September 2026
Vertical SaaS Without Support Makes No Sense
AI is making software more efficient than it's ever been. And at the same time, it's making support more important than it's ever been. Those two things sound like they should cancel each other out. They don't. We're running 90-day AI challenges across every department at Omniboost right now. The gains are real: faster answers, fewer manual steps, less time lost to reconciliation. I'm all in on it. AI is almost always right. But "almost always right" is doing a lot of work in that sentence.
AI is cold. That's the problem.
AI makes cold decisions. Not wrong ones, necessarily. Cold ones. It optimizes for the pattern it sees, not for the guest standing at the front desk with a problem the pattern doesn't cover. That's not a flaw you fix with a better model. It's a gap you fill with a person. We see this constantly. AI can flag that a transaction doesn’t reconcile, that tax logic changed, or that a rate got overwritten. It can’t tell you why the property changed their tax setup last month, or why the front desk manager overrode that rate in a way that made sense in the moment but breaks the pattern everywhere else. That context lives with our support team, not in the model. It's why partners stay with Omniboost after the integration is "done." The integration handles the 95%. Our people handle the part where something doesn't look right, and someone actually needs to understand why, not just flag it and move on.
Confident, not arrogant
One of our core values at Omniboost is Confident, not Arrogant. I think about that value differently since AI entered every client conversation we have. Confidence says: I trust this system, and I still check it. Arrogance says: the system is right, so the outcome doesn't need a human look. AI, left to run on its own, behaves like the arrogant version. It doesn't hedge. It doesn't sense when something feels off. It just answers. Human governance is where confidence actually lives. Someone has to be willing to say "this doesn't look right" even when the system is 99% sure it is. That's not bureaucracy. That's the difference between a tool you can trust and a tool you're hoping works out.
The second violin still matters
I wrote about the second violin before: the idea that its value comes from being the foundational layer that makes everyone else perform better. Support is exactly that. It's the second violin that shows up when the guest, or the client, needs someone to actually care. Willingness to serve is an emotional decision, not a logical one. It's the choice to care about someone else's problem as if it were your own, at 11 pm, when there's no efficiency argument for doing so. AI can execute the fix. It can't decide to care about the fix.
The guest is still the point
Every hospitality vertical we serve exists because someone, somewhere, is trying to take care of a guest. That's the whole point of the industry. Efficiency should exist to protect that focus, not replace it. A lot of AI-first thinking in vertical SaaS gets this backward. The pitch is: automate the interaction so no human has to touch it. But in hospitality, the human touch isn't overhead. It's the product. You automate everything up to the moment that matters, and then you get out of the way for the part that doesn't scale: someone showing up for another person, on purpose. Vertical SaaS that strips out support because AI is good enough is optimizing for the 95% of cases and abandoning the 5% that decide whether someone trusts you at all. Trust doesn't get rebuilt by a faster response time. It is rebuilt by a person willing to serve.
The partnership dilemma
Here's where it gets tested for real. In some of our partnership models, we've agreed that first-line support sits with the partner, not with us. Our AI bots are instructed accordingly. If a ticket falls under that agreement, neither the bot nor our teams jump in.
On paper, that's fair. In practice, it's the tricky part. The end-user, the hotel, or the restaurant doesn't see a contract. They just see a problem, and the person who happens to know the answer isn't always the one who's allowed to pick it up.
But "not our job" and "not our problem" aren't the same thing. If the end-user walks away unserved, it doesn't matter whose line item it was supposed to be.
Our support and onboarding teams feel that pull anyway. They want to help. Incidentally, that's just good instinct at work and nobody minds. The problem is when it becomes structural. So the AI bot, in a narrow sense, is right: first-line support isn't our job there. But the customer comes first. That same instinct is a big part of what's driven our success. It's not a bug we manage around. It's the thing that got us here.
This is the part worth celebrating, not fixing. The fact that our team feels that pull despite the agreement, despite the cost, tells me partnership at Omniboost isn't just a word in a contract. It's a behavioral trait. Our partners and we exist for the same reason: to help and serve the end-user. So the answer isn't to enforce the boundary harder. It's to keep reiterating a model, together with our partners, that actually works for everyone.
Confident in the tech. Never arrogant about it. That's the balance.
