If an AI tool cannot connect to your practice management system, it is worthless to your practice. That is my honest position after years of these conversations, and a new piece in MIT Technology Review, the technology magazine published by MIT, makes the same argument for the whole healthcare industry.
The article, Healthcare AI's next test is integration, argues that the next decade of healthcare AI will be defined by integration, and that model capability alone will not get there. Worth noting: the piece was produced by Ensemble, a healthcare revenue cycle company, so it is making a case for its own approach. The case still holds up.
The conversation I keep having
The number of dental and healthcare practice owners who call me excited about a new AI tool is high. They saw an ad for it, or a post on social media, and they want it in their practice next week.
Then we hit the wall. Healthcare practices run on specific practice management software, the system that holds your patient records, schedule, and billing. In dental there is a handful of them. They are HIPAA compliant, generally old, hard to use, and hard to connect to anything. The shiny AI tool sits on the outside looking in.
The article says the quiet part out loud
The piece makes a point most AI marketing skips. Healthcare's administrative problems come from fragmented information, fragmented workflows, and fragmented accountability, and no shortage of information. The industry spent decades buying systems that each capture one slice of activity: electronic health records, billing platforms, scheduling systems, and more. Few of those systems were built to reason across the whole chain of decisions.
The newest AI models are genuinely impressive. The article notes they can process long clinical records, interpret complex terminology, and generate clean summaries from piles of information. None of that helps you if the model cannot see your patient database. The article's closing point matches mine: healthcare intelligence cannot live in a separate interface.
My 5 percent rule
Here is my rough estimate from the field. Of all the AI tools being pitched to healthcare practices right now, maybe 5 percent can actually integrate with your practice management system. That 5 percent is where your attention belongs. The other 95 percent will give you a second screen, a second login, and a second copy of your data that drifts out of sync within a month.
So my filter for any owner evaluating a tool is one question: does it write into the system my front desk already lives in? If the vendor cannot answer that clearly, the demo does not matter.
What this looks like with your ad leads
We run into this constantly on the ads side. As a Google and Meta Partner for over 10 years, we have set up campaigns for a lot of healthcare practices, and every one of those campaigns produces leads that have to land somewhere useful.
Our job is to find the simplest possible path from ad click to your practice management system. Sometimes that path includes a manual step, and I am fine with that, because most AI platforms are not building custom integrations for dental practice management systems. A reliable manual step beats an automated tool that dumps leads into a dashboard nobody checks.
I expect the tools will catch up eventually, and the vendors that build real integrations for the old, stubborn systems will win the healthcare market. Until then, judge every AI pitch by one thing: whether it can reach your patient data. Everything else is a demo.
