Center for Practical AI

A health system, anywhere

Tuesday, 6:42 AM

A children's hospital. A community system. An academic medical center. Pick yours — the morning looks the same.

The people who run it are about to spend hours on work their AI tools could already help with. Most of them have never been shown how.

This is one Tuesday, for four of them.

The people in this story are illustrative composites. The work — and the friction — is real in every health system we've trained.

Scroll ↓

Today

6
wks

Six weeks later

Each Tuesday below is told twice: as it is today, and as it runs six weeks after each person built a capstone — a working AI workflow of their own design, presented at the end of their cohort.

“You don't leave with a certificate of attendance. You leave with a real AI workflow for your role.”

R

Renee · Clinical Nurse

Her Tuesday runs 7:00 AM to — officially — 5:30 PM.

7:10 AMToday

Fifteen tabs deep

Renee's first patient is complex, and the family traveled hours to make the 8:00 appointment. So she's in the chart at 7:10 — flowsheets, consults, med lists, two sets of labs — building the picture one click at a time.

Open in the chart

FlowsheetsConsultsMed adminLabs 06/09NotesLabs 5/28SummaryMessagesOrdersVitalsHistoryImagingCare planAllergiesResults

She'll do this five more times before lunch.

7:10 AMSix weeks later

The picture, assembled

Her capstone ran overnight: the draft summary is waiting, flags on top. Renee verifies it against the chart — that's the job now, verification, not excavation — and is in the hallway early when the family arrives.

The chart — 15 tabs

Encounter notes (47)

Specialist consults — pulm, cardio, neuro

Med admin record · 3 active, 2 PRN

Labs 06/09 · Labs 05/28

Prior admission summary (11 pp.)

Pre-visit summary — drafted for Renee's review

Why they're here: 6-mo follow-up, post-op course stable

What changed since last visit: new PRN added 5/28

⚑ Flag: weight percentile drop — confirm with family

⚑ Flag: pending lab from outside clinic — chase result

Meds reconciled ✓ · Allergies confirmed ✓

Verified by Renee, 7:14 AM

Same chart. Same kid. Different morning.

Renee verifies every draft against the source chart before it touches care. The tool drafts; the nurse decides.
6:50 PMToday

The shift that ends at 5:30 ends at 7

Renee's shift ended at 5:30. She's still here, typing handoff summaries from memory and a day's worth of notes — the most important writing she does, done entirely after her job is over.

Handoff summaries · shift ended 5:30

Rm 4012 — signed
Rm 4013 — signed
Rm 4014 — not started
Rm 4015 — not started
Rm 4016 — not started
Rm 4017 — not started

Dinner waits again.

6:50 PMSix weeks later

She goes home at 5:30

Her capstone's second act: handoff drafts assembled from the day's notes, morning flags carried forward automatically. She reviews, fixes the nuance only she knows, signs — and is out the door at 5:31.

A day of notes — 6 patients

Vitals q4h × 6 patients

Med admin events (31)

Family conversations — 3 significant

New orders after 3 PM (4)

Pending: outside lab, imaging read

Handoff drafts — for Renee's review & signature

Rm 4012: post-op day 2, pain managed, mom staying overnight

⚑ Rm 4015: new PRN after 3 PM — watch response

Rm 4018: discharge likely AM, education started

⚑ Pending lab from outside clinic — night shift to chase

Reviewed & signed by Renee, 5:24 PM

One workflow, both ends of the shift. Capstones compound.

Every handoff is reviewed, corrected, and signed by Renee. The draft saves her time; her judgment makes it safe.

Renee's ledger — her Tuesday's documentation load

11525min

The capstone — a workflow Renee designed, week 6

The Pre-Visit & Handoff Workflow

Modern EHRs increasingly ship with AI that can draft chart summaries and end-of-shift notes — that part is product, not magic. Renee's capstone is the workflow she designed around it: when drafts run, what gets verified against the chart before anything touches care, and how morning flags carry automatically into the evening handoff.

Built on your EHR's own AI drafting features, deployed through clinical informatics — the governed path your clinical tools already follow.

Frameworks Renee drew on:
M

Marcus · Patient Access

His queue was longer when he arrived than when he left last night. It always is.

9:30 AMToday

The queue grew overnight

Each prior auth is the same ritual: pull the order, find the payer's requirements, hunt the documentation, assemble, submit, log. Twenty minutes when it's clean. Most aren't — and the on-hold queue of families blinks on his other monitor.

23

authorizations waiting

+6 overnight

+ 18 more below the fold

Every packet he's assembling is a family waiting to schedule care.

9:30 AMSix weeks later

The packet, pre-assembled

His capstone drafts the ritual's first half — requirements matched to documentation, gaps in red. Marcus reviews, fixes, submits. Seven minutes when it's clean, and the recovered time goes where it should have been all along: on the phone with a family, walking through what their coverage means.

Prior-authorization packet — drafted for review

Physician order
Payer criteria — matched
Clinical notes, 2 of 3 located
!⚑ Missing: outside imaging report
Demographics & coverage verified
Reviewed & submitted by Marcus, 9:41 AM

The queue still grows. He's faster than it now.

Every packet is human-reviewed before submission. The workflow assembles; Marcus is accountable for what goes out.

Marcus's ledger — his morning auth block

5010min

The capstone — a workflow Marcus designed, week 6

The Prior-Auth Packet Assembler

A workflow Marcus designed in the enterprise AI assistant his team already licenses: it drafts each authorization packet from the payer's published requirements — documents matched, gaps flagged red, nothing submitted without his review.

Runs in the enterprise AI assistant most health systems already deploy (e.g., a BAA-covered Microsoft 365 Copilot tenant).

Frameworks Marcus drew on:
P

Priya · HR Generalist

Three open reqs, and every clinical posting competes with systems three states away.

11:45 AMToday

Three open reqs and a buried answer

The posting drafts are stale copies of old postings. Before she can touch them, a manager calls with a float-pool policy question. She knows the answer is in the handbook — somewhere. The reqs age another day.

Open requisitions

RN, Peds ICUDay 12
Resp. TherapistDay 9
Patient Access RepDay 5

The candidates she's losing won't wait for the handbook search.

11:45 AMSix weeks later

Drafts that start at version three

Her capstone drafts the postings from approved templates — three reqs out before lunch. The float-pool question takes four minutes: her handbook workflow answers in plain language with the policy citation, she reads the cited section, confirms, calls back.

“Can a part-time RN in the float pool pick up weekend ICU shifts?”

Yes, if they hold current ICU competency validation and the shift keeps them under the part-time hours cap for the pay period.

Source: Team Member Handbook § 4.2.3 — Float Pool Eligibility
Source confirmed by Priya before responding

The handbook didn't get shorter. The path through it did.

Every posting is human-edited before publishing; every policy answer carries its citation, and Priya reads the source before she relies on it.

Priya's ledger — postings + policy lookups

3510min

The capstone — a workflow Priya designed, week 6

The Posting & Policy Desk

Two workflows Priya designed and presented as one capstone: postings drafted from the approved template library so her first edit is judgment, not transcription — and a handbook answerer she built to refuse any answer without a section citation.

Runs in the enterprise AI assistant your teams already license, against your own approved templates and handbook.

Frameworks Priya drew on:
D

Dana · Clinic Manager

Two call-outs by 1:00. The afternoon clinic is a puzzle with a deadline.

1:15 PMToday

Schedule Tetris

She works the options in her head, on a sticky note, in the scheduling screen: who's cross-trained, who's at their cap, whether the 2:15 block survives. It takes most of an hour. The option she picks is fine — she'll never know if it was the best one.

This afternoon · 2 uncovered

?
?

2:15 block — family rescheduled twice already

Every unfilled slot this afternoon is a family rescheduling a trip.

1:15 PMSix weeks later

Three options, eyes open

Her capstone lays out three coverage scenarios side by side — overtime risk, throughput, who gets stretched. Dana still decides; she's the one who knows Kim is burned out and the 2:15 family has been rescheduled twice. Twelve minutes, door to door, and the afternoon clinic runs full.

1:15 PM — two call-outs

Afternoon clinic

2 gaps · 2:15 block at risk

Option analysis

in Dana's head + a sticky note

Time to decision

~50 minutes

1:27 PM — decision made

Afternoon clinic

fully covered · 2:15 block runs

Option analysis

3 scenarios, costs shown side-by-side

Time to decision

12 minutes — decision Dana's

Selected & entered by Dana

The judgment was always hers. Now it has options to judge.

The workflow proposes scenarios; Dana weighs what no roster shows and makes the call. The schedule of record stays hers.

Dana's ledger — the coverage decision

4510min

The capstone — a workflow Dana designed, week 6

The Coverage Scenario Builder

A workflow Dana designed in her enterprise AI assistant: she gives it the constraints — who's out, who's cross-trained, hours caps, the afternoon template — and it lays out three coverage scenarios with what each costs. She decides, and enters the schedule in the system of record herself.

Runs in the enterprise AI assistant your teams license. The schedule of record stays in your scheduling system, entered by Dana.

Frameworks Dana drew on:

Now multiply one Tuesday.

Across four people, 245 minutes of Tuesday friction became 55.

If proficiency returns even 20 minutes a week to each team member, a 3,000-person system recovers roughly 48,000 hours a year.

Those hours don't disappear — they go back into care, into families, into the work that mattered before the friction.

Four people. Four capstones. One Tuesday transformed.

Renee · Clinical Nurse

The Pre-Visit & Handoff Workflow

verified against the chart, deployed with informatics

Marcus · Patient Access

The Prior-Auth Packet Assembler

human-reviewed before every submission

Priya · HR Generalist

The Posting & Policy Desk

refuses to answer without a citation

Dana · Clinic Manager

The Coverage Scenario Builder

the decision stays human

A cohort graduates up to 40 capstones.

Run the program across your workforce and the library compounds.

That's a growing library of working AI workflows your system would own — built by your people, for your work, reviewed by your standards.

Most health systems already have the ideas: an innovation team, a pipeline of staff suggestions, pockets of brilliant local fixes. What's missing is the last mile — the skills to let every team build its own answer.

Innovation centers concentrate capability in one room. Proficiency training distributes it to every department.

The certificate hangs on a wall. The capstone shows up for work on Tuesday.

Every one of those capstones started in the same place: the certification.

This is what the certification produces.

Every workflow in this story is the kind of thing a graduate builds — not because they learned to code, but because they learned when to trust AI, when to question it, and how to wrap human judgment around it.

The curriculum is live and public. Each framework below opens onto the actual lesson your teams would use.

Every role gets its own workshop.

The four role-based workshops below are tailored to real work — built with your teams, for your scenarios. Tell us which roles to start with.

The plan, in one scroll

Three layers: baseline proficiency for everyone, role-based application per team, and executive training for leaders. The curriculum below is live — every module links to the actual lesson your teams would use.

1

All team members

Baseline AI Proficiency

6 live, 1-hour sessions in role-based cohorts — or internal trainers, or interactive online modules

2

All team members, tailored by role

Persona-Based Application

1 live, 2-hour workshop per team — the four role workshops in this story are working examples

3

Senior leadership

Executive Training

Tailored session & coaching for your leaders

The six modules — live, not described

Every framework below is a door. They open onto the actual public lessons your cohorts would use.

Mod 01AI Foundations
FoundationsEthics
Mod 02Applied Basics
FoundationsInteraction
Mod 03When, Where, & Whether
EthicsJudgment
Mod 04Reimagined Workflows
InteractionOrchestration
Mod 05Applications of Co-Intelligence
JudgmentEvaluation
Mod 06Capstone Presentation
EvaluationOrchestration

A representative rollout — tailored to your calendar

Scheduling adapts to team availability and business priorities while preserving the cadence. Most systems run baseline cohorts first, then advanced role workshops in waves.

Weeks 1–2
Plan & schedule
Weeks 3–8
First cohorts · baseline training
Weeks 9–10
Review & build role workshops
Weeks 11–14
First cohorts · advanced workshops
Weeks 11–18
Next cohorts · baseline training

What's at stake in each domain

Companions to the live curriculum, not replacements for it. Each interactive shows what goes wrong when one of the six domains doesn't take hold — try one and you'll feel the gap the lessons close.

Start the conversation

Bring this to your teams.

Tell us about your system and the roles you'd start with. We respond within two business days to schedule a free discovery call — no obligation, no slides you've seen before.

We respond within 2 business days. Or email contact@cp-ai.org directly.

Your mission is to care for patients and families.

Ours is to give the people who do it their Tuesdays back.

Request a consultation