The Art of the Exact Prompt: How Specificity Saves Schedules (and Relationships)
Karen is my partner but is not a code share.
She spent many years on Cape Cod and wanted to help a neighbor of ours who is going to visit there. She threw some generic prompts into Gemini. It came back with a list of the usual "iconic" sights to see. When I tried to recreate her exact search later, I continually got completely different results—which, of course, allowed Karen to label me as "moronic."
That's the thing with standard AI prompts: they are a moving target. But when you're dealing with commercial airline schedules, you can't afford moving targets. You need repeatable, complete, and exacting results.
We have a client with its own hub and a commuter "bridge" to a codeshare partner's hub. I wanted to write a prompt tying the bridge to the partner hub to ensure local traffic was prioritized and the timing was spot-on with the banks it was designed to connect into.
Here was my first attempt:
PROMPT 1:
Prioritizing a complimentary shuttle schedule between ORIGIN and DESTINATION for us and our partner carrier combined, what changes should we make to the week of November 2 (The Winter Schedule) to increase connectivity with partner banks at DESTINATION - inbound and outbound? Exclude code shares.
Maestrow came back with some solid data looking at the current schedule against the partner's "waves":
Code Share Partner Connection Windows (DESTINATION)
| Wave | Inbound Peak (Arrivals) | Outbound Peak (Departures) |
|---|---|---|
| Wave 1 | 07:00 – 08:30 | 09:00 – 10:30 |
| Wave 2 | 11:00 – 12:30 | 13:00 – 14:30 |
| Wave 3 | 15:00 – 16:30 | 17:00 – 18:30 & 21:00 |
Current Shuttle vs. Partner Waves
The analysis showed our current "spread" shuttle was well-distributed but had several sub-optimal connection windows:
- Flight 281 (08:00 – 12:30): Tight. Misses most Wave 1 arrivals.
- Flight 285 (11:40 – 16:10): Sub-optimal. Falls off-peak between Waves 1 and 2.
- Flight 287 (13:55 – 18:25): Excellent. Connects smoothly from Wave 2 arrivals.
- Flight 283 (19:40 – 00:10): Good. Connects from Wave 3 arrivals.
- Flight 284 (07:50 – 10:40): Excellent. Feeds Wave 2 departures.
- Flight 286 (10:05 – 12:55): Tight. Feeds late Wave 2 or Wave 3.
- Flight 282 (15:50 – 18:40): Excellent. Feeds the 21:00 departure wave.
The Recommendation
Maestrow suggested re-timing existing flights (e.g., moving Flight 281 to 09:30 to capture massive long-haul traffic from Wave 1 arrivals). The Executive Summary concluded: "To maximize the partnership... transition from a 'spread' schedule to a 'pulsed' schedule that mirrors the partner hub waves."
The Pivot: Adding the "Exacting" Details
While that first response was good, it didn't fully execute what I actually intended. It gave me data, but it didn't give me the full breadth of operational levers. So, I rewrote the prompt and added a critical line specifying the exact types of changes allowed:
PROMPT 2 (The Exacting Prompt):
Prioritizing a complimentary shuttle schedule between ORIGIN and DESTINATION for us and our partner carrier combined, what changes should we make to the week of November 2 (The Winter Schedule) to increase connectivity with partner banks at DESTINATION - inbound and outbound? Changes should include retiming existing flights, adding flights, eliminating flights and changing equipment. Exclude code shares.
By adding that single, definitive layer of constraints, Maestrow didn't just analyze the schedule—it completely optimized the operational fleet strategy:
Maestrow Optimized Output
- Shift Outbound Peaks: Move departures to the 09:30, 13:30, and 18:00 windows to lock into the three primary partner arrival banks.
- Shift Inbound Peaks: Target arrivals at 08:30, 11:30, and 18:30 to feed the three primary partner departure banks.
- Capacity Alignment: Deploy the more efficient 7MAX fleet on the retimed Wave 1 and Wave 3 flights to handle the expected surge in high-yield connecting traffic.
- Connectivity Gain: These precise adjustments are projected to increase viable connection pairs with the partner by 25-30%—without increasing the total number of aircraft frames required.
The Reality Check
Being highly specific doesn't just improve your results; it makes those projections repeatable and operational.
Oh, and as for Karen? Armed with my newly refined prompting skills, I rewrote her Cape Cod prompts, got her the exact repeatable results she actually needed, and earned myself a well-deserved "atta boy."
Now, isn't that ironic?