If you spent any time in the hallways at ASAE Annual this year, you probably had this conversation at least three times. Not "should we be doing AI" — that argument's over. The conversation now is about competitiveness. Members expect more value tied to their actual job, faster answers, a better experience than what they're already getting from the free chatbot open in their other tab or on their second computer screen. And more than one exec told me, more or less unprompted, that next year is when this either turns into a real edge or a real problem.
And almost every time, somebody in the room eventually said some version of: "We've got a dev team. We're already looking at building this ourselves."
I get why. It feels like control instead of a vendor contract, like the responsible, grown-up choice. But it's usually the wrong bet, and it's worth saying so plainly before next year's budget gets locked in.
The model was never the moat
Building something that answers questions reasonably well isn't the hard part anymore. The models are good; they keep getting better on a schedule nobody controls, and anyone with an API key can plug into one. Your team can absolutely put together something that looks great in a board meeting. That was true a year ago too. But it won’t work.
The hard part starts after the demo ends. Somebody has to take years of fragmented content, education material, PDFs, webinar recordings, expert interviews, research, whatever's been sitting in that one folder nobody's touched since 2019, and turn it into something an AI can actually reason over without making things up. Somebody has to earn member trust back the first time it gets an answer wrong. Somebody has to keep tuning it against how people actually use it, not how you assumed they would. And somebody, ideally with real authority, has to own the thing internally so it doesn't become the project everyone loves in theory and nobody's job in practice. None of that shows up in a demo. All of it is where in-house builds tend to stall, usually about eighteen months in, right around the time the person who championed it takes a new job.
Nobody's writing their own payment processor, or standing up their own email deliverability system in-house, because those aren't problems worth reinventing. Somebody already solved them well, at scale, a hundred times over, and every organization that insists on solving them alone just ends up paying the "first time" tax everyone else paid years ago. AI isn't different. The model was never your differentiator. What you do with it is.
And while your team spends the next year rebuilding things that are already built, your real competition isn't waiting around. It doesn't need to out-build you. A venture-backed startup, or the association down the street with a bigger AI budget, just needs to be better at one thing: faster answers, a smoother path through certification, a first ninety days that doesn't feel like homework. Members have already proven they'll go wherever the friction is lower. Loyalty to the logo isn't a given anymore. What gets compared now is the outcome.
What a model can't do, and two hundred reps can
Here's what tends to get lost in the excitement: AI can do a lot, but it can't replace a team that's already launched this exact thing two hundred times. Your developers are good at their jobs. They're also, necessarily, doing this for the first time at your organization, which means they're going to learn some of it the hard way. A team with two hundred launches behind them already knows which content trips up a model, which member segment needs a slower rollout, which VP will kill the whole thing if they're not brought in by week two, and roughly how long it takes to go from "cool demo" to "staff actually trust this enough to use it without double-checking."
That kind of knowledge doesn't come from a better model. It comes from having done this enough times to know where the bodies are buried, and it's the one thing your in-house team can't buy, can't hire for on short notice, and definitely can't shortcut with a clever prompt.
What the smart ones are doing before next year
The associations that separate themselves in the next twelve months probably won't be the ones with the most technical staff. They'll be the ones treating this like a portfolio decision instead of a pride project. A few patterns kept coming up in post-Annual conversations.
Somebody's finally been named the actual owner of AI internally, instead of leaving it to whichever department happened to run the pilot, because "IT will figure it out" is how good pilots quietly die on the vine. The smart ones are watching what members actually click on and ask about, not just what they say on a survey, because behavior tells you things people won't. They're taking an honest look at how much of their best content is still locked behind a login nobody bothers with anymore, while that same answer is one search away on a public AI tool. And they're being straight with themselves about which parts of this are genuinely core to the mission and worth owning, and which parts, like implementation, are better left to someone who's already run this play a hundred times.
The right question isn't "can we build it"
Almost any association with a decent dev team can build something that resembles an AI assistant. That was never really the question. The real one, the one worth putting on the agenda for your next leadership meeting, is who's going to make sure it still works, still gets used, and still earns its line in the budget on day one hundred, not just day one.
That's not really a technology question. It's a change management question, and it's the one nobody puts a line item on until it's already the reason the whole thing stalled out. The associations leaving this year's Annual with actual momentum aren't the ones with the most ambitious roadmap. They're the ones who figured out early, and said out loud, that their own team's time is better spent somewhere other than reinventing a wheel two hundred other people have already built.
