The Stance Lab

Aldermere General did 600 fewer surgeries than planned

From the CEO, to the executive team: We completed 600 fewer scheduled surgeries than planned from July to September. Please look into it from where you sit and come back with what you think the problem is and what we should do.

Aldermere General Hospital · Scheduled surgery results · July to September

Scheduled surgeries completed

5,400

Plan 6,000

−600 (−10%)

By campus and month

Campus	Jul plan	Jul actual	Aug plan	Aug actual	Sep plan	Sep actual	Difference
Main campus	1,600	1,420	1,600	1,445	1,600	1,465	−470
Riverside campus	400	400	400	400	400	270	−130
Total	2,000	1,820	2,000	1,845	2,000	1,735	−600

By specialty

Specialty	Plan	Actual	Difference
Orthopedics (hips, knees)	1,800	1,530	−270
General surgery	1,500	1,390	−110
Eye surgery	1,200	1,010	−190
Cancer surgery	900	880	−20
Cardiac surgery	600	590	−10
Total	6,000	5,400	−600

How this works. Everyone agrees on the 600. Take one executive's seat, open that executive's data, and decide what the problem is. Then see what the other three concluded, what all the data shows together, and what AI could and could not do about it. About 25 minutes.

Aldermere General, the data and every number are invented, so you may upload these files to an AI tool. Based on how domain experts formulate problems in Joshi, Su & Zbaracki (2026), Organization Science.

Step 1 · take a seat

Which executive are you?

You will see only the data this person normally sees.

Step 2 · your data

These are your files. Look through them in Excel, or upload them to Claude with the CEO's page. Decide for yourself what to ask. Take about 10 minutes.

In one sentence, what problem should Aldermere solve? And what should Aldermere do about it?

Step 3 · same shortfall

Four seats, four problems

We gave Claude each executive's files and the CEO's page, and asked: “What problem should we solve?”

Real replies from Claude, shortened; one of two runs per seat, November 2026.

Notice. Nobody disagrees about the 600. Each seat's data leads to a different problem, each backed by real numbers. And none of them can say what connects them.

Step 4 · the solution follows the problem

What would you do, and which AI would you use?

Notice. Each problem leads to a different fix and a different use of AI, or to none. These are four very different projects, and each would look sensible to the person proposing it.

Step 5 · all ten files together

Then someone put all the data in one place

We gave Claude all ten files and asked the same question, “What problem should we solve?”

Claude · real reply, shortened

Solve the shortage of inpatient beds at the Main campus, not the operating room shortfall. Since late June, more older patients have been admitted through Emergency. Medicine wards were already full, so these patients are filling the surgical wards. Surgical patients then get stuck in recovery, the operating day runs late, and the last cases are cancelled.

[…] Emergency's bed wait fell from 10 to 8 hours, but surgery paid for that improvement.

[…] What the data can't show: why older patients' visits rose, how long they stay, or what is holding up their discharge.

One of three runs, November 2026. All three found this chain.

Read the full reply

Part of the 600 shortfall	Surgeries	Status
Riverside sterilizer failure (three rooms closed Sep 3–21)	-130	Over
Two new Main rooms opened Aug 17, not Jul 1	-160	Over
Main's existing rooms	-310	Getting worse

The CEO's page shows Main improving (-180, -155, -135), but the new rooms hide the trend. Without them, I estimate Main's gap went from about -13 to -105 to -190. That splits the new rooms' 340 cases by working days.

Why beds are the cause

Emergency: daily visits rose from 228 to about 249, almost all aged 65+. Admissions rose from 28 to 36 a day and have stayed there.
Surgical wards: Emergency placements rose from 77 in June to 283 in September. Occupancy went from 86% to 99%, with staffed beds unchanged.
Cancellations: "Ran out of operating time" at Main rose 61, 92, 122 (275 in total), about 70% of it orthopedics. These are mostly the last cases on the list, cancelled after 3 pm.
Daily link, July–September:

Surgical beds free	Days	Recovery wait (orthopedics, general)	"Ran out of time" cancellations per day
None	20	93 min	6.1
6–8	4	39 min	1.0

Comparison group: at the same campus, eye, cancer and cardiac recovery waits stayed near 20 minutes.

Why the label misleads: operating time isn't what's short. Main's room use fell from 84% to 77%, and with the new rooms open it still completes about 71 cases a day, the same as June. More room capacity won't help until beds free up. Emergency's bed wait fell from 10 to 8 hours, but surgery paid for that improvement.

Handle separately

"Patient not ready" cancellations: steady at about 65 a month at Main and unrelated to beds. Pre-op nurses fell from 5.9 to 4.0, though incomplete tests haven't risen yet.
Sterilizer B: it has failed three times in 18 months, and replacement was recommended in May. It's a quick risk to close.

What the data can't show: why older patients' visits rose, how long they stay, or what is holding up their discharge. It also doesn't show which wards each specialty uses, and cancellation reasons start only in July.

Those are the next questions to answer. If nothing changes, September's pace means roughly 400 fewer surgeries next quarter, assuming the same plan.

What no single seat could see

Emergency patients fill the surgical wards→Patients leaving surgery wait in recovery for a bed→The operating list runs late→The last patient is sent home

Those cancellations were recorded as “ran out of operating time,” so they looked like a scheduling problem. Emergency's numbers looked excellent, and nobody was wrong to move Emergency patients faster.

Where the 600 went

310

160

130

Main’s existing roomsBeds are the leading explanation; the exact share isn’t established

New operating rooms opened lateOver; the runs suggest beds would have limited them anyway

Riverside sterilizer failureOver, could recur

No file records a patient held for lack of a bed, so the chain is a strong inference, not proof. One of the three runs split the numbers differently; all three put beds first.

The tool did not do the hard part. It read all ten files because someone got them into one place. No executive had that access. It still needs a person to confirm the link, and to decide what to do about it.

Step 6 · can AI solve this problem?

Partly, and not the part that matters most

A better problem statement: “Emergency and surgical patients compete for the same beds, and nobody manages that trade-off for the whole hospital. How do we decide who gets beds so that we meet both goals?”

What this needs	Who or what can do it
Record “no bed available after surgery” as a cancellation reason	People. Until then, any AI learns from records that hide the problem
Predict Emergency admissions and discharges by ward, one to three days ahead	Predictive AI, built on linked patient and bed records
Schedule surgeries around the beds that will be free	Predictive AI plus scheduling tools, with people approving
Find what is holding up discharges, and draft discharge summaries	Generative AI can help, with clinicians checking
Decide who waits when there is no bed	People. It is a clinical and ethical choice
Name one owner across Emergency, Medicine and Surgery; fund more medicine capacity	People
Replace the sterilizer; match the plan to confirmed funding	People. No AI needed

The same AI, aimed differently. A bed tool built for “fill beds faster” would have cancelled more surgeries. A bed tool built for “balance beds across the hospital” could help. The problem statement decides what the AI works toward. This is our view, and a better problem statement is still a choice about whose goals count.

Who should have been in the room before Aldermere wrote its problem statement?

Bring your answer to the next class.

Keep going

Try another seat

Your answers stay saved.

Run all ten files yourself

Upload every file to Claude with the CEO's page. Ask: What problem should we solve? Yours will differ from ours.

Same shortfall. Four problems.