Lesson objective
Apply a written test to every slide, demonstration and handout in a session, cut what fails it, and allocate the surviving minutes by the risk of the work rather than by the interest of the content.
Plain-language explanation
Why this matters now
Backwards design produces a small design. That is its main practical benefit and the reason most instructors abandon it halfway. Once the duties and the evidence are fixed, most of the material anybody would naturally prepare turns out to serve neither, and cutting it feels like admitting the session is thin.
It is not thin. It is the difference between a room that watched something and a room that did something. But the cut has to be made on a written test rather than on taste, because taste protects the material its owner enjoys.
There is also a question of where the surviving minutes go. Not every duty in a practice carries the same consequence, and an hour spent evenly across four duties is an hour that gave the riskiest one a quarter of the attention.
Guidance (voluntary, not binding): Joint Commission and CHAI, Guidance on the Responsible Use of AI in Healthcare, released 17 September 2025. Its fourth element says monitoring should be risk-based, with tools that inform or drive clinical decisions checked more often and tools that help with administrative or documentation tasks checked periodically. That is guidance about monitoring a tool, not about allocating a session, so borrowing its logic for time allocation is this programme's inference and must be labelled as one. The guidance is voluntary, is not a Joint Commission accreditation standard, describes itself as "an initial, high-level document", and was written mainly for hospitals and health systems, so it reaches a small practice by analogy.
Guidance (voluntary, not binding): OIG Compliance Program Guidance for Individual and Small Group Physician Practices, 65 Fed. Reg. 59434, 5 October 2000. It says explicitly that, unlike other OIG guidances, it "does not suggest that physician practices implement all seven components of a full scale compliance program". It also distinguishes innocent errors, which it calls erroneous claims, from reckless or intentional conduct, which it calls fraudulent claims, and it defines upcoding as "billing for a more expensive service than the one actually performed". It is archived, dated 2000, pre-dates the Affordable Care Act, says nothing about AI, and is voluntary throughout.
A question to sit with
You have thirty-five minutes of material and twenty-two minutes of room. The twelve minutes you would cut are the twelve you are proudest of. What written test decides, so that the decision is not made by whichever of you is more tired?
What goes wrong
1. Cutting by trimming. Every segment loses two minutes and the session keeps all its content at a shallower depth. Nothing gets practised twice, which means nothing gets practised properly once.
2. Cutting the practice round. It is the longest single block and the easiest to shorten, and it is the only part that produces evidence. Cutting it turns a design back into a talk.
3. Keeping context because it feels irresponsible to drop it. A history of the technology, a survey slide, a vendor landscape. None of it serves a duty. Some of it can live in a handout the practice keeps.
4. Even time across unequal risk. Four duties, four equal blocks, regardless of what each one touches. The clinician's signature and the front desk's disclosure decision do not carry the same consequence, and pretending they do is a decision, not neutrality.
5. Cutting the honest caveats. The status, the date, whom a rule binds and what is not recorded are the first things to go when a session runs long, and they are the things whose absence turns a session into misinformation. They are never candidates for the cut.
6. Cutting the debrief. The rounds run and the clock ends, so nothing is said about what happened in them. The debrief is where a learner finds out what they missed and why.
The practical answer
Best practice: cut in three passes, in this order. First pass, remove everything that serves no duty statement. Second pass, remove everything duplicating something already practised. Third pass, remove whole duties rather than shortening the rounds that remain, and record where each removed duty will be taught instead. Never take the third pass inside a round.
Best practice: allocate the surviving minutes by the consequence of the work, and label the reasoning. This programme's recommendation, by analogy to risk-based monitoring, is a rough ordering rather than a formula.
| Work the duty touches | Rough share of practice time | Why |
|---|---|---|
| Anything reaching a patient as clinical information, or the record and signature | The largest share | The consequence is a patient and a record, and the clinician owns both |
| Anything reaching a payer on a claim | Substantial | Documentation support, query practice and the obligations that follow an error |
| Disclosure decisions about how much of a record travels | Substantial | Made many times a day by the people given the least training time |
| Internal administrative output reviewed before use | The smallest share | Reviewed inside the practice before it reaches anyone |
That ordering is an inference from voluntary guidance about monitoring tools, not a rule, and no source here sets session time allocation at all. Present it as this programme's recommendation and invite the practice to reorder it for their own work.
Best practice: put the cut material somewhere the practice keeps. A one-page handout with the context, the sources page, the supervision slide and the honest silences will outlive the session. Cutting from the room is not the same as destroying, and saying so makes the cut easier to accept.
Best practice: protect the debrief in the plan itself. Write it as a numbered segment with its own minutes rather than as time at the end, or the overrun will eat it. Two minutes per round is usually enough to ask what was missed and why.
Best practice: keep the design small enough to be repeatable. A small practice will not run a ninety-minute session twice a year. It might run a twenty-minute one every other month, which is where a duty per session becomes an advantage rather than a compromise. OIG's 2000 guidance for small practices says it does not suggest implementing all seven components of a full scale compliance program; the same modesty applies to a training design, as this programme's recommendation rather than as anything OIG said about training.
Remember this
- Cut on a written test, not on taste; taste protects the material its owner enjoys.
- Three passes: material serving no duty, duplication, then whole duties. Never cut inside a practice round.
- Allocate minutes by the consequence of the work, labelled as this programme's inference from voluntary guidance about monitoring, not as a rule.
- Caveats, statuses, dates and honest silences are never candidates for the cut.
- Cut material goes to a handout the practice keeps; say out loud what you removed and why.
Medical practice example
Ifeoma Castellanos has built a fifty-two minute design for the fictional Meridian Street Dermatology and has been given twenty-five minutes. She writes the four-question cut test at the top of her notes and works down the plan. Out goes a nine-minute opening on how drafting tools produce text: it serves no duty statement, produces no evidence, and nothing in the room becomes harder without it. Out go two slides of adoption figures: same result, and they move to page one of the handout with their bases and dates intact. Out goes a six-minute vendor comparison she did not want to run anyway. That first pass leaves thirty-one minutes. The second pass finds duplication: she had learners cut a fictional drafted message down twice, once in a warm-up and once in the round, so the warm-up goes, leaving twenty-seven. The third pass is the painful one. She has four duties and room for three, so she removes the administrative sampling round entirely rather than shortening any of the others, and writes on the plan that it moves to a twenty-minute session for the manager and the compliance contact in six weeks. That leaves twenty-two minutes. She then allocates by consequence rather than evenly: eight minutes on the clinician round touching the record and the signature, seven on the front desk disclosure decision, four on the coding documentation round, and three minutes of debrief written as its own numbered segment so the overrun cannot eat it, which totals twenty-two. She checks that every caveat survived: the status and date on each rule, whom each binds, the sentence that no AI tool can supervise anyone, and the three questions she has no answer to. All of them are still there, because none of them was ever a candidate. At delivery she opens by naming what she removed and why, and hands out the four-page appendix. The practice owner, Dr Suniti Raval-Okoro, tells her afterwards that the list of what was cut was the first time anyone had explained why a training session was short.
Responsible-use warning
Never cut by trimming every segment: a session that keeps all its content at a shallower depth practises nothing properly. Never shorten a practice round to save time; remove a whole duty instead and record where it will be taught. Never cut a caveat, a status, a date, a statement of whom a rule binds or an honest silence. Never present the risk-based time allocation as a rule: it is this programme's inference from voluntary guidance about monitoring tools, and no source here allocates session time.
Human-review requirement
The instructor owns the cut and must be able to show the test applied to each removed item. The practice manager confirms the surviving design fits the time actually available and that the appendix will be kept where staff can find it. The practice owner decides which duties move to a later session and when. Where a duty touching the record, the signature or a patient communication is moved rather than taught, a licensed clinician confirms the practice is content to wait, and owns that decision.
Practical exercise
Cut an overlong fictional design down to the room you have, on a written test.
- Write a fictional session plan of at least forty minutes for a fictional practice, listing every segment with its minutes, including at least two segments you would enjoy delivering.
- Write the room you actually have as a number of minutes, no more than twenty-five.
- Write the four-question cut test at the top of your plan.
- Pass one: mark every segment serving no duty statement, and record for each which of the four questions it failed.
- Pass two: mark every segment duplicating something already practised elsewhere in the plan.
- Pass three: if you are still over, remove whole duties rather than shortening rounds, and record where each removed duty will be taught and when.
- Total the surviving minutes and show the addition.
- Allocate the practice minutes by the consequence of the work, write one sentence justifying each allocation, and label the reasoning as this programme's inference rather than as a rule.
- Write the debrief as its own numbered segment with its own minutes.
- List every caveat, status, date and honest silence in the design and confirm in one line that none was cut.
- Build the appendix contents page from everything removed, and write the two sentences you will say at the start of the session naming what you cut and why.
Knowledge check
-
What are the four questions of the cut test?
Answer guidance: Which duty statement does this serve; which piece of evidence does it help produce; what act would become harder without it; and is it a caveat, status or date attached to something being taught.
-
Why must a session never be cut by trimming every segment equally?
Answer guidance: Because the result keeps all the content at a shallower depth and nothing is practised properly once, which removes the only part of the session that produces evidence of a duty being performed.
-
How should an instructor label the risk-based allocation of session time?
Answer guidance: As this programme's inference by analogy from voluntary Joint Commission and CHAI guidance about risk-based monitoring of tools, since no source here addresses how session time should be allocated.
-
What does the 2000 OIG guidance for small physician practices say about scale?
Answer guidance: That unlike other OIG guidances it does not suggest physician practices implement all seven components of a full scale compliance program; it is voluntary, archived, dated 2000 and says nothing about AI.
Sources and further reading
- Joint Commission and Coalition for Health AI (CHAI) — Joint Commission and Coalition for Health AI (CHAI) Guidance on The Responsible Use of AI in Healthcare (RUAIH) (2025-09-17). Accessed 2026-09-15.
- U.S. Department of Health and Human Services, Office of Inspector General (HHS-OIG) — OIG Compliance Program Guidance for Individual and Small Group Physician Practices (65 FR 59434, October 5, 2000) (2000-10-05). Accessed 2026-09-15.
- AI Coalition Network — Certification Educational Standard: Human Review of AI-Assisted Professional Work. Accessed 2026-09-15.
- National Institute of Standards and Technology — Artificial Intelligence Risk Management Framework (AI RMF 1.0), NIST AI 100-1 (2023-01-26). Accessed 2026-09-15.