Making AI earn its place in your dental practice

A couple of interesting questions from clients yesterday:

1. Please can you help produce a  priority list for the areas where you see AI should be introduced into dental practices.

2. The pace of AI innovation is frightening.  Should we have a dedicated AI budget or % of budget overall? Or perhaps we should look at each area of expenditure and dedicate varying% to AI expenditure.

3. Who should be responsible for AI introduction within any practice hierarchy  or do we say AI affects every area  so every strata has to look at it but have one person who outlooks it all?

followed by:

Use of AI to create efficient systems. Not sure what AI tool is best - Claude vs Chat GPT, is pro worth it. How can I use them to help reception, practice manager, marketing, etc?

My view is that AI should earn its place in a dental practice by saving time, improving consistency and helping people make better decisions.

I use it extensively for analysis, writing and developing systems. I also know how easily an interesting experiment can swallow an afternoon. The discipline is to finish one useful improvement, embed it and then move on.

My suggested order of priority is:

1. Practice management and systems. Turn the way you actually work into clear procedures, checklists, induction materials and training guides. Give the practice manager help with the repetitive writing and organising that consumes management time.

2. Reception and patient communication. Develop approved enquiry responses, telephone scripts, follow-up templates and practice scenarios for training. The purpose is to release people for hospitality, reassurance and conversations that build trust.

3. Financial understanding. Use reconciled accounts and anonymised performance data to explain trends, test fee scenarios and prepare questions for your accountant. Check every important calculation against the source.

4. Marketing. Build useful answers to the questions patients genuinely ask. Turn your expertise into FAQs, newsletters, blogs and video scripts, with a named person checking accuracy and tone before publication.

5. More complex automation. Introduce approved call-handling, diary and clinical-support systems only after testing their suitability, safeguards and integration. Clinical decisions remain with qualified clinicians.

On budget, I would start with a capped 90-day development budget covering subscriptions, training and protected implementation time. Allocate it to two or three specific problems, each with an expected benefit.

I would avoid setting an arbitrary percentage of turnover. Once a system proves useful, put its ongoing cost into the relevant departmental budget while keeping one central register of AI expenditure. Measure hours genuinely released, errors reduced and improvements in patient experience or conversion. Time saved becomes valuable when we decide how to use it.

Responsibility should be shared, with one person accountable for coordination. The owner sponsors the programme and approves expenditure. The practice manager, or a capable operations lead, oversees implementation, supported by a champion in each department. Give that lead protected time, authority and a monthly review meeting.

On Claude versus ChatGPT, my everyday starting point is ChatGPT. I would test both on the same three real tasks and compare accuracy, usefulness and the editing required. Then standardise the team's approach. I have many clients who are very happy with Claude and feel it's a matter of personal style - similar to Microsoft v. Apple.

A paid subscription is worthwhile when the work justifies it. ChatGPT Pro makes sense for intensive individual use requiring its additional capabilities or allowance. For a practice team, assess a business or team plan with appropriate administration and data controls. The word “Pro” means different things across providers.

Start with anonymised or fictional examples. Identifiable patient information requires an approved workflow and data-protection assessment; buying a business subscription does not, by itself, make a workflow compliant. Keep human review of patient-facing output.

My recommendation is to choose one problem this week, appoint its owner, agree the measure of success and review the result after 30 days. Build confidence through useful, repeatable progress before adding another new tool.