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)

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?

“You look like the Michelin Man with jaundice.”

Ok then. New jacket it is.

Gemini Prompt:

Armed with a list of brands, I went to the store and met a 33 year sales veteran who calls himself Red:

“I’m looking for this [Insert Expensive Brand] jacket.”

“Why?” said Red.

“Ummm, I want something fashionable?”

Red said: “You don't really want fashion. You want comfort.”

Fifteen minutes later, I had a goose down parka that cost 38% less than I thought. How had I screwed up? Simple: The Prompt. I had over-indexed on "fashion" and neglected the "local expertise" constraint.

From the Slopes to the Runway

This brings us to our Scheduling conundrum of the day.

I recently sat down with Maestrow to analyze a carrier revamping its network in Q1 2026. The goal? Shedding unprofitable routes. My initial prompts focused heavily on frequency reductions. The results were great - graphics, analysis, and intelligence on where the carrier was pulling back.

But I made the "Yellow Jacket" mistake again. I was looking at the cuts, but I hadn’t asked the most important question: “What is happening to the planes?”

When I adjusted the prompts to look for insight into next steps - the "dry" data became a much different tale. The real story was a Fortress Hub expansion. The carrier wasn't just shrinking; they were reallocating to out-schedule on primary routes.

I almost missed this!

Your Intellectual Partner

Using an AI platform like Maestrow is a constant exercise in discovery. Our CIO, Chris, and I often "get into it" (sometimes in a collegial manner) regarding “Teaching the User” vs. “The User Experience.”

Chris argues that the system should be intuitive—and it is. Maestrow gets smarter daily by digesting our documentation libraries and existing prompt histories. However, I have to (grudgingly) admit that Chris is right: Prompt Context is everything. If you ask for a "fashionable" schedule, you’ll get one that looks good on a slide deck.

But if you treat the system as your intellectual partner and ask about the effects of asset reallocation, you get the "Red" version: the truth that saves you 38% and reveals the competitor's next move.

So, what have we learned?

1. Always ask what happens to the planes when the route is cut.
2. The right prompt turns "dry" data into actionable strategy.
3. My doctor has officially confirmed I do not have jaundice.

This wasn't just interesting news; it was a strategic trigger. The key question immediately became: How do we find out if these new planes will be used to add new routes, or if they'll be used to shift capacity on existing ones? Manually sifting through thousands of flights at a hub airport to compare schedules would take a team of analysts hours, if not days.

This is where prompt engineering turns a tedious task into actionable market intelligence. We turned to an analytical prompt to do the heavy lifting.

Prompt:

For [Origin Airport Code], compare the current published schedule with 
the previous one. 
Identify and list any new or removed destinations over the next 10 months
Exclude code shares and include the carriers involved and the 
monthly change in flight count. 
Format your response in a clear table."

This single command contains several key elements of a powerful prompt:

1. A Specific Target: [Origin Airport Code] tells the AI exactly where to focus.
2. A Clear Task: compare the current published schedule with the previous one.
3. Precise Constraints: Exclude code shares is crucial for avoiding misleading data.
4. Structured Output: Include the carriers involved and the monthly change in flight count ensures the result is immediately useful for analysis.

Answer:

 There are 2 destinations added and 1 destination removed from your 
 target airport’s schedule. 
 Would you like a detailed breakdown of these changes?

A simple Yes, please from my end was all it took for Maestrow to present the full, detailed report.

This is the real power of modern AI. It’s not about replacing human insight; it's about amplifying it. In a matter of minutes, we had the data needed to understand a competitor’s moves and begin formulating a counter-strategy—all without spending hours on a spreadsheet.

This is just the beginning of what's possible with prompt engineering in airline analytics. In upcoming posts, I’ll share how we’re using similar techniques to compare public schedules against private, non-published ones and how we monitor for the subtle signs of network shifts.

For now, this intelligence is more than enough for a team to take action. And with that, I’m going to have another cup of coffee.

Agentic AI stands apart with its ability to:

• Autonomously monitor a variety of inputs—schedules (published and unpublished), traffic data, revenue and cost data and any private airline data a user would want included in the various analyses performed in the platform.
• Continuously re-optimize routes and schedules on the fly as conditions evolve. For instance, as one industry report notes, agentic systems can “continuously re-optimize routes and schedules as new data comes in, improving reliability and reducing costs.”

This autonomous orchestration enables airlines to plan not just once per season—but to evolve schedule integrity and network efficiency as new data arrives.

How Can Maestrow AI Lead the Next Wave of Route & Schedule Planning?

1. Establish Live Data Feeds & Feedback Loops Build connections to real-time sources: APIs, traffic data, revenue & cost data, demand signals. Agentic AI thrives on freshness—continuous inputs make continuous optimization possible.

2. Deploy Proprietary Specialized Agents Trained On Route & Schedule Planning

   • Aircraft Agent: Finds aircraft groups, categories, and types.
   • Station Agent: Finds stations/airports.
   • Carrier Agent: Finds carriers, types, alliances, and interline agreements.
   • Flight Agent: Finds direct and connecting flights.
   • MCT Agent: Finds minimum connecting times (MCT).
   • SSIM Agent: Finds SSIM objects.
   • General Agent: Finds general objects.
   • Route Agent: Finds routes.
   • Analytics Agent: Performs analytics and count queries.
   • Document Agent: Retrieves snippets and definitions from the knowledge base documentation.
   • Google Agent: Answers questions using Google Search (with user permission).
   • Image Agent: Generates and retrieves images.
   • Chart Agent: Creates and generates charts.

3. Iterate with Human-Approved Actions Present agent-proposed adjustments (e.g., shifting a flight by 15 minutes to preserve critical connections or respond to changes in a competitor’s schedule). Planners retain full authority to accept, modify, or reject each recommendation.

4. Orchestrate the Impossible Now, airlines can easily...

·      Assess competitor performance for benchmarking.

·      Track and evaluate schedule changes over time.

·      Leverage optimal codeshare connections.

·      Identify opportunities for new air services.

·      Predict travel demand trends.

·      Enhance and optimize flight networks.

·      Improve planning strategies and revenue models.

Final Thoughts: Why Now—and Why Maestrow.ai?

• **AI Maturity = Unmatched Opportunity—**LLMs, reinforcement learning, and multi-agent frameworks are ready.
• Productivity—Agentic AI can compress weeks of planning into hours—creating a game-changing leap in efficiency and competitiveness.
• Enduring value—Agentic AI amplifies network competitiveness and resilience, empowering planners to do more, faster, with foresight that is “Beyond Information To Intelligence With Actionable Insights & Recommendations.”

At Maestrow.AI , adopting agentic AI in route and schedule design allows us to revolutionize network planning—from seasonally revised schedules to continuous, data-driven orchestration, resulting in optimized schedules with trusted human oversight.