At Pearl AI, the next test is not whether software can spot a finding on a radiograph. It is whether that finding improves case acceptance, cuts billing errors and saves staff time.
In September 2026, Pearl launched Second Opinion 3D, a commercial 3D-image analysis product that made the company's radiologic AI available to general dentists for the first time.
Diagnosis was only the starting point
Radiologic AI has advanced quickly, Tanz says. Systems can detect more pathologies on one image. Accuracy has improved, and the software now connects with more imaging platforms.
Pearl AI has embedded radiologic AI into roughly 40 imaging platforms. The next shift is broader integration with practice management systems.
Pearl says it has developed FDA-cleared computer vision for 2D and 3D dental images since 2019. Its products are used by clinicians, practice owners, laboratories and insurers, according to its company description.
The reason is workflow. A dentist spots a finding on a radiograph. The dentist explains it to the patient, records it in the chart and uses it to shape a care plan and recall schedule. Each handoff can reduce the precision of the original information.
When a finding is established on the image, the same information can move into the case presentation, clinical note, care plan and long-term patient record.
That is the gap.
The American Association of Orthodontists' September 2026 position paper says a licensed orthodontist remains responsible for any clinical use of AI. It also calls for disclosure of training data, validation methods and known limitations, plus notification when algorithm or dataset changes could affect performance. The document is professional guidance rather than a regulation.
Adoption is lagging behind capability
Pearl AI's Oral Health Report found that two in ten UK oral healthcare practitioners used AI in their practice day to day. Tanz says many of those uses are narrower than the available technology.
Some sceptical dentists change their view after two weeks with the software. They may spot something they would have missed. They may also see a patient understand treatment more clearly.
Office managers can have a similar experience. Eligibility checks or claim evidence can stay inside the workflow instead of passing through repeated manual tasks.
Small changes can add up.
Pearl's portfolio now goes beyond its radiologic Second Opinion product. It includes Practice Intelligence, tools that automate insurance verification and systems that analyse data from practice-management software, according to an overview of Pearl's products.
That does not make AI infallible. Tanz calls the technology a "consistency layer, not an oracle". Practitioners who expect perfect detection may abandon it after early disagreements. Those who use it as a second set of eyes may see its practical value sooner.
The next frontier is the conversation and the balance sheet
Tanz sees substantial room for development in three areas. Ambient AI could capture and structure the clinical conversation. That could include examination findings, periodontal observations and treatment discussions that might otherwise be recorded late or rebuilt from memory.
Revenue-cycle work is another target. It covers billing, collections, eligibility and claims. In the UK, NHS reimbursement has its own documentation and claim requirements. Private coverage adds another layer of administrative work.
CBCT is also a major frontier for Tanz. Two-dimensional radiograph analysis is relatively mature. Three-dimensional interpretation is still developing.
General dentists may find CBCT harder to interpret consistently. Tanz also says it could connect dental AI with oral-systemic health. Pearl's September launch of Second Opinion 3D, reported by Becker's Dental, expanded that capability commercially to general dentistry.
The commercial test is concrete. Practices can track case acceptance, denial rates, days in accounts receivable, write-offs and staff time spent on eligibility and claims.
Tanz is not saying every AI tool will deliver those gains. His point is narrower. A practice does not need an ideology to test whether a workflow improves its results.
Numbers tell the story.
For clinicians and practice owners, the issue is whether a particular system removes repeated work without weakening clinical judgement, documentation or patient communication. Dentistry's software infrastructure makes that difficult. Imaging systems and practice management platforms use inconsistent data standards. Teams must also change habits built over time.
Practices must check the FDA status and intended use of each product. A company-wide marketing claim is not enough. The relevant module may have a specific authorisation for its clinical purpose. The practice must also confirm that it works with the practice-management system already in use.
Tanz's account points to a clear direction. Dental AI is moving from a separate diagnostic application toward a thread that links examination, explanation, documentation, treatment planning and follow-up.
Tools that fit existing workflows will have the strongest case. Their value should show up in clinical consistency and practice performance.
The serious test is simple. Does the system make the next patient, the next record and the next monthly result measurably better?
