)
    • Read biopsy report D2
    • Adenocarcinoma, moderately differentiated.
    • Mismatch-repair (MMR/MSI) status is not reported. Immunohistochemistry should be requested, as it may change the options discussed.

    D2

  • Radiology agent

    Ready
    Tool calls (2)
    • Read MRI pelvis D3
    • Read CT chest and abdomen D4
    • MRI: features consistent with cT3 N1, mesorectal fascia clear.
    • No distant disease on CT, but the scan is from 21/07, older than 6 weeks. Updated staging may be needed.

    D3D4

  • Colorectal surgery agent

    Ready
    Tool calls (2)
    • Read MRI pelvis D3
    • Read colonoscopy report D1
    • Resectable on imaging. Sphincter preservation appears feasible at about 8 cm.
    • Raises neoadjuvant treatment before total mesorectal excision for discussion.

    D3D1

  • General surgery agent

    Ready
    Tool calls (2)
    • Read referral and history
    • Read laboratory results D5
    • ECOG 1. Diabetes and hypertension recorded as controlled.
    • Pre-operative anaesthetic assessment and stoma counselling to be arranged if surgery is planned.

    D5

  • Genetics agent

    Ready
    Tool calls (2)
    • Read biopsy report D2
    • Read referral and history
    • No family history of cancer is recorded in the referral. A three-generation family history should be taken.
    • If the tumour is MMR deficient, germline testing for Lynch syndrome and genetic counselling could be offered.

    D2

  • Medical oncology agent

    Ready
    Tool calls (3)
    • Read biopsy report D2
    • Read MRI pelvis D3
    • Read laboratory results D5
    • Candidate for neoadjuvant therapy. Sequencing is open for discussion.
    • MMR/MSI result and CEA (pending) could change the options discussed.

    D2D3D5

Board preparation

Prepared by AI agents for discussion. Not a recommendation.

Agreement

Locally advanced mid-rectal cancer on MRI. Colorectal surgery and medical oncology both raise neoadjuvant treatment before surgery.

Divergence

Colorectal surgery and medical oncology raise treatment sequencing differently. This is left open for the board to discuss.

Missing before decision

  • MMR / MSI status
  • CEA result
  • Updated staging CT
  • Completion colonoscopy
  • Family history

Human review

Decided by MDT coordinator

The board discusses the case and records its decision. The agents never propose one.

5 items to request before the meeting

Demo only. Nothing is saved.

Design intent

Design intent

  1. 01

    Case summary with sources

    A concise, source-referenced summary of history, imaging, pathology, and prior treatment.

  2. 02

    Perspectives by discipline

    Separate agents prepare the considerations radiology, pathology, surgery, and oncology would each raise.

  3. 03

    Agreement and divergence

    Points of consensus and disagreement are laid side by side, not averaged into one answer.

  4. 04

    Missing information

    Gaps such as a pending biomarker or an outdated scan are listed before the meeting, not discovered during it.

  5. 05

    Decision record

    The board’s decision and rationale are recorded by clinicians, and the preparation is kept for audit.

How it works

  1. 1

    Assemble

    Case documents are gathered and summarised with references.

  2. 2

    Prepare perspectives

    Each discipline’s considerations are drafted separately.

  3. 3

    Compare

    Agreement, divergence, and gaps are laid out for the coordinator.

  4. 4

    Decide

    The board discusses and records its decision. The system does not vote.

Built for

  • Tumour boards and other MDT meetings
  • MDT coordinators
  • Hospital quality and governance leads

Safeguards

  • Agent outputs are labelled as preparation material, never as recommendations of the board.
  • Each statement references the document it relies on.
  • Disagreement is preserved and shown, not resolved automatically.

Boundaries

  • This is a working title for a product in design. Capability and availability are to be confirmed.
  • It does not make treatment decisions, and its outputs are not clinical advice.

Related reading

  • Decision support3 min read

    AI for tumour board preparation

    Multidisciplinary team meetings change how patients are assessed and treated, and they take a lot of specialist time. Where multi-agent AI can help with preparation, and where it must stay out.

  • Decision support3 min read

    Showing uncertainty in clinical AI

    A prediction without a sense of how sure it is invites both over-trust and dismissal. Why calibration, case-level uncertainty, and the option to abstain belong in clinical AI.

  • Evaluation and safety3 min read

    Making clinician review work

    “A clinician reviews every output” is only a safeguard if the review is designed to work. What research on automation bias says about building it well.

Request a demo

Request a demo

Multidisciplinary Board Platform is already selected. Change it if you’d like to see something else.

A sentence about your team and workflow is plenty. Please don’t include patient information.

We use these details only to reply to your request.