Main site

Healthcare AI agent · Capacity & flow

Capacity & Patient-Flow AI Agent (Command Centre)

Hold a live picture of beds, pending discharges and incoming demand, propose placements against configured unit and isolation rules, surface each discharge's barrier, and prepare transfer decisions — every one stays with a person.

Typical delivery
4–6 weeks
Deployment
Your stack
Human confirmation
Advisory-only
After launch
Agent Care

What this agent does

Proposes the placement, not the release

In

  1. 01

    A bed request arrives from the ED, OR or an accepting unit, with the live census and ADT feed.

  2. 02

    Normalise unit, isolation flag, ratio configuration and stated arrival time across sources.

Reason

  1. 03

    Read predicted discharges, current occupancy and the configured unit and isolation rules.

  2. 04

    Apply the ratio limits and cohorting rules configured for each unit, never inferred.

  3. 05

    Score candidate beds against the request and surface the barrier behind each pending discharge.

Decide

  1. 06

    A transfer call comes in; flag what the picture shows and what it cannot confirm.

  2. 07

    Route every placement and every transfer read to the house supervisor or charge nurse.

Out

  1. 08

    Retain the census snapshot, the candidate beds, the barrier and the decision against the request.

  2. 09

    Execute write actions only inside the approval boundaries agreed during implementation.

Product statement

The agent proposes a placement or a transfer read; the house supervisor, charge nurse or physician decides, and a predicted date stays a planning signal, never a clinical conclusion.

Example workflow

One bed request, arrival to placement

  1. 1Bed request receivedADT feed, ED board, OR schedule or transfer-centre call
  2. 2Picture assembledCensus, pending discharges, isolation flags, ratio configuration and unit rules
  3. 3Placement proposedCandidate bed, discharge barrier, transfer read and confidence
  4. 4Controls appliedRatio checks, isolation checks, staleness checks and confidence threshold

No human action required

Stages 1 to 4 run unaided, and no bed is held at any of them — the agent is proposing, and the human lane opens at the confidence gate.

5DecisionBranches at the confidence threshold
High confidence

Goes to the house supervisor or charge nurse to confirm.

Low confidence

Adds a capacity-manager read first.

Human confirmation

The placement is held with its candidate beds, the barrier and the confidence.

Confirm · Reassign · Escalate

Confirmed — bed board updated

  1. 6ADT and bed board updatedOnly where write access and approval policy allow it
  2. 7Outcome evaluatedPlacement accuracy, reassignments, escalations and time-to-placement

Reassignments

Every reassignment made in confirmation is counted in the evaluation.

What should not run autonomously

Human approval stays in control

Outside the boundary — human approval required

8 items

  • Accepting or refusing a transfer request.
    • Deciding a patient is ready for discharge.
    • Overriding a configured nurse-to-patient ratio.
    • Determining medical necessity for admission.

  • Automation boundary

    Agent acts unaided

    • ✓Hold a live picture of beds, discharges and incoming demand.
    • ✓Propose placements against configured unit and isolation rules.
    • ✓Track the barrier behind each pending discharge.
    • ✓Prepare the picture a transfer decision needs, and hold.

    Any write happens inside the boundaries agreed at implementation, never ahead of a decision.

    • Predicting clinical deterioration or acuity.
    • Conducting the physician-to-physician transfer call.
    • Performing the discharge-planning evaluation.
  • Changes to unit, isolation or ratio configuration.

Example output

One bed request, annotated

Everything the agent proposes is attached to the request it was drawn from.

Placement output · single bed request

Illustrative example
Request
ED admit request
Proposed placement
Bed 4B-12, single room, contact-precaution cohort
Unit ratio
1:4
Candidate unit
4B – Med-surg
Confidence
88%
Isolation status
Contact precautions cleared

As received

Taken from the ADT feed and the unit's configuration — nothing on this side is inferred by the agent.

Evidence used

  • ADT bed-status feed
  • Unit isolation flag
  • Configured ratio limit

Why this candidate

The ratio and isolation rule are configured for the unit, not decided for this one request.

What the score decides

Below the configured threshold the placement picks up a capacity-manager read.

Action

ConfirmReassignEscalate

Value

Where AI adds value

The same four claims, placed at the point in the workflow where each one applies.

Where the value lands

Value 01 – 04

Every bed request

From the census and ADT feed

03Placement

Score against unit rules

Weigh the request against configured unit, isolation and ratio rules for each candidate bed.

01Approved path

Capacity is a legal statement

The rule that cleared a placement travels with it — the record is the deliverable, not just the bed.

02Human review

Point the supervisor at the tight cases

Ratio-tight units and low-confidence placements are marked, so the supervisor's read starts where the bed is hardest to place.

04

Build an evidence trail

The placement, the constraint behind it and the charge nurse who accepted it stay on the bed request.

Integrations

Typical integrations

Five system groups connect to the same agent. Which of them are in scope is decided in discovery.

  • ADT and bed systems

    • Epic ADT
    • Cerner Millennium
    • MEDITECH
    • HL7 ADT feeds

  • Bed-management platforms

    • TeleTracking
    • central logistics
    • Bed board and housekeeping

  • ED and OR scheduling

    • Epic ASAP
    • ORMIS
    • Perioperative and ED tracking boards

Agent

Capacity & patient-flow command centre

  • Reads the census
  • Proposes the placement
  • Holds for confirmation

  • Transfer centre and discharge

    • TeleTracking Transfer Center
    • Case-management systems

  • Observability & evaluation

    • OpenTelemetry
    • Langfuse
    • Supported monitoring/evaluation sources

Integration availability depends on the client's existing systems and API access.

Agent controls

Six layers between the model and the bed board

Each control wraps the one inside it. What the stack does not catch is named in the map below it.

L6 · Outermost — last line of defence

Inward → L1 · closest to the model

L6Rollback / safe modeFall back to census reporting only when evaluation or production signals degrade.Roll back
L5Version monitoringTrack model, prompt, unit-rule and ratio-configuration changes.Track
L4TraceabilityRecord the census read, the candidate beds, the constraint and the confirmation.Record
L3Charge-nurse approvalHold placements for the named charge nurse; it governs release, not whether the candidate bed is right.Gate
L2Policy guardrailsTest placements against configured ratio, isolation and unit rules; a failure returns the candidate.Restrict
L1Confidence thresholdsRoute low-confidence placements to a capacity-manager read before the charge nurse sees them.Require review
Model corePlacement proposed — candidate bed, discharge barrier, transfer read and confidence
L1 – L2
Test whether a placement may stand
L3
Puts the release in a charge nurse's hands
L4 – L5
Keep the placement and the constraint behind it
L6
Falls back to census reporting when signals degrade

How Nestack evaluates it

Evaluate the placement workflow — not only the final bed.

Coverage runs the whole depth of the workflow, and every layer is cut by slice.

Surface — the placement the unit sees

  1. E1Final-output evaluationDid the proposed bed match the request's isolation and ratio rules?
  2. E2Step-level evaluationDid the agent use the right census, unit rules and ratio configuration?
  3. E3Tool evaluationDid it read and write the correct bed, unit and ADT record?
  4. E4Confidence calibrationDo low-confidence placements actually draw more reassignments?
  5. E5Slice evaluationHow does performance change across specific unit types?
  6. E6Business outcomeHow many placements needed a reassignment or an escalation?

Floor — the outcome the hospital answers for

Failure modes

Where each failure originates in the agent

Seven failure modes, mapped to the stage each one starts at.

Agent lifecycle

Direction of processing →

  1. 01 · Retrieval

    1 mode

    • FW-03

      Stale bed-status feed

      Housekeeping or ADT status hasn't posted the bed's current state.

    Stage gathers

    Census, ADT feed, isolation flags and unit configuration

  2. 02 · Reasoning

    2 modes

    • FW-04

      Deterioration language creep

      A logistics signal is phrased as a claim about clinical risk or acuity.

    • FW-06

      Discharge date read as order

      A predicted date is treated as the clinical discharge decision, not a planning signal.

    Stage proposes

    Candidate bed, discharge barrier and confidence

  3. 03 · Tool / write

    2 modes

    • FW-02

      Placement written past a ratio limit

      A bed offer writes to a unit whose staffing ratio the placement would breach.

    • FW-05

      Isolation rule bypassed on write

      A bed assignment writes back before the unit's configured isolation rule clears it.

    Stage writes

    Only where write access and approval policy allow it

  4. 04 · Output

    1 mode

    • FW-01

      Constraint dropped on handoff

      The rule that shaped a placement doesn't travel with it to the receiving unit.

    Stage returns

    The placement the charge nurse confirms

  5. 05 · Change / Version

    1 mode

    • FW-07

      Silent threshold drift

      A configuration change loosens which units or ratios a placement will clear without review.

    Stage tracks

    Model, prompt, unit rules and ratio configuration

  • Sev-1 · writes outside the boundary
  • Sev-2 · a wrong placement reaches the unit
  • Sev-3 · feed degrades, routes to review
See the complete Healthcare failure-mode catalogue

Affected slices

A house-wide placement score hides the unit underneath it

A house-wide placement-rework rate can look acceptable while one unit or a single cohort absorbs most of the reassignments. Nestack reports the rate by slice, not only in total.

Slice performance — reported separately, not only in aggregate

Illustrative example
SliceFailure rateLiftLift vs. thresholdStatus
Isolation and cohorting requirements7.5%3.9×Review
Off-service and overflow placements5.4%2.8×Review
Predicted same-day discharges3.3%1.7×Watch
Routine med-surg placements1.9%0.8×Normal

Bar: placement-rework rate lift vs. the routine-med-surg baseline · scale 0–4.0× · tick marks the 2.0× review threshold

2 of 4 slices over threshold

Evidence-linked improvement

The loop closes on the bed, not the metric

A cycle closes when the bad placement is a case the next release has to survive. That suite is what the next bed assigned is measured against.

Improvement cycle · five stages

Switchback — the path turns at Improve and returns at Learn

  1. 01

    Detect

    Placement-rework rate rises in a unit slice.

  2. 02

    Diagnose

    The bed that showed as clean and was not is traced to the feed, the rule or the workflow that missed it.

  3. 03

    Improve

    The fix ships against a version, with the placements that exposed it attached.

  4. 04

    Verify

    Release is held until the affected placement cases pass again.

  5. 05

    Learn

    The case joins the permanent suite and the placement rules move with it.

  6. Learn → Detect

    The return edge. The next census runs against a suite one placement case longer.

Typical build scope

Twelve workstreams across six weeks

The build scope read against the delivery timeline. Week structure follows the six-week plan — discovery, sources, placement workflow, evaluation, integration, then production validation and handover.

Workstream

Week 1Week 2Week 3Week 4Week 5Week 6

  1. 01Capacity workflow and boundary definition and it is logged..Week 1
  2. 02ADT and bed-system source assessment.Week 1, Week 2
  3. 03Unit, isolation and ratio-rule mapping and rule mapping.Week 1, Week 2
  4. 04Census ingestion and normalisation.Week 2
  5. 05Placement retrieval and scoring logic.Week 2, Week 3
  6. 06Confidence scoring and escalation routing.Week 3
  7. 07Charge-nurse confirmation workflow.Week 3, Week 4
  8. 08ADT and transfer-centre integration.Week 5
  9. 09Placement regression cases.Week 4, Week 5
  10. 10Guardrails and escalation controls.Week 4
  11. 11Bed-trail instrumentation.Week 5, Week 6
  12. 12Deployment, documentation and Agent Care handover.Week 6

12 workstreams · 6 weeks · bar shows the weeks a workstream is active — several run in parallel

Final scope and sequence confirmed in discovery

Engagement tiers

What each tier includes

Rows are the capabilities named in each tier's scope. Higher tiers include everything below them.

Capability✓ in scope · — not at this tierPilotOne unit, one campusProductionProduction ADT integrationAdvancedMultiple campuses / systems
Introduced at Pilot
Placement recommendations✓✓✓
Human confirmation✓✓✓
Placement-accuracy baseline✓✓✓
Introduced at Production
Reporting by unit—✓✓
Confirmation workflow in your systems—✓✓
Approved bed write-back—✓✓
ADT-system integration—✓✓
Introduced at Advanced
Multi-campus placement rules——✓
Multi-stage confirmation chains——✓
High placement volume——✓
Multi-campus capacity controls——✓
Build priceFrom $5,000From $8,000Custom quote
Final build price
Confirmed after discovery based on integrations, workflow complexity, transaction volume, approval controls and deployment requirements.
Separate from build
Build pricing is separate from recurring Agent Care, which covers managed monitoring, evaluations, incidents and verified improvements after launch.

What we need from you

What you bring, and what we build with it

Each input maps to a piece of build scope and a week in the delivery timeline.

  1. 01Your live census, ADT feed and bed-board layout

    Census ingestion and bed-board mapping

    Week 1

  2. 02Representative recent placements and transfers

    Placement baseline and constraint extraction

    Week 2

  3. 03Your unit, isolation and ratio rules

    Unit, isolation and ratio-rule mapping

    Week 1

  4. 04Access to relevant APIs, feeds or exports

    ADT and transfer-centre assessment, then integration setup

    Week 2

  5. 05Placements you would not want made

    Transfer cases and failure-mode testing

    Week 4

  6. 06What no placement may decide alone

    Confidence scoring, escalation routing, guardrails and approval controls

    Week 3

  7. 07Named house supervisors and charge nurses to confirm

    Charge-nurse confirmation workflow, then pilot and production validation

    Weeks 5–6

Nothing else is required

Deployment, documentation and Agent Care handover are ours.

Delivery timeline

Four phases across six weeks

Each band covers the weeks the work really takes, which is why week 5 carries evaluation and pilot together.

Phase

W1

W2

W3

W4

W5

W6

  1. DiscoveryW1
  2. BuildW2 – W3
  3. EvaluateW4 – W5
  4. Pilot & LaunchW5 – W6

Week focus

W1Capacity workflow discovery, unit-rule mapping and boundary

W2ADT integration and the placement baseline

W3Placement logic, confidence scoring and escalation controls

W4Evaluation suite, guardrails and failure-mode testing

W5Transfer-centre integration, pilot placements and corrections

W6One census cycle run under the house supervisor, then Agent Care handover

Reading the band
A bar covers the weeks its work is named in, and nothing else. The week 5 overlap is real, not padding.
At the end of W6
Validation closes on live placements, and Agent Care picks up monitoring.
Duration
Six-week plan shown · typical delivery 4–6 weeks depending on scope confirmed in discovery.

Next step · Healthcare AI agent

Build a capacity and patient-flow agent around your bed board.

Show us your bed board, your unit and ratio rules, and who confirms a placement. If your census spans multiple units or a transfer centre, we'll map the workflow around it and name what stays with a person.

Typical delivery
4–6 weeks
Deployment
Your ADT and bed systems
Human confirmation
Advisory-only
After launch
Agent Care available

Nestack Agents · Capacity & patient-flow

AGT-HC-12 · Agent Care available after launch

Build a capacity and patient-flow agent

Four questions so the call starts from your actual setup — then pick a time that suits you.

About you
Your census

No obligation. We’ll map the workflow, name the automation boundary and tell you if an agent is the wrong answer.