V•VITALHEALTHCARE AI PLAYBOOK
Back to overview

Clinical documentation / PLAY 01

Ambient clinical notes

Work through the evidence. Make a decision you can explain.

YOUR ASSIGNMENT / Routine case

The note before the signature

A clinician has finished a fictional visit. The source includes symptoms, explicit negative history, an unconfirmed assessment, and a follow-up plan. Your job is to preserve what was said without turning uncertainty into a diagnosis.

THE DECISION TO MAKE

Is the draft faithful enough to send to the clinician for review?

What you will produce

A source-linked note and a record of what needs verification.

What a careful reviewer checks

Check negative statements, uncertainty, and the clinician’s plan against the encounter facts.

Acceptance criteria

  • Every included statement has a source.
  • Negative history is preserved.
  • The clinician remains responsible for signing.

The example data

6 source records
Inspect all 6 source records
Next: calculate the output and trace it to the source.
Synthetic case study · Fixed rules, no live clinical AI · Reports stay in this page session.
VITAL Healthcare AI, with intent.Educational planning · No patient data entry
VITAL / HEALTHCARE AI

Decision brief · AMBIENT ·

Ambient clinical notes

Turn the consultation into a draft. Keep the clinician in control.

1. Pilot proposition

Accountable sponsor: CMIO + clinical service lead

One specialty, a defined clinician cohort, and a baseline period. Compare documentation time and note quality before expanding.

Primary measure: Net documentation minutes per encounter

2. Controls and decision evidence

Human accountability: A clinician verifies the draft against the encounter before signing. No autonomous chart filing.

Data: Approved encounter audio, clinician edits, consent or notice workflow under local policy, and EHR audit events.

Validation: Blinded note-quality review; omissions, unsupported facts, negation errors, subgroup performance, and clinician correction time.

Stop rule: Pause if a material fabricated fact reaches a signed note, consent controls fail, or patient identity is mismatched.

3. Illustrative value case (base)

8,000 eligible tasks/month; 70% adoption; 6 minutes gross release; 2 minutes review; $85/hour; 35% financial realization. Recurring cost $8,000/month; setup $25,000.

Monthly net capacity: 373 hours. Year-one net: $12,280. Break-even: 8.0 months.

Net capacity = tasks × adoption × (gross minutes − review minutes) ÷ 60. Year-one net = 12 × (capacity × hourly value × applied realization − monthly cost) − setup. Applied realization is 100% for negative net capacity so added work is fully costed. Constant run rate; no ramp, discounting, taxes, or clinical-outcome valuation. Assumptions are not observed outcomes.

4. Evidence gate record (0/12 self-attested)

[OPEN] Define a bounded intended use

Clinical or operational sponsor: One-page charter: users, population, exclusions, current workflow, and non-AI alternative.

[OPEN] Establish the baseline

Quality + finance: Measure the primary outcome, quality balancing measures, denominator, and staff time including review.

[OPEN] Clear the data and privacy route

Privacy + security: Approved data-flow map, access design, retention terms, and applicable agreements before patient data is used.

[OPEN] Set approval and stop criteria

Safety lead + governance committee: Document intended-use review, accountable sign-off, incident escalation, and a tested fallback.

[OPEN] Validate on representative local cases

Independent clinical or quality reviewers: Test report covering failures, subgroups, uncertainty, missing data, and a comparison with the baseline.

[OPEN] Rehearse a failure and rollback

Service owner + IT operations: Recorded downtime drill, incident ownership, access controls, and a demonstrated manual workflow.

[OPEN] Prepare people and workflow

Change lead + frontline champions: Training, role-based instructions, review capacity, patient communication, and feedback routes.

[OPEN] Approve a controlled pilot

Accountable governance committee: Signed limited-pilot decision with scope, conditions, staffing, monitoring, and regulatory review where applicable.

[OPEN] Measure benefit and unintended effects

Quality + finance: Observed results with uncertainty, adoption, rework, subgroup outcomes, safety events, and full costs.

[OPEN] Make the scale, revise, or stop decision

Executive sponsor + clinical safety: Document the decision against pre-agreed criteria. A calendar deadline is not a reason to pass a gate.

[OPEN] Operationalize monitoring and change control

Named service owner: Thresholds, review cadence, vendor-change notices, revalidation triggers, and an incident runbook.

[OPEN] Verify value after deployment

Finance + benefit owner: Separate released time, realized financial value, and clinical outcomes; revisit the case using actual observations.

5. Decision and sources

Decision: ____________________ Approver: ____________________ Review date: ___________

No deployment is authorized by this brief. Confirm intended use, local requirements, independent validation, and governance approval.

HHS · OCR: HIPAA & cloud computing
https://www.hhs.gov/hipaa/for-professionals/special-topics/health-information-technology/cloud-computing/index.html

FDA: Clinical Decision Support Software
https://www.fda.gov/regulatory-information/search-fda-guidance-documents/clinical-decision-support-software

ONC: HTI-1 final rule
https://healthit.gov/regulations/hti-rules/hti-1-final-rule/

NIST: AI Risk Management Framework
https://airc.nist.gov/airmf-resources/airmf/5-sec-core/

Primary sources reviewed 18 September 2026. US-oriented educational planning, not clinical advice or compliance certification.