Portside POV: Build vs. Buy

Roman Mogylatov

5–8 minutes
Blue glass-effect illustration of a balance scale comparing custom software development and ready-made aviation software solutions, with a private jet flying overhead.

AI can build it. But who’s going to run it?

AI coding tools have changed the build-or-buy question. Someone on your team can describe a scheduling screen or a trip-request portal to an AI tool and get a working version back, fast. People call this “vibe-coding.” So it’s fair to ask, “Why keep paying a vendor when we could build our own?”

It’s a good question, and sometimes building is the right answer. But the screen is the easy part. Here are seven things to think through first.

1. AI can write the first version. The hard part is what comes after.

AI really does make the first version fast. That part isn’t hype, and it’s useful for testing ideas.

But a fast first version isn’t a finished system. Fabien Cros, chief data and AI officer at the consulting firm Ducker Carlisle, put it this way: “It’s very easy to build something that is shiny, but those things don’t run properly. They are vibe-coded. They are not on a proper IT infrastructure. They are not secure.” [1] He was talking about software in general, but the point carries over.

2. The bill doesn’t stop when the software goes live. Upkeep is over half the cost.

Most of what software costs comes after launch. Published research puts upkeep at more than half of a system’s lifetime cost, with estimates running up to 80% or more [2].

That upkeep is ordinary, steady work. Phones and browser updates. Security holes turn up and need fixing. The outside data you depend on changes: weather, airport notices, the other tools your team uses. Each change can break something. None of that shows up in the build estimate. So when you weigh building against buying, compare what the system will cost over its whole life with what a subscription costs over the same years. Someone has to own this work. Without a dedicated team, it lands on top of someone’s day job. It also means paying to host and run it. Servers, cloud bills, backups and uptime are ongoing costs, and someone has to read the logs and watch performance monitoring to catch problems before your dispatchers do. That’s a specialist skill, and it rarely appears in the build estimate.

3. Many software projects run over. A few run way over.

Overruns are common: McKinsey and the University of Oxford found that large IT projects (starting budgets above $15 million) ran 45% over budget on average and delivered 56% less value than predicted [3]. The average isn’t the scary part. A separate study of 1,471 IT projects found that about one in six was a “black swan,” an extreme overrun nobody planned for. Those projects ran roughly 200% over budget and almost 70% over schedule [4]. These studies cover large IT projects in general, so read them as a warning, not a forecast. The lesson: risk isn’t spread evenly, and it’s hard to know in advance which project will be the one that goes badly.

4. The biggest operators built their own. That doesn’t mean you should.

It’s natural to look at the largest operators and copy them. Some run software they built themselves. We call the urge to follow them the Follow-the-Giants Fallacy.

Here’s the trap. A very large fleet can spread the cost of a software team across hundreds of aircraft. That cost doesn’t shrink much when the fleet does. A 30-aircraft operation still has to secure, run, fix and update its systems. So does a flight department with three jets.

For the giants, building can be the right call. For most operators, the smarter move is to copy the experience the giants give their owners and passengers, not how they staff their IT team.

5. Safety software has to hold up when someone checks.

A spreadsheet risk form can look finished. The real test comes when an auditor, insurer or investigator asks what happened and who signed off. That’s the job of your safety management system (SMS) and flight risk assessment (FRAT): a written record of each risk and each decision. If you build your own, it’s on you to keep it current and compliant.

Be careful with general-purpose AI, too. It can give confident wrong answers. Anything touching your manuals or a go/no-go call needs a person checking it.

  • If you hold a Part 135 certificate: you must have an SMS that meets the FAA’s Part 5 rule in place by May 28, 2027 [5].
  • If you fly only under Part 91: the rule doesn’t currently cover you. However, insurers, boards, owners and the executives you fly may still want to see a safety program. 

6. If you build it, it’s your job to keep it running and keep the data safe.

Your systems hold sensitive information: who’s flying, where they’re going, when, and who’s paying. If you build your own, protecting all of it falls to you.

Ask the hard questions early:

  • Who secures the system and tests it for weak spots? 
  • Where does the data live, and who can see it? 
  • Who fixes a security hole the day it’s found?
  • Who watches the logs and performance monitoring, and who gets the call at 2 a.m. when something slows down or fails?

Others may ask, too. Your company’s IT security team, your insurer or a large charter client may want to know how you protect data before they sign off. With a home-built system, someone on your team has to have those answers.

7. Aviation software doesn’t work alone. It connects to everything around it.

Scheduling, safety, maintenance, weather, flight data and your other systems all need to share information. Portside partners with many companies and integrates with them, so you see a full picture of your operation. We also bring what we’ve learned from 1,100 customers and 15,000 aircraft, so every customer benefits from what works across the industry, and the whole industry gets better. A system built for one operation sees only that operation. And every connection you build is yours to secure and fix when the other side changes.

8. When building makes sense

Building can be the right call. It’s worth serious thought when all three are true:

  • You have a standing team to own it for years, not just to launch it.
  • You have a workflow that’s truly unique, and no vendor offers it.
  • You have a fleet and budget that can carry the cost for the system’s whole life.

If that’s you, you may consider building. If it isn’t, talk to people who build and run this kind of software full time, and ask them the same hard questions. At Portside, our 200+ team members bring more than 2,500 combined years of software experience and 2,750 in aviation. We’re here to support scheduling built for your operation, with safety built in, and AI built to work for you, all connected to the partners and systems you already rely on for a complete picture.


Sources

  1. FAA, 14 CFR Part 5, and the 2024 SMS final rule (published April 26, 2024). 
  2. TechTarget, “SaaSpocalypse? Maybe not, but SaaS applications are changing,” March 5, 2026. https://www.techtarget.com/news/SaaSpocalypse-Maybe-not-but-SaaS-applications-are-changing/857248/
  3. Published software-maintenance estimates, including S. Schach (1999, about 67% of lifetime cost) and T. Pigoski (2001, over 80%). [VERIFY primary texts]
  4. McKinsey & Company with the University of Oxford (M. Bloch, S. Blumberg, J. Laartz), “Delivering large-scale IT projects on time, on budget, and on value,” October 1, 2012. https://www.mckinsey.com/capabilities/tech-and-ai/our-insights/delivering-large-scale-it-projects-on-time-on-budget-and-on-value
  5. B. Flyvbjerg and A. Budzier, “Why Your IT Project May Be Riskier Than You Think,” Harvard Business Review, Vol. 89, No. 9 (September 2011), pp. 23-25 (1,471 IT projects). https://hbr.org/2011/09/why-your-it-project-may-be-riskier-than-you-think Preprint: arXiv:1304.0265.

Back to Insights