I tested my rooming list GPT on 200 attendees.

GPT-5.4 dropped on Friday, and I've been stress-testing all of my custom GPTs all of this week

Some of the improvements are subtle, this one was not.

The Rooming List Manager GPT was one of the first tools I built — and honestly, it's always been useful but kind of annoying to use. The original version could only reliably handle about 25 names at a time. Anything more and it'd start losing data, mixing up assignments, or just giving up halfway through.

25 names at a time is still faster than doing it by hand. But for a 200-person conference, that means 8 separate runs, stitching the outputs together manually, and hoping nothing got scrambled in between. It saved time, but it also added a new kind of headache.

So when GPT-5.4 launched, the Rooming List Manager was the first thing I went back to test.

🤖 GPT of the Week: Rooming List Manager GPT

THE PROBLEM: A rooming list for a 200-person conference across 4 hotels is a frustrating data entry task. ADA-needing guests must get ADA rooms, VIPs must be at the right properties, and room inventory can't exceed each hotel's capacity. Special requests must be standardized in the hotel's PMS system. Miss any of it, and you're dealing with the fallout on-site.

THE SOLUTION: The Rooming List Manager GPT runs constraint-based optimization across your full attendee list — matching accessibility needs to room types, placing VIPs, respecting inventory limits, and flagging issues before you send anything to the hotels.

All 200 names. One run.

I uploaded the full list — 200 attendees, 4 hotels, real complexity — and let it go.

The accuracy report: 7 out of 8 checks passed.

✅ All 200 attendees present — perfect match

✅ All 54 VIPs, Speakers, and Executives placed at the correct hotel

✅ All 38 ADA-needing attendees matched to ADA rooms

✅ Service animal ground-floor notes applied across all 7 attendees

✅ Room inventory within limits at every hotel

✅ Hotel notes column added cleanly

✅ Zero data loss — 200/200 rows intact

❌ One specific attendee swap — failed

The one miss: I asked it to move attendees #15 and #30 to a different hotel. It moved the wrong people. Likely interpreted the row numbers differently than intended — and one of those attendees was an Executive, so it may have kept him at the VIP hotel intentionally without flagging the conflict.

One mistake across 200 names. With the old version, running 200 names in 8 separate batches, the odds of something going wrong multiplied with every pass.

Who this is actually built for

This tool hits hardest for a very specific group of people — and if you're in it, you already know the pain I'm describing.

Independent meeting planners and hotel sourcing associates. People working with companies like ConferenceDirect, HelmsBriscoe, or Global Cynergies — firms where associates operate as independent contractors, building their own book of business on a commission-based model.

Here's how the economics typically work: you source the hotel block, negotiate the contract, and earn a percentage of the rooms commission when the event closes. The hotel might advance 5% at signing, with the balance settled after the program. Your income is tied directly to how many deals you can close and service.

The problem? Rooming list management is one of the most time-consuming parts of delivering on those deals — and many associates have historically had to outsource it, paying someone else a cut to handle the coordination, ADA matching, VIP placement, and hotel submissions. That fee comes straight out of your commission.

Now there are two ways this changes:

Either your housing specialist can use AI to do the same work in a fraction of the time — improving their own margins without sacrificing quality. Or you, the associate, can start handling it yourself, keep your full percentage, and deliver just as clean of a list.

This also applies directly to in-house corporate event teams managing multi-hotel blocks for annual conferences, and to third-party planning agencies that handle housing as part of a broader event management contract. Anywhere rooming list management is either being outsourced or eating up hours of a coordinator's week — this is the fix.

The 4 hours → 45 minutes math

The GPT handled the full 200-person assignment in about 12 minutes. Bulk updates took another 3. A proper human audit to catch anything it missed? Call it 30 minutes.

Total: ~45 minutes vs. 4 hours.

For an independent contractor where time literally equals commission, that math compounds fast across a full event calendar.

The lesson from the one miss

The GPT nailed the math. It failed on the nuance.

Constraint optimization — ADA matching, inventory limits, VIP rules — that's pure knowledge work. AI crushes it. But "move attendee #15 to Hyatt" requires interpreting ambiguous human intent. That's where it stumbled.

The fix is simple: always reference attendees by name, not row number. That one adjustment, and this tool is about as close to hands-off as anything I've tested (but you still need to watch for mistakes).

I score every event task on my AI Readiness Scale from 0–10. With this update, rooming lists have moved into the Green Zone at 9/10. The ceiling keeps rising — and the professionals who build these habits now are the ones who'll feel it most when the next model drops.

Till next time,

Noah Cheyer

Do More With Less Using AI

PS: Are you an independent associate managing your own housing? I'd love to know how you're currently handling rooming lists — and whether a tool like this would change your workflow. Reply and let me know.