You Want Your Nights and Weekends Back.
How to choose an AI partner who serves your team, not the technology

I recently sat with a controller at an air charter and maintenance company who closed the books across multiple entities by hand every month. Each cycle swallowed nearly two weeks as she moved numbers between systems with no interest in speaking to one another and chased receipts through email one at a time. 

She looked at me and said, “I don’t want AI. I want my nights and weekends back.”

That is the whole thing, really. Nobody walks into your company wanting “artificial intelligence.” But you want the quote to go out before the customer calls somebody else. You want the work order typed once. You want to stop rebuilding the same report every month for a person who may or may not read it. 

NATA, Tuvoli and ERAU just hosted the first AI Symposium for Business Aviation. If you were there, I suspect you left with more questions than you arrived with—which is usually a good sign. Now that the booths are packed up, let me put a few honest things in writing, from someone who is admittedly in the business of selling AI.

You know something no vendor does 

Here is what makes this decision genuinely hard: nearly everyone who pitched you is capable, and nearly every demo you saw was real. But it was also running on their data. Clean, labeled, sitting in one place, tidied up by somebody who spent a week preparing it for exactly that purpose.

Yours lives in a scheduling system, a maintenance tracker, a fuel system, an accounting package, and roughly four inboxes. Two of them disagree about how to spell your best customer’s name. One of them is a whiteboard.

So the question you need to answer is not, “What can this do?” It is, “What can this do here, in my building, with my people and my systems?” 

Nobody on an exhibit floor can answer that for you—which sounds like bad news, but actually isn’t. The hard part of this decision has never been technical. It is about how your business runs, and on that subject you are the most qualified person in the conversation.

You came looking for someone with answers. What you needed was better questions, and you already have everything required to ask them. 

Three kinds of company will pitch you, and all three serve a purpose 

Over twenty-five years in the software industry, I have come to realize most vendors fall into three buckets. Each is genuinely excellent at one thing and quieter about the other two. 

The niche provider knows your business cold. They have been at every conference for a decade, they speak business aviation fluently, and they are often already in your stack. Their AI is usually a feature added to something you already own, which makes it fast and cheap to switch on. For a narrow problem, it is frequently the right buy. 

The drawback: you outgrow them quickly. Their AI works on the data inside their product and nowhere else, so the second and third things you want become a roadmap date, and after a few years you are running four narrow tools that nobody has ever connected.

The development shop is prepared to build with real engineers, real code, real schedules, and a refreshing willingness to give you a date. If you already know precisely what you need, this is a genuinely good use of money.

The drawback: you become the product manager, whether or not that was the plan. They will build exactly what you specified—including the parts you specified wrong—because nobody in that room has the standing or the context to tell you otherwise.

The consultant will ask you the sharpest questions in the process. Six weeks in you will have a roadmap, a maturity model, and a much clearer view of your own operation than when you started.

The drawback: you pay for the thinking twice. Once to them, and again to whoever eventually builds it—by which point the roadmap is nine months old and your business has evolved. The deliverable is a document, and documents do not reduce anyone’s workload. 

What you want is a company that understands how you make money, can build the thing, and will tell you when the thing is a bad idea. Two out of three is the common shape in this market, and it is perfectly workable if you know which one is missing and who is covering it. 

The part that will save you the most money 

Here is the advice that costs me business, so weigh it accordingly. 

With most technology, you buy the capability. With AI, a surprising amount of it arrives with your own people, trained, using tools you may already pay for. Turning a forty-message thread into something a busy person will read. Asking questions of your own contracts and manuals. Doing the research before you commit somebody’s entire week to it. None of that needs an integration, a project plan, or a purchase order.

Where you genuinely need help is the other half: connecting systems, anything that acts rather than drafts, customer data at scale, and designing who reviews what when something comes back wrong.

There is a third category nobody puts on a slide, and it deserves saying out loud. Plenty of work you could do yourselves, you won’t, because nobody has the time to own it. That is a legitimate reason to hire someone. It is a far better reason than believing the work is beyond you. 

So before you buy anything, train a few of your people on ordinary, low-stakes work. Not customer-facing. Not regulated. Expense coding, meeting notes, first drafts. What you are buying in that first month is not output. It is judgment. Your people learn where the tool is strong and where it is confidently wrong. And they learn it in an area where being wrong is cheap. The person who can glance at an output and know in three seconds that it is nonsense is the most valuable thing in this whole business, and none of us can sell you one.

Three good questions for your AI partner 

Each one is aimed at a different weakness, and none of them is a trap. Good companies enjoy all three. 

“Before I tell you anything, what do you think is costing a business like mine the most hours?” 

The domain test. Someone who knows your world will name something specific and slightly unglamorous, which you will recognize immediately. Someone who doesn’t will ask you for requirements or describe their product. Both are honest answers to a question they cannot answer, and now you know.

“What’s the last thing you put into production, and what went wrong with it?” 

The engineering test. The second half is the important half. Everything that delivers goes wrong somewhere. A company that has actually executed will tell you about the integration that ran three weeks long—and sound almost weary describing it. A company that has only advised will change the subject to methodology.

“What will my team be able to do without you in six months?”

The independence test. This is the one that tells you whether you are buying a capability or a dependency. The right answer involves handing things over and teaching your people. If the question itself seems strange to them, that is your answer.

Ask all three at my table too. If we can’t answer them, we don’t deserve the work.

That controller, incidentally, got her weekends back. Not from a transformation. From an expense app that reads a photo of a receipt. The integration took three weeks longer than anyone promised, which is true of every project I have ever shipped, and which nobody puts in the proposal.

Start small. Start with your people. Then keep going, because technology was never the point.

The point is to make work more meaningful.