Prompts • Skills • Connectors • Agents • Health System workflows • Tools
AI Playbook
for Hospital & Health System Administrators
For finance, revenue cycle, operations, supply chain, and administrative leaders at hospitals, health systems, and community health centers. A practical guide to budgeting and cost reporting, revenue cycle and denials, supply chain, staffing and labor costs, grant and program reporting, and board reporting.
Who this playbook is for
This playbook is for the finance, revenue cycle, operations, and administrative leaders who run hospitals and health systems, not for clinical staff. It covers the business side: budgets, cost reports, revenue cycle, supply chain, labor costs, grants, and board reporting. Clinical work such as documentation, prior authorization clinical review, and patient flow is covered in the Healthcare playbook.
- CFOs, controllers, and finance teams
- Revenue cycle, denials, and managed care leaders
- Supply chain and purchasing leaders
- Workforce, HR, and labor finance teams
- Grants, compliance, and administrative leaders at hospitals and community health centers
- Physicians, nurses, and clinical documentation (see the Healthcare playbook)
- Clinical decision support or patient care
- Patients seeking medical or billing advice
Start With Your Role
Choose the role closest to your day-to-day work. Start with one useful prompt, make it repeatable, then add the information it needs.
- Start with: monthly variance narrative or finance committee report
- Save as a skill: variance method and report format
- Connect: aggregate financial exports
- First agent: monthly variance narrative drafter
- Measure: reporting days, edits
- Start with: denial root-cause analysis or KPI narrative
- Save as a skill: denial mapping and KPI definitions
- Connect: aggregate denial and KPI reports; PHI workflows only under BAA
- First agent: denial triage & deadline watch
- Measure: denial rate, overturns, AR days
- Start with: cost report checklist or cost per case
- Save as a skill: cost report readiness format
- Connect: workpapers, cost model outputs
- First agent: cost report readiness tracker
- Measure: on-time filing, adjustments
- Start with: supply spend analysis
- Save as a skill: spend and contract rules
- Connect: PO and contract data
- First agent: off-contract spend monitor
- Measure: contract compliance, savings
- Start with: contract labor reduction plan or productivity report
- Save as a skill: productivity targets and report format
- Connect: timekeeping exports
- First agent: contract labor & overtime watch
- Measure: contract labor cost, overtime
- Start with: grant opportunity summary or UDS data check
- Save as a skill: report and budget formats
- Connect: grants tracker, aggregate program data
- First agent: grant deadline & burn rate monitor
- Measure: on-time reports, funding secured
100 AI Prompts for Hospital & Health System Administrators
Ready-to-use prompts for budgeting, cost reporting, revenue cycle, denials and underpayments, supply chain, staffing and labor costs, grants and program reporting, board reporting, compliance, and operations. Copy one, add your organization's context, and review the result.
Default to aggregate and de-identified data, use PHI only in tools approved under a BAA, and review the result before it reaches payers, regulators, or the board.
Prompts for operating budgets, volume forecasts, variance narratives, and long-range plans.
Separating volume from rate and efficiency makes variances clear to leaders and the board.
Risks: Explanations need confirmation. Control: Department leaders and the controller confirm.
A structured framework focuses budget meetings on decisions.
Risks: Assumptions drive results. Control: CFO and finance committee approve.
Volume drives revenue, staffing, and supply budgets.
Risks: Market changes are hard to predict. Control: Service line leaders review.
Rolling forecasts keep decisions current.
Risks: Trends can reverse. Control: Review monthly.
Flexing shows real efficiency, not just volume effects.
Risks: Fixed/variable splits are estimates. Control: Agree on splits with leaders.
Speeds budget review and keeps conversations specific.
Risks: Departments have context. Control: Budget office meets with leaders.
Payer mix shifts can move margins more than volume.
Risks: Contract terms vary. Control: Managed care team confirms rates.
Long-range plans reveal structural issues early.
Risks: Projections are uncertain. Control: Stress-test with leadership.
Shows where margin comes from.
Risks: Service decisions involve mission and community need. Control: Leadership and board decide.
Clear kickoffs produce better budget submissions.
Risks: Guidelines must be approved first. Control: CFO approves.
12 Claude-Ready Health System Business Skills
These are downloadable Claude Skill packages, not just longer prompts. Each includes a SKILL.md your team can tailor with your policies and data rules. A copy-and-paste version works in ChatGPT and other AI tools.
Connect Your Work
AI can only help with the information it can see. On the business side, start with de-identified and aggregate data, and bring AI near PHI only under a BAA.
Easy start
Start with a limited, low-risk connection your team can test quickly.
- Access: scheduled exports of financial statements, volumes, KPIs, spend, and labor hours with no PHI, into an approved shared location
- Useful for: variance narratives, KPI summaries, spend analysis, and board packets
- Setup & limit: confirm reports contain no PHI; name an owner for each export
- Access: Outlook, SharePoint, Excel, and Teams connections, or Microsoft Copilot in your tenant
- Useful for: board materials, budgets, policies, and contracts
- Setup & limit: IT controls access; keep PHI out unless the tenant is approved for it
- Access: Gmail, Drive, Sheets, and Calendar connections in Claude and ChatGPT
- Useful for: documents and reporting for organizations using Google
- Setup & limit: follow data classification rules
- Access: connectors to the CMS Coverage Database (NCDs and LCDs), ICD-10 code lookup, and the NPI Registry, announced January 2026
- Useful for: coverage policy lookups for denial prevention, code reference, and provider data checks
- Setup & limit: public reference data, not your patients' records; confirm availability on your plan
Needs an administrator
Useful, but someone needs to set access and permissions first.
- Access: Claude for Enterprise is offered as HIPAA-ready for healthcare organizations; other vendors offer similar enterprise programs
- Useful for: approved workflows that may involve PHI, such as appeal preparation
- Setup & limit: requires an enterprise agreement, a signed BAA, and privacy officer approval before any PHI is used
- Access: Epic's revenue cycle AI (e.g., Penny for denials and prior authorizations) and Agent Factory work inside Epic; other vendors offer similar features
- Useful for: denials and revenue cycle work inside the system of record
- Setup & limit: governed by your EHR agreements; review with your AI governance committee
- Access: Workday, Oracle, Strata, and Axiom offer their own AI and reporting; NetSuite has an AI Connector Service for smaller organizations
- Useful for: finance, supply chain, and cost questions
- Setup & limit: read-only reporting roles; postings stay with finance
Integration project
Plan the use case, source data, permissions, and owner before connecting.
- Access: no official direct Claude or ChatGPT connector to Epic or Oracle Health records is confirmed here for business use
- Useful for: governed, de-identified data from your analytics warehouse
- Setup & limit: an IT, privacy, and governance project; minimum necessary data and logging required
- Access: contract modeling systems and payer portals
- Useful for: expected reimbursement models and remittance data for underpayment analysis
- Setup & limit: claims data may contain PHI; use approved tools under a BAA
- Access: GHX and GPO data portals
- Useful for: contract pricing and spend data
- Setup & limit: follow GPO data terms
Good to know
Limits and honest gaps.
- Access: patient records, claims with patient detail, and clinical documentation
- Useful for: nothing in general AI tools; only BAA-covered, approved systems
- Setup & limit: business analysis defaults to aggregate and de-identified data
- Access: documentation, prior authorization clinical review, and patient flow are covered in the Healthcare playbook
- Useful for: not covered here; this playbook is for business and administrative teams
- Setup & limit: keep business and clinical AI governance consistent
Tools
Choose the AI tools and health system software that fit your teams. The goal is to make your existing systems easier to use, not replace everything at once.
AI Assistants
4EHR & Revenue Cycle Platforms
4Finance & Decision Support
5Supply Chain
5Grants & Community Health
4Board & Compliance
412 AI Agents for Hospital & Health System Administrators
An agent is an assistant for one recurring job. It gathers the right information, follows your process, and prepares the work for someone to review and decide.
A good first agent handles one task your department already does, reads aggregate data or approved sources, never submits anything, and hands its work to a named person, with approval from your AI governance committee.
# Monthly Variance Narrative Drafter ## Job to be done Cut the time to produce monthly financial narratives. ## When it runs Monthly after close. ## Approved context Approved financials, volumes, budget, prior year. ## Operating loop Compute volume, rate, and efficiency variances; flex budgets; draft narrative and questions. ## What it delivers A draft narrative. ## Human owner Controller and CFO. ## What it must never do Never distribute reports or change budgets. ## How to measure it Prep time, edits, on-time reporting. ## Operating rules 1. Use only approved sources and follow the linked health system skill. 2. Cite report, claim, or record IDs and separate facts from assumptions. 3. Route compliance questions, PHI exposure risks, and material variances to the human owner immediately. 4. Keep an audit record of trigger, inputs, output, reviewer, decision, and override. 5. Stop and ask for human review when regulation, contract terms, or approval authority are unclear. 6. Never process PHI outside tools approved under a BAA by your privacy officer.
# Cost Report Readiness Tracker ## Job to be done Keep cost report inputs on schedule and defensible. ## When it runs Monthly; weekly in the 90 days before filing. ## Approved context Checklist, statistics, bad debt listings. ## Operating loop Track input status; flag statistic swings and documentation gaps. ## What it delivers A readiness report. ## Human owner Reimbursement manager. ## What it must never do Never prepare or file cost reports. ## How to measure it On-time filing, audit adjustments. ## Operating rules 1. Use only approved sources and follow the linked health system skill. 2. Cite report, claim, or record IDs and separate facts from assumptions. 3. Route compliance questions, PHI exposure risks, and material variances to the human owner immediately. 4. Keep an audit record of trigger, inputs, output, reviewer, decision, and override. 5. Stop and ask for human review when regulation, contract terms, or approval authority are unclear. 6. Never process PHI outside tools approved under a BAA by your privacy officer.
# DNFB & KPI Watch ## Job to be done Keep cash flowing by spotting revenue cycle bottlenecks. ## When it runs Daily for DNFB; weekly for KPIs. ## Approved context DNFB by hold reason, revenue cycle KPIs. ## Operating loop Summarize DNFB by reason and age; flag KPI drift; assign owners. ## What it delivers A revenue cycle status report. ## Human owner Revenue cycle director. ## What it must never do Never touch accounts, codes, or documentation. ## How to measure it DNFB dollars, days in AR, clean claim rate. ## Operating rules 1. Use only approved sources and follow the linked health system skill. 2. Cite report, claim, or record IDs and separate facts from assumptions. 3. Route compliance questions, PHI exposure risks, and material variances to the human owner immediately. 4. Keep an audit record of trigger, inputs, output, reviewer, decision, and override. 5. Stop and ask for human review when regulation, contract terms, or approval authority are unclear. 6. Never process PHI outside tools approved under a BAA by your privacy officer.
# Denial Triage & Deadline Watch ## Job to be done Work the highest-value denials before deadlines. ## When it runs Daily as remittances post. ## Approved context Denials with codes, appeal deadlines, outcome history. ## Operating loop Categorize new denials; prioritize by expected recovery and deadline; organize evidence lists. ## What it delivers A prioritized denial queue. ## Human owner Denials manager. ## What it must never do Never submit appeals or write clinical arguments. ## How to measure it Overturn rate, dollars recovered, missed deadlines. ## Operating rules 1. Use only approved sources and follow the linked health system skill. 2. Cite report, claim, or record IDs and separate facts from assumptions. 3. Route compliance questions, PHI exposure risks, and material variances to the human owner immediately. 4. Keep an audit record of trigger, inputs, output, reviewer, decision, and override. 5. Stop and ask for human review when regulation, contract terms, or approval authority are unclear. 6. Never process PHI outside tools approved under a BAA by your privacy officer.
# Underpayment Detector ## Job to be done Catch payer underpayments systematically. ## When it runs Weekly. ## Approved context Remittances, contract models. ## Operating loop Compare payment to expected; flag variances and patterns; summarize by payer. ## What it delivers An underpayment report. ## Human owner Managed care analyst. ## What it must never do Never contact payers. ## How to measure it Underpayments recovered, variance trends. ## Operating rules 1. Use only approved sources and follow the linked health system skill. 2. Cite report, claim, or record IDs and separate facts from assumptions. 3. Route compliance questions, PHI exposure risks, and material variances to the human owner immediately. 4. Keep an audit record of trigger, inputs, output, reviewer, decision, and override. 5. Stop and ask for human review when regulation, contract terms, or approval authority are unclear. 6. Never process PHI outside tools approved under a BAA by your privacy officer.
# Off-Contract Spend Monitor ## Job to be done Keep purchasing on contract and on price. ## When it runs Weekly. ## Approved context PO and invoice data, contract prices. ## Operating loop Flag off-contract and price-variance purchases; summarize by department and category. ## What it delivers A spend exception report. ## Human owner Supply chain director. ## What it must never do Never place or cancel orders. ## How to measure it Contract compliance %, price variance dollars. ## Operating rules 1. Use only approved sources and follow the linked health system skill. 2. Cite report, claim, or record IDs and separate facts from assumptions. 3. Route compliance questions, PHI exposure risks, and material variances to the human owner immediately. 4. Keep an audit record of trigger, inputs, output, reviewer, decision, and override. 5. Stop and ask for human review when regulation, contract terms, or approval authority are unclear. 6. Never process PHI outside tools approved under a BAA by your privacy officer.
# Backorder & Expiry Watch ## Job to be done Warn early about supply shortages and expiring stock. ## When it runs Daily. ## Approved context Backorder reports, days on hand, expiry dates, approved substitutes list. ## Operating loop Rank shortage risk; list approved substitute status; flag expiring items. ## What it delivers A supply risk report. ## Human owner Materials manager. ## What it must never do Never approve substitutions or change pars. ## How to measure it Stockouts, expired waste. ## Operating rules 1. Use only approved sources and follow the linked health system skill. 2. Cite report, claim, or record IDs and separate facts from assumptions. 3. Route compliance questions, PHI exposure risks, and material variances to the human owner immediately. 4. Keep an audit record of trigger, inputs, output, reviewer, decision, and override. 5. Stop and ask for human review when regulation, contract terms, or approval authority are unclear. 6. Never process PHI outside tools approved under a BAA by your privacy officer.
# Contract Labor & Overtime Watch ## Job to be done Spot labor cost drift every pay period. ## When it runs Each pay period; mid-period projection. ## Approved context Hours, contract labor, overtime, productivity targets. ## Operating loop Report productivity vs. target; project overtime; quantify contract labor premium. ## What it delivers A labor cost report. ## Human owner COO, CNO, and finance. ## What it must never do Never change schedules or staffing levels. ## How to measure it Contract labor cost, overtime %, productivity. ## Operating rules 1. Use only approved sources and follow the linked health system skill. 2. Cite report, claim, or record IDs and separate facts from assumptions. 3. Route compliance questions, PHI exposure risks, and material variances to the human owner immediately. 4. Keep an audit record of trigger, inputs, output, reviewer, decision, and override. 5. Stop and ask for human review when regulation, contract terms, or approval authority are unclear. 6. Never process PHI outside tools approved under a BAA by your privacy officer.
# Grant Deadline & Burn Rate Monitor ## Job to be done Never miss a grant report and keep spending on plan. ## When it runs Weekly; 45 days before report deadlines. ## Approved context Award tracker, budgets, expenditures. ## Operating loop Track deadlines; compute burn vs. time; draft report outlines with [NEEDED] flags. ## What it delivers A grants status report. ## Human owner Grants manager. ## What it must never do Never submit reports or draw funds. ## How to measure it On-time reports, unspent balances. ## Operating rules 1. Use only approved sources and follow the linked health system skill. 2. Cite report, claim, or record IDs and separate facts from assumptions. 3. Route compliance questions, PHI exposure risks, and material variances to the human owner immediately. 4. Keep an audit record of trigger, inputs, output, reviewer, decision, and override. 5. Stop and ask for human review when regulation, contract terms, or approval authority are unclear. 6. Never process PHI outside tools approved under a BAA by your privacy officer.
# Board Packet Assembler ## Job to be done Deliver complete board and committee packets on time. ## When it runs Two weeks before each meeting. ## Approved context Financials, dashboards, committee reports, prior minutes. ## Operating loop Collect inputs; draft reports and agenda; list missing items. ## What it delivers A packet draft. ## Human owner CFO and board secretary. ## What it must never do Never distribute materials or include executive session content. ## How to measure it On-time packets, prep hours. ## Operating rules 1. Use only approved sources and follow the linked health system skill. 2. Cite report, claim, or record IDs and separate facts from assumptions. 3. Route compliance questions, PHI exposure risks, and material variances to the human owner immediately. 4. Keep an audit record of trigger, inputs, output, reviewer, decision, and override. 5. Stop and ask for human review when regulation, contract terms, or approval authority are unclear. 6. Never process PHI outside tools approved under a BAA by your privacy officer.
# Compliance Calendar & BAA Monitor ## Job to be done Keep filings on time and vendors covered before PHI access. ## When it runs Weekly; when a new vendor or AI feature is requested. ## Approved context Compliance calendar, vendor inventory, BAA status. ## Operating loop Flag upcoming deadlines and vendors with PHI access lacking BAAs or with new AI features. ## What it delivers A compliance status report. ## Human owner Compliance and privacy officers. ## What it must never do Never sign agreements or file anything. ## How to measure it Missed deadlines, vendors without BAAs. ## Operating rules 1. Use only approved sources and follow the linked health system skill. 2. Cite report, claim, or record IDs and separate facts from assumptions. 3. Route compliance questions, PHI exposure risks, and material variances to the human owner immediately. 4. Keep an audit record of trigger, inputs, output, reviewer, decision, and override. 5. Stop and ask for human review when regulation, contract terms, or approval authority are unclear. 6. Never process PHI outside tools approved under a BAA by your privacy officer.
# AI Use Case Intake Reviewer ## Job to be done Give the AI governance committee consistent, scored requests. ## When it runs When a new AI use case is submitted. ## Approved context Use case requests, criteria, data classifications, vendor info. ## Operating loop Score requests; list data and oversight questions; check BAA status. ## What it delivers A scored intake packet. ## Human owner AI governance committee. ## What it must never do Never approve use cases or enable tools. ## How to measure it Time to decision, approved pilots, incidents. ## Operating rules 1. Use only approved sources and follow the linked health system skill. 2. Cite report, claim, or record IDs and separate facts from assumptions. 3. Route compliance questions, PHI exposure risks, and material variances to the human owner immediately. 4. Keep an audit record of trigger, inputs, output, reviewer, decision, and override. 5. Stop and ask for human review when regulation, contract terms, or approval authority are unclear. 6. Never process PHI outside tools approved under a BAA by your privacy officer.
AI for Budgeting & Cost Reporting
Use AI to draft variance narratives, build forecasts, and organize cost report inputs. Finance and reimbursement leaders own the numbers and filings.
- Prompts: monthly variance narrative, budget framework, cost report checklist
- Skills: operating budget & variance analysis; cost reporting & cost accounting
- Connect: ERP and decision support exports (Workday, Oracle, Strata, Axiom), cost report workpapers
- Agent: monthly variance narrative drafter; cost report readiness tracker
- What AI does: Volume, rate, and efficiency variances
- Helps: Faster, clearer reporting
- Control: Leaders confirm causes
- What AI does: Rolling and service line forecasts
- Helps: Current decisions
- Control: Review monthly
- What AI does: Department efficiency vs. volume
- Helps: Fair accountability
- Control: Agree splits
- What AI does: Inputs, statistics, and documentation
- Helps: Defensible filings
- Control: Specialists certify
- What AI does: DRG cost drivers
- Helps: Target opportunities
- Control: Clinical leaders interpret
- What AI does: Margins, liquidity, and covenants
- Helps: See gaps early
- Control: Stress-test
Finance Implementation Checklist
ControlsBefore You Start
- Use aggregate financial data
- Standardize adjusted volume metrics
- Collect cost report workpapers
- Baseline: reporting days
- Confirm tool approval
After You Start
- Reconcile every figure
- Track prep time
- Review forecasts monthly
- Close cost report gaps early
- Brief auditors on AI use
- Filings: Cost reports and tax filings are prepared and certified by people.
- Accuracy: Figures tie to the general ledger.
- Service decisions: Profitability data informs; leadership and boards decide with mission in mind.
- Data: Financial analysis uses aggregate data, not PHI.
- Approvals: Budgets are leadership and board decisions.
AI for Revenue Cycle & Denials
Use AI to analyze denials, organize appeals, detect underpayments, and track KPIs. Coders, physician advisors, and revenue cycle staff decide and submit.
- Prompts: denial root-cause analysis, appeal organizer, underpayment detection
- Skills: revenue cycle performance; denials & appeals preparation; payer contract & underpayment review
- Connect: revenue cycle reports from your EHR (Epic's Penny and other built-in AI work inside Epic), contract management system, Claude for Healthcare connectors (CMS Coverage Database, ICD-10, NPI)
- Agent: DNFB & KPI watch; denial triage & deadline watch; underpayment detector
- What AI does: Preventable denials by owner
- Helps: Fix front-end causes
- Control: Consistent mapping
- What AI does: Evidence and deadlines
- Helps: More overturns
- Control: Physician advisors write clinical content
- What AI does: Payments vs. contract models
- Helps: Recover lost revenue
- Control: Managed care verifies
- What AI does: Days in AR, DNFB, clean claims
- Helps: Focused teams
- Control: Standard definitions
- What AI does: NCD/LCD requirements
- Helps: Fewer medical necessity denials
- Control: Compliance interprets
- What AI does: Plain-language estimates and statements
- Helps: Better collections
- Control: Compliance reviews
Revenue Cycle Control Checklist
ControlsBefore You Start
- Confirm BAA coverage for any tool touching claims
- Map denial codes to categories
- Load contract models
- Baseline: denial rate, AR days
- Name owners by category
After You Start
- Review queues daily
- Track overturns and recoveries
- Report root causes monthly
- Update payer rules
- Audit a sample of outputs
- PHI: Only in tools approved under a BAA; default to claim IDs and aggregates.
- Coding: Certified coders decide codes; AI never assigns codes for billing.
- Medical necessity: Physician advisors and clinicians decide clinical arguments.
- Patients: Follow financial assistance (501(r) for nonprofits), EMTALA, and collection rules.
- Submissions: Staff submit claims and appeals.
AI for Supply Chain & Purchasing
Use AI to find spend savings, track contracts, and prepare value analysis. Committees and clinical leaders make product decisions.
- Prompts: supply spend analysis, contract renewal pipeline, value analysis summary
- Skills: supply chain spend & value analysis
- Connect: ERP and materials management exports, GPO contract data (Vizient, Premier, HealthTrust), GHX
- Agent: off-contract spend monitor; backorder & expiry watch
- What AI does: Off-contract and price variance
- Helps: Quick savings
- Control: Value analysis for clinical items
- What AI does: Contract pipeline and notice dates
- Helps: No surprise renewals
- Control: Contracting reviews
- What AI does: Consistent request summaries
- Helps: Faster committee decisions
- Control: Committee decides
- What AI does: Cost variation by procedure
- Helps: Standardization discussions
- Control: Physician leaders lead
- What AI does: Shortage risk and substitutes
- Helps: Prevent disruptions
- Control: Clinical leaders approve substitutes
- What AI does: Par and expiry review
- Helps: Less waste
- Control: Managers confirm
Supply Chain Checklist
WorkflowBefore You Start
- Clean item master
- Load contract prices
- Define value analysis process
- Baseline: contract compliance, price variance
- Name category owners
After You Start
- Review exceptions weekly
- Track savings realized
- Monitor backorders daily
- Report to leadership monthly
- Verify vendor bank changes independently
- Clinical products: Value analysis committees and clinical leaders decide.
- Substitutions: Clinical leaders approve during shortages.
- Fraud: Verify vendor bank changes by phone with known contacts.
- Contracts: Follow GPO and vendor terms.
- Orders: AI never places or cancels orders.
AI for Staffing & Labor Costs
Use AI to report productivity, reduce contract labor, and plan the workforce. Clinical leaders set staffing levels.
- Prompts: labor productivity report, contract labor reduction plan, overtime analysis
- Skills: workforce & labor cost analysis
- Connect: timekeeping and HR exports (UKG, Workday, Symplr, ShiftWise), productivity system
- Agent: contract labor & overtime watch
- What AI does: Hours per unit of service vs. target
- Helps: Staff to demand
- Control: Targets reflect acuity
- What AI does: Premium cost and reduction plan
- Helps: Biggest labor savings
- Control: Safe staffing first
- What AI does: Causes and fatigue risk
- Helps: Lower cost and burnout
- Control: HR confirms rules
- What AI does: Hard-to-fill roles and bottlenecks
- Helps: Faster hiring
- Control: Recruiters update
- What AI does: Patterns by role and manager
- Helps: Retention
- Control: Stay interviews
- What AI does: New service staffing timelines
- Helps: Launch on time
- Control: Clinical models
Workforce Implementation Checklist
ControlsBefore You Start
- Agree productivity targets with leaders
- Export hours and contract labor
- Define overtime causes
- Baseline: contract labor %, overtime
- Keep personnel data in HR systems
After You Start
- Report each pay period
- Track contract labor reduction
- Review turnover quarterly
- Validate targets annually
- Share with nursing leadership
- Safe staffing: Clinical leaders set staffing; AI never recommends cuts below their levels or requirements.
- Personnel data: Stays in HR systems; use IDs.
- Labor law: Overtime, breaks, and union agreements apply.
- Fairness: No decisions based on protected characteristics.
- Decisions: Leaders and HR decide.
AI for Grants & Program Reporting
Use AI to summarize opportunities, draft budgets and reports, and pre-check UDS and grant data. Grants and program leaders submit.
- Prompts: grant opportunity summary, budget justification, UDS data check
- Skills: grants & program reporting
- Connect: grants tracker, finance system, aggregate program data
- Agent: grant deadline & burn rate monitor
- What AI does: One-page fit and requirements
- Helps: Faster go/no-go
- Control: Read full announcement
- What AI does: Clear, allowable budgets
- Helps: Fewer questions
- Control: Grants finance verifies
- What AI does: Answers to funder questions
- Helps: On-time reporting
- Control: Verify numbers
- What AI does: Cross-table consistency (health centers)
- Helps: Cleaner submissions
- Control: Current UDS manual
- What AI does: Spending vs. time
- Helps: Avoid overruns
- Control: Program leads confirm
- What AI does: Organize required processes
- Helps: Compliance readiness
- Control: Compliance and counsel
Grants Implementation Checklist
WorkflowBefore You Start
- Inventory awards and deadlines
- Collect funder rules
- Use aggregate program data
- Baseline: late reports
- Name grant owners
After You Start
- Track deadlines weekly
- Review burn rates monthly
- Verify every reported number
- Close out on time
- Report to leadership quarterly
- Accuracy: Never invent outcomes or statistics.
- Federal awards: Follow Uniform Guidance (2 CFR 200) and award terms.
- Data: Use aggregate data; PHI stays in approved systems.
- Health centers: Follow HRSA program requirements.
- Submissions: Grants managers submit.
AI for Board & Executive Reporting
Use AI to assemble packets, narrate dashboards, and draft minutes. Executives and boards govern.
- Prompts: finance committee report, executive dashboard narrative, board agenda
- Skills: board & executive reporting; administrative compliance management
- Connect: board portal, financial reports, dashboards
- Agent: board packet assembler; compliance calendar & BAA monitor
- What AI does: Metrics, risks, and decisions
- Helps: Better governance
- Control: Controller verifies
- What AI does: Balanced executive view
- Helps: See risks early
- Control: Quality leaders verify quality data
- What AI does: Timed agendas and draft minutes
- Helps: Efficient meetings
- Control: Secretary verifies
- What AI does: Compliance and cushion
- Helps: Early warnings
- Control: Treasury and counsel
- What AI does: Plain-language briefs
- Helps: Informed trustees
- Control: Verify current policy
- What AI does: Calendar and committee summaries
- Helps: Oversight
- Control: Compliance officer edits
Board Reporting Checklist
WorkflowBefore You Start
- Standardize dashboard metrics
- Set packet deadlines
- Define executive session handling
- Baseline: prep hours
- Confirm board portal access rules
After You Start
- Verify every figure
- Keep confidential matters out of AI
- Board approves minutes
- Survey trustees on packets
- Review metrics annually
- Confidentiality: Executive session, personnel, and patient matters stay out of AI tools.
- Accuracy: Figures tie to audited and management reports.
- Quality data: Verified by quality and clinical leaders.
- Public entities: Open meetings and records laws may apply.
- Distribution: Executives send materials.
Set Up & Get Running
If you already use prompts, here is how to turn one into a repeatable skill and then a useful first agent. Start with a task that already works.
From prompt to first agent
- Choose a recurring business task
- Use aggregate data with no PHI
- Good first choices: variance narrative, supply spend analysis, or denial root causes from aggregate data
- Write down inputs, output, and reviewer
- Test on three real cycles
- Improve when results are unclear
- Start with aggregate exports
- Bring PHI-capable tools through governance and a BAA
- Use vendor AI inside systems where approved
- Start with a monitoring task
- Approve through the AI governance committee
- Track time saved and dollars recovered
It's useful, measurable, and uses no PHI. It compares purchases to contract prices and flags off-contract and price-variance buys by department. It doesn't place or cancel orders.
Copy these builders
Use the first after a prompt works. Use the second only after the skill has been tested on real work.
Days 1-30: Foundation
- Name an executive sponsor and involve the privacy officer
- Confirm the business AI policy draft and data rules
- Set up enterprise accounts
- Build aggregate exports
- Pick 1-2 pilots without PHI
- Record baselines
Days 31-60: Expand
- Save pilots as skills
- Bring one PHI-adjacent use case (e.g., appeals) through governance
- Review vendor AI features and BAAs
- Train finance and revenue cycle teams
- Measure time saved and dollars recovered
Days 61-90: Standardize
- Adopt the business AI policy
- Pilot one review-only agent with audit logs
- Write SOPs
- Report results to the CFO and governance committee
- Maintain the AI use case inventory
- Plan next wave
Implementation Success Measures
Measurement30-Day Targets
- No PHI in unapproved tools
- Pilots running
- Baselines recorded
- Governance engaged
60-Day Targets
- Time or dollars measured
- One governed PHI-adjacent use case reviewed
- Vendor AI reviewed
- Teams trained
90-Day Targets
- Policy adopted
- First agent piloted with logs
- SOPs written
- Inventory current
- Decision on next wave
How the Pieces Fit Together
Start with a prompt. Save the ones that work. Add the systems and files they need. Then use an agent for recurring work, with someone checking the result.
- Ask AI to do one clear task
- Use it for a quick draft, analysis, or explanation
- Check the result before you use it
- Save the step-by-step instructions for a task your team repeats
- Add your policies, examples, and review steps
- Update it when your process or regulations change
- Bring the right systems, files, and reports into the task
- Stop copying and pasting information by hand
- Start read-only, with only the sources the task needs
- Run a skill on a schedule or when something happens
- Gather the information, prepare the work, and flag issues
- Leave coding, clinical, staffing, and financial decisions with people
Where AI Can Help Health System Business Teams
Use AI to get routine business work into a clear first draft. Keep coding, clinical, staffing, and financial decisions with qualified people.
- Variance narratives and forecasts
- Cost report preparation
- Cost per case analysis
- Board finance reports
- Denial root causes and queues
- Appeal organization
- Underpayment detection
- KPI narratives
- Off-contract spend and renewals
- Contract labor and overtime analysis
- Grant budgets and reports
- UDS and 340B organization
- Coding and medical necessity
- Clinical staffing and products
- Filings and certifications
- Board and payer decisions
AI Capabilities Explained
No jargon. What each kind of AI does, and what it looks like on the business side of a health system.
Claude, ChatGPT, Gemini, and Copilot write, summarize, and follow instructions.
In a health system: variance narratives, board reports, policy drafts, grant narratives.
Enterprise AI offered for regulated health data under a BAA.
In a health system: approved appeal preparation workflows.
Connections to public reference data.
In a health system: CMS Coverage Database, ICD-10, and NPI Registry lookups.
AI agents built into EHR and revenue cycle systems.
In a health system: Epic's revenue cycle AI working denials inside Epic.
Compares records to rules and flags exceptions.
In a health system: payments vs. contracts, purchases vs. GPO prices.
Projects numbers from history and assumptions.
In a health system: volumes, cash, and labor needs.
Guardrails: PHI, Compliance & Accountability
Simple rules for business-side AI use in hospitals and health systems. Patients, regulators, and communities expect you to protect data and get the numbers right.
- No PHI in tools without a BAA and approval
- Default to aggregate and de-identified data
- Minimum necessary in every export
- Privacy officer approves exceptions
- Certified coders assign codes
- Physician advisors write clinical appeals
- Clinical leaders set staffing
- Committees decide clinical products
- Cost reports certified by people
- UDS, 340B, and grant reports submitted by staff
- Price transparency reviewed by compliance
- Tax filings by tax professionals
- AI governance committee approves use cases
- Vendor AI reviewed for PHI and actions
- Audit logs for agents
- Accuracy monitoring
- Financial assistance publicized and applied
- EMTALA always respected
- Collection actions follow rules
- Plain-language communication
- Coding and medical necessity
- Staffing levels and clinical decisions
- Financial approvals and filings
- Payer negotiations and board decisions
Business-Side AI Readiness Checklist
ControlsLeadership & Policy
- Executive sponsor for business AI
- AI governance committee includes finance and revenue cycle
- Business AI policy adopted
- BAAs in place for PHI-capable tools
- Role-based training delivered
Practice & Review
- Aggregate data by default
- Human approval on filings, appeals, and financial decisions
- Vendor AI features reviewed
- AI use case inventory maintained
- Quarterly review of outcomes and incidents
Scope: Finance, revenue cycle, supply chain, workforce, grants, compliance, and administrative functions. Clinical AI use is governed by a separate clinical AI policy.
Approved tools: [e.g., enterprise AI accounts; HIPAA-ready environment under BAA for approved PHI workflows; reviewed AI features in our EHR and ERP]
Data rules: Aggregate and de-identified data by default. PHI only in tools approved under a BAA, for approved use cases, with minimum necessary data.
Human decisions: coding, medical necessity, clinical appeal content, staffing levels, clinical product choices, financial approvals, regulatory filings, payer negotiations.
Governance: New use cases go through the AI governance committee. Agents keep audit logs. Outputs are monitored for accuracy.
Approval: [CFO / Chief Compliance Officer / AI Governance Committee], [date]
AI Maturity Model for Health System Business Teams
See where your business functions are today. Choose a realistic next step.
Curious + Ad Hoc
- Individuals use consumer AI
- No business AI policy
- Vendor AI unreviewed
- No inventory
Enterprise Accounts & Rules
- Enterprise accounts
- PHI rules and BAAs
- Prompt libraries
- Aggregate exports
Embedded Workflows
- Skills in finance, revenue cycle, supply chain
- Governance approval process
- Vendor AI reviewed
- Measured results
Governed Agents
- Review-only agents with audit logs
- HIPAA-ready environment for approved PHI workflows
- Accuracy monitoring
- Annual governance review
Maturity Self-Assessment
AssessmentLeadership
- Sponsor
- Governance committee
- Policy
- Training
Practice
- Prompt library
- Skills
- SOPs
- Exports
Protection
- PHI rules
- BAAs
- Human decisions
- Audit logs
Measurement
- Baseline
- Dollars recovered
- Time saved
- Quarterly review