The AI sales pitch has become a weekly event in the dental technology space. At Centaur, we receive integration requests from AI dental receptionist companies, imaging analysis platforms, patient communication tools, clinical scribes, and scheduling assistants at a rate that has more than doubled in the past year. Every single one promises the same thing: save time, cut costs, transform your practice.”
And for every request that lands with us, we know it has already made its way to a few hundred dental practices across Australia.
So, I want to ask you something before we go any further. Have you bought an AI tool in the past twelve months? If you have, what problem were you solving when you bought it? Not the problem the salesperson identified for you. The problem you measured, quantified, and confirmed was costing your practice real money.
If you hesitated on that question, read on.
The Numbers Are Telling a Story Most Vendors Will Not
In 2025, the world spent $1.5 trillion on AI (Gartner, 2025). IBM’s global study of 2,000 CEOs across 33 countries found that only one in four of those investments delivered the expected returns (IBM, 2025). MIT’s research across more than 300 enterprise deployments found that 95% of generative AI pilot programs delivered zero measurable impact on profit and loss (MIT NANDA, 2025). Zero. S&P Global reported that 42% of companies scrapped most of their AI initiatives in 2025, up from just 17% the year before (S&P Global, 2025).
This is not a story about bad technology. The technology, in many cases, works perfectly well. It is a story about buying solutions before diagnosing problems.
The dental industry is not immune. Vendors have entered the market at pace. AI-powered patient communications, automated recalls, smart imaging tools, AI scribes, chatbots, and a flood of AI receptionists are all competing for a budget that most practices have not specifically set aside for AI at all. Orgvue’s April 2025 survey of 1,163 C-suite leaders, including Australian respondents, found that 64% of executives admitted that the fear of falling behind, not a clear business case, drove their AI spending decisions (Orgvue, 2025). That needs to stop.
The Case That Should Give Every Practice Owner Pause
In 2023, Klarna, a $14.6 billion fintech company with direct access to OpenAI’s best technology, replaced 700 customer service staff with an AI chatbot. By early 2024, the CEO was publicly celebrating the result. The chatbot was handling 2.3 million customer conversations per month.
By late 2024, customer satisfaction had dropped. Complaints were rising. By May 2025, CEO Sebastian Siemiatkowski publicly admitted that the company had focused too much on efficiency and cost, resulting in lower quality. Klarna began rehiring human agents (Klarna, 2025).
If Klarna, with OpenAI in its corner, could not replace human judgement by flipping a switch, a dental practice of five staff cannot either.
The lesson from Klarna is not that AI failed. It is that AI was applied without first asking the right question. That question is simple. What problem, exactly, are we solving?
Introducing SCOPE
Over the past year, I developed a structured approach to AI adoption decisions for dental practices. I call it SCOPE. It stands for Stability, Capacity, Objective, Persona, and Ethics. It is a pre-flight checklist, not a philosophy. Run through it before you sign any AI contract, and you will know whether you are ready to invest or whether you need to build your foundations first.
Stability Check
AI amplifies what you already do. If your processes are stable and well-documented, AI can make them even better. If they are inconsistent, AI makes the inconsistency faster and more expensive.
Before you consider any AI solution, ask whether your current workflows are documented and followed consistently by every team member, regardless of who is on shift. Can you quantify your top three operational pain points with actual data? Do you know your current no-show rate? Your average call conversion?
I call this the Document Test. If it is not written down and measured, AI cannot fix it.
Capacity Evaluation
If everyone is the AI champion, then no one is.
Before committing to any implementation, apply what I call the Tuesday Test. Can you or a designated team member spare 2 hours every Tuesday for the next 3 months to own implementation, training, and troubleshooting? If the honest answer is no, you do not have the capacity to do this properly right now. Doing it badly is more costly than not doing it at all.
And budget beyond the subscription price. Factor in the training time, the workflow disruption during transition, staff overtime in the first few weeks, and the ongoing maintenance overhead. The advertised price is rarely the real price.
Objective Definition
This is where most purchasing decisions come apart. A vendor pitches their solution. You get excited about the capabilities. You sign up. Then, three months later, nobody can clearly say whether it is working.
For every AI solution you consider, write this statement before you engage a vendor.
[Problem] causes [impact], costing our practice approximately [dollar amount or hours] per [week/month].
If you cannot fill in every field in that sentence, park the conversation. I call this the So What Test. What does this problem actually cost you in dollars, time, or patient experience? If you cannot quantify it, it is not ready for an AI solution. And ask yourself whether you have tried non-AI options first. Sometimes a better checklist, a process change, or a fifty-dollar-a-month scheduling tool solves 80% of the problem without a single integration.
Practice Persona
Your patients chose your practice for a reason. Maybe it is the warmth of your front desk. Maybe it is the way your team remembers that a patient’s daughter just started school. When you hand patient communications to an AI system, something needs to carry that personality into every output.
Before you deploy any AI tool that faces your patients directly, run what I call the Read Aloud Test. Take a piece of AI-generated patient communication and read it out loud. Does it sound like something your reception team would actually say? If it sounds generic, your patients will feel the difference, even if they cannot name it.
The Australian Office of the Information Commissioner has made its position clear. AI-generated or inferred information about individuals is personal information under the Australian Privacy Principles (OAIC, 2025). This is not just a brand risk. It is a compliance obligation.
Ethics and Evaluation
Four phrases in a vendor conversation should stop you immediately. “Proprietary algorithm” means you cannot audit how decisions are being made about your patients. “Limited time offer” is a pressure tactic that signals the value proposition is weak. “No trial period” means the vendor knows what happens when customers get to test it properly. And “we cannot export your data if you leave” means you are a hostage, not a customer.
The Australian regulatory context matters here. The Privacy Act 1988 and the Australian Privacy Principles apply to all AI use involving personal information. From June 2025, a new statutory tort will allow individuals to sue directly for serious invasions of privacy, with penalties of up to $50 million for organisations (Privacy and Other Legislation Amendment Act 2024). The Therapeutic Goods Administration has flagged that AI scribes proposing diagnoses beyond those identified by a clinician may constitute unregistered medical devices (TGA, 2025). The Dental Board of Australia, the TGA, the Privacy Act, and State and Territory Civil Liability Acts all intersect in this space (Australian Dental Journal, 2025).
When you evaluate a vendor, confirm where patient data is stored, who can access it, what jurisdiction governs it, and what happens to your data if you cancel. Any vendor that cannot answer all four questions clearly should not handle your patient records.
Where AI Actually Earns Its Keep
The practices getting real value from AI share a common pattern. The wins are specific, measurable, and concentrated in high-volume repetitive tasks.
Radiographic analysis is the most clinically credible application. Multiple systematic reviews confirm that AI caries detection achieves sensitivity (the ability to correctly identify decay when it is present) and specificity (the ability to correctly rule out decay when it is absent) comparable to or better than those of unassisted clinicians. An umbrella review covering 14 systematic reviews and 137 primary studies found a pooled sensitivity of 0.85 and a specificity of 0.90 (Dashti et al., 2024). A 2025 study by researchers at the University of Western Australia found AI demonstrated superior sensitivity and equal specificity compared to bitewing radiography for caries detection (Abbott, Saikia and Anthonappa, 2025).
Smart appointment management has a growing evidence base. A peer-reviewed study tracking over 135,000 appointments found a 50% reduction in no-show rates after AI implementation, with patients 57% less likely to miss appointments (Al-Shamsi et al., JMIR Formative Research). The American Dental Association reported an average 25% reduction in no-show rates across clinics implementing AI-driven predictive analytics and automated patient engagement tools (ADA, 2022). Results vary by practice size, patient mix, and implementation quality, but the impact on chair utilisation is direct and quantifiable.
Then there is the admin grind. HICAPS reconciliations, CDBS processing, DVA paperwork, recall management, referral letters. High-volume, rule-based, and repetitive. This is where AI genuinely earns its place in an Australian practice. Not glamorous, but it is where your team’s time is being consumed week after week.
It is also worth acknowledging that some AI dental receptionist products are genuinely impressive. Having evaluated a number of them, there are solutions in the Australian dental market that handle patient communications with a sophistication and natural fluency that would surprise most practice owners. The issue is not the technology. It is whether your practice has the call volume, the after-hours demand, and the defined problem to justify it. The right product applied to the right problem is a completely different conversation to buying one because everyone else seems to be.
Notice the pattern. Each win is specific. Each has a measurable baseline. None of them began with “let’s try AI and see what happens.”
What to Do This Week, This Month, and This Quarter
This week, list every AI tool and subscription your practice currently pays for. Circle the ones your team uses daily. The gap between what you pay for and what you actually use is your first saving, and it is waiting for you right now.
This month, pick your single biggest operational pain point. Not the one a vendor identified. Yours. Plot it against two questions: how often does this problem occur, and how much does it cost when it does? The problems sitting in the high-frequency, high-impact corner of that grid are your legitimate AI candidates. Everything else can wait.
This quarter, if your SCOPE assessment says you are ready, pilot one solution. Not three. Not five. One. Set a 90-day measurement framework before you start. Days 1 to 30 are adoption: Is your team actually using it? Days 31 to 60 are for effectiveness: Is it moving the specific metric you identified? Days 61 to 90 is the moment of truth. Calculate total cost against measurable benefit. If the return is not clear and demonstrable by day 90, execute your exit strategy.
IBM’s research found that AI pilot programs showing early returns of around 31% settled to a more realistic 7% as they scaled, below the typical 10% cost-of-capital hurdle rate (IBM, 2025). The honeymoon ends. Your 90-day test gives you structured decision points so you are not stuck paying for something that looked great in the demo but could not survive contact with reality.
The Practices That Get This Right Are Not the Ones That Move First
They are the ones who move deliberately.
Being deliberate is not the same as being slow. It is not falling behind. It is building a foundation that AI can actually stand on. The practices that will get genuine, measurable value from AI over the next three years are the ones asking hard questions today. What is broken? What does it cost? What does success look like in 90 days?
Start there. Everything else follows.
References
Abbott, L.P., Saikia, A. and Anthonappa, R.P. (2025) ‘Artificial intelligence platforms in dental caries detection: a systematic review and meta-analysis’, Journal of Evidence-Based Dental Practice, 25(1), p. 102077.
Al-Shamsi, S. et al. (2021) ‘Reducing patient no-show rates using artificial intelligence: a retrospective cohort study’, JMIR Formative Research. Available at: https://www.ncbi.nlm.nih.gov [Accessed March 2026].
American Dental Association (2022) AI in Dental Practice: Predictive Analytics and Patient Engagement. Chicago: ADA.
Australian Dental Journal (2025) ‘Artificial intelligence in dentistry: a scoping review of regulatory and ethical considerations for Australian practitioners’, Australian Dental Journal, August.
Dashti, M. et al. (2024) ‘Examining the diagnostic accuracy of artificial intelligence for detecting dental caries across a range of imaging modalities: an umbrella review with meta-analysis’, PMC, October.
Gartner (2025) Gartner Forecasts Worldwide AI Spending to Reach $1.5 Trillion in 2025. Available at: https://www.gartner.com [Accessed March 2026].
IBM Institute for Business Value (2025) CEO Decision-Making in the Age of AI. Armonk: IBM. Available at: https://www.ibm.com/thought-leadership/institute-business-value [Accessed March 2026].
Klarna (2025) CEO Statement on AI and Customer Service Strategy, May 2025. Available at: https://www.klarna.com/newsroom [Accessed March 2026].
MIT NANDA Research Group (2025) Generative AI in the Enterprise: Pilot Outcomes and P&L Impact. Cambridge: Massachusetts Institute of Technology.
Office of the Australian Information Commissioner (2025) Guidance on Artificial Intelligence and the Privacy Act 1988. Canberra: OAIC. Available at: https://www.oaic.gov.au [Accessed March 2026].
Orgvue (2025) The AI Workforce Report: C-Suite Perspectives on AI and Employment. London: Orgvue. Available at: https://www.orgvue.com [Accessed March 2026].
Privacy and Other Legislation Amendment Act 2024 (Cth).
S&P Global (2025) Corporate AI Adoption Survey 2025. New York: S&P Global Market Intelligence.
Therapeutic Goods Administration (2025) Artificial Intelligence as a Medical Device: Regulatory Guidance. Canberra: TGA. Available at: https://www.tga.gov.au [Accessed March 2026].
About the Author
Sean Perera
Chief Technology Officer
