Artificial Intelligence
Ask AI vendors about your data before you ask about features
Trehan Stenton tells multi-site dental groups to judge AI vendors on data access, real intelligence and an 18-month roadmap, then prove the choice in a pilot.
The Advantage of AI, Why multi-site DSOs That Wait Will Be Left Behind
Video from The DSO Hygienist.
Generated from the canonical interview transcript and validated against source data by Guests on Air.
How should a multi-site dental group choose an AI vendor?
Start with the data, not the demo. Trehan Stenton asks every vendor how data can be pulled out of its system, at minimum as a CSV or Excel export and ideally through an API. He then asks what intelligence the tool applies, checks its roadmap against where the business will be in 18 months, and tests the winner in a small pilot.
Every software vendor now says its product has AI, and for a dental group running five to 40 locations, keeping track of what is on the market takes time few operators have. On The DSO Hygienist podcast, Trehan Stenton, who grew a dental group in Queensland, Australia to $4 million in annual revenue, explained how he cuts through that noise.
His starting point is not the feature list. Attaching AI to a product is easy from a marketing point of view, he said, but the real test comes when you drill into what AI means for the organization and how it is actually used. In almost every case that conversation comes back to the same place: the data.
What follows are the questions Trehan Stenton puts to vendors, why data access sits at the top of his list, and how he turns a promising tool into a pilot rather than a group-wide rollout. The advice applies to any growing multi-location business getting ready for AI, but the examples come straight from dental operations.
Key takeaways
- Ask every vendor how data leaves its system: a CSV or Excel export at minimum, an API or open connection ideally.
- A tool that only offers screens and screenshots leaves a multi-location group unable to see what is happening across its sites.
- Ask what intelligence the tool applies to its findings, such as missed cases, verified insurance and compliant notes.
- Plan 18 months ahead, compare several vendors and their roadmaps, and prove the choice in a small pilot first.
1. Make data access the first question
When Trehan Stenton talks to vendors, the first thing he wants to know is how the data can be pulled out of their systems. If a tool only offers a visual dashboard and the team ends up taking screenshots, a group with many locations has no practical way to see what is happening across the business. The data has to be freely available.
His minimum is a CSV or Excel export. The ideal is an API or an open connection to the data, so it can be brought together on one platform with everything else the group runs. A system that locks data down, or that cannot work with other systems, leaves the group stuck just as AI starts producing far more information than before.
He also warns about what goes into AI. Dropping an Excel spreadsheet into an uncontrolled environment is a poor idea, because the tool will not know the full context of the data and may amplify errors already in it. Rubbish in equals rubbish out, so someone needs to check the data with the team before anyone relies on the answers.
2. Ask what the intelligence actually does
Once the data question is settled, Trehan Stenton asks what intelligence is being applied to the findings. With radiography AI, for example, he wants to know which cases the practice has missed, which past cases could be brought through, how much money is on the table and what could realistically be converted. Useful AI surfaces that kind of answer without being asked.
Insurance offers simpler tests. Has the insurance been verified before the patient arrives at the office? Is a form created as part of the treatment notes compliant with the requirements of that particular insurer? Insurers are tightening their rules and may want specific words in the notes, so a vendor offering generic notes needs to explain how its AI handles that before a claim goes in.
These questions also show the wider effect of a tool. A vendor tends to look only at the deployment of its own product and the benefit it promises. Trehan Stenton looks upstream and downstream instead: if AI takes calls or notes off a team member, where does the freed time go, and does it lift patient conversions, speed up financing and give patients a better experience?
3. Buy for the business you will be in 18 months
Trehan Stenton does not choose tools only for the here and now. He thinks about 18 months ahead, asks what the business will look like then, and works out what he should be asking vendors today to get there. Data stays at the centre of that conversation, because without access to it the extra information AI produces quickly becomes impossible to use.
His advice is to talk to a range of vendors, because each has different capabilities and experience. Ask what they understand AI to mean, what their roadmap looks like and how far away that development really is. The aim is a partner the group can trust and rely on, with credibility and visible investment in technology that will still help the business later.
Even then, he would not roll a new tool out to everyone at once. He starts with a pilot, lets the team test the options and gives them ownership of the choice, so they buy into the solution and help sell it to their colleagues. If it does not work, the group drops it, keeps the lessons and moves on to the next opportunity.
