On 13 November 2026, Ars Technica reported that the US Federal Aviation Administration is expanding AI advice for managing air traffic congestion across the entire US airspace. Airlines praised the tool. The air traffic controllers’ union wasn’t involved.
I couldn’t load the full article, so I’m working from that summary. I won’t guess at the tool’s design or its results. The summary already holds the part that matters for any business.
The excitement is about the tool
The airline reaction is easy to understand. Congestion costs airlines money, and a tool that advises on it sounds like a clear win. Airlines are the customers of the system, so their praise tells you the tool looks useful from the outside.
Praise from the people who benefit is a weak test of whether a rollout will work. I ask a different question first. Is there a real need? Congestion is a real problem, so yes. The next question is what changed. Here, the advice now covers the whole airspace instead of a part of it. What didn’t change is who has to act on that advice every day. Controllers are the people who run the process. A tool that advises them is only as useful as their willingness to work with it.
The technology is the easy part
I’ve said this in my talks, and I mean it plainly. The technology was the easy part. Changing how people think and work was the hard part, even for a tech team.
Building an advice tool is a solvable engineering problem. Getting the people who do the job to trust it, question it and use it is a different kind of work. It takes time, and it can’t be done after the fact by announcing the tool is live.
When the union that represents those people wasn’t involved, the FAA left that work for later. I can’t say how it will turn out. I can say what happens in ordinary businesses when the same step is skipped, because the pattern is familiar.
What this looks like in a normal business
Take a clinic that buys an AI tool to manage appointment bookings. The owner picks it after seeing a demo. The front desk team hears about it on the day it goes live. They weren’t asked how double bookings really happen, which patients need a phone call, or which doctors never accept a gap between appointments.
So they work around it. They keep their own notes, they override the tool’s suggestions quietly, and after a few weeks the owner decides the tool doesn’t work. The software did what it was built to do. The people who know the process were never brought in, so it was set up without their knowledge.
A retail shop, a service business or a marketing team follows the same path. The staff who run the process hold the details that decide whether the tool fits the work. Bring them in early and those details go into the setup. Leave them out and you find the details later, one complaint at a time.
Name who decides
Involving staff doesn’t mean everyone gets a vote. Someone has to decide, and the business should say who.
For each part of the process, write down who accepts the AI’s suggestion, who can override it and who is responsible for the outcome. If the answer is vague, staff will fill the gap with their own habits. In my own approach to AI, you direct, the AI produces, and you check and decide. You are responsible for every output that leaves your hands. That only works when “you” is a named person.
Keep rules where a wrong call is costly
Some decisions can go wrong cheaply. A badly worded reply can be corrected. Others can’t. Refunds, discounts, delivery promises and anything involving safety or money belong in fixed rules, not in the AI’s judgement.
I build this into AeroChat that way. The merchant sets limits, such as a refund amount, and anything above the limit goes to the team with the order details attached. The staff who handle those cases helped decide where the limits sit, because they know which cases cause trouble.
Air traffic is an extreme case of the same idea, since a wrong call there is costly in the highest sense. I don’t know how the FAA has drawn that line, and the summary doesn’t say. A business should still know where its own line is before the rollout, and the people who run the process are the ones best placed to help draw it.
What a business should do
Before you roll out an AI tool, do these things in order.
- List the people who run the process today and talk to them first. Ask where it breaks, what they work around and what they would never hand over.
- Let them see the tool early and test it on real cases. Change the setup based on what they find.
- Name one person who decides for each step, and write down who can override the AI.
- Put hard limits in the system for anything where a wrong call is costly, such as money, promises and dates. Send those cases to a person with the full context.
- Check the results after a few weeks with the same staff. Fix the brief, then run it again.
Praise from the customers is nice to have. The sign that the rollout is working is that the people doing the job use the tool and say it helps.
Source: Ars Technica: FAA expands AI advice to manage air traffic congestion across entire US airspace