Everything published so far
Numbered in the order it was written. The series builds, so the early ones are the place to start.
- What AI Actually Does Inside a UK Dental Practice Today 1,716 words Separates the four genuinely deployed categories — radiograph analysis, triage, notes, front desk — from marketing fiction, and gives principals a map of which ones are worth attention first.
- How AI Reads a Dental Radiograph, Explained Without the Maths 1,945 words Argues that clinicians need a working mental model of pixel-level pattern detection — not deep learning theory — to judge when an AI output deserves trust.
- The Six Front Desk Jobs AI Can Realistically Take Over 1,859 words Breaks reception work into discrete tasks and shows that automation succeeds on bounded, repetitive jobs and fails on anything requiring judgement about a distressed patient.
- Is Your Dental AI a Medical Device? A Principal's Guide to UKCA 1,711 words Makes the case that diagnostic claims are the dividing line: if a tool tells you a lesion is present, it is regulated, and buying an unregistered one puts the clinician on the hook.
- What an AI Dental Scribe Actually Captures During an Exam 1,818 words Walks through a real check-up transcript to show what ambient capture gets right, what it invents, and why the dentist still writes the diagnosis line.
- Teledentistry in NHS Practice: What You Are Allowed to Do 1,513 words Clarifies that remote assessment is permitted and claimable in narrower circumstances than vendors imply, and maps the boundary before a practice builds a workflow on it.
- Sensitivity, Specificity and Why Your AI Flags Caries That Isn't There 1,771 words Argues that a high-sensitivity detector tuned for research benchmarks will over-call in a low-prevalence NHS recall population, and shows how to read the threshold setting.
- What Dental AI Software Really Costs Per Surgery Per Year 1,801 words Builds a full cost model including per-image fees, training time and integration work, and shows why headline per-month pricing understates the first-year figure by a wide margin.
- GDC Standards and AI-Assisted Diagnosis: Where Responsibility Sits 1,508 words Establishes that the registrant carries the clinical decision regardless of what the software said, and what that means for how you document disagreement with a tool.
- Cutting Failed Appointments With Automated Reminders 1,833 words Uses reminder timing and channel data to argue that message cadence, not message wording, drives most of the FTA reduction practices report.
- Designing a Triage Form That Sorts Real Dental Emergencies 1,519 words Shows that the value of digital triage comes from a small number of red-flag questions, and that longer forms reduce completion without improving sorting.
- Who Signs an AI-Generated Clinical Note? 2,083 words Argues that the sign-off step must be an active clinical read, not a click, and proposes a practical check that takes under a minute per patient.
- Bitewings, Periapicals and Panoramics: Where AI Adds Most 1,918 words Contends that detection performance varies sharply by view type and that practices should switch AI on for bitewings first rather than across the whole radiograph library.
- An Eight-Week Pilot Protocol for Any Dental AI Tool 1,743 words Provides a structured pilot with baseline metrics and a pre-agreed kill criterion, arguing that pilots without a defined failure condition always convert to purchases.
- UK GDPR and Patient Photographs Sent to an AI Service 1,986 words Works through lawful basis and special category conditions to show that consent is usually the wrong basis for clinical images, and what to rely on instead.
- Consent and Channel Choice for SMS, Email and WhatsApp Recalls 1,765 words Shows that messaging preference records, not marketing consent, govern recall contact, and that most practices hold the wrong permissions on file.
- AI Phone Answering Versus a Human Answering Service 1,574 words Compares cost, call containment and escalation quality, and argues that AI wins on volume overflow while outsourced humans still win on complaints and pain calls.
- How to Read a Dental AI Vendor's Clinical Evidence Pack 1,969 words Teaches buyers to check dataset origin, reader study design and whether the comparison group resembles a UK general practice — the three places evidence packs are weakest.
- When the AI and the Dentist Disagree: Building a House Rule 1,763 words Proposes a written practice policy for resolving disagreement, arguing that ad hoc resolution produces inconsistent charting across associates in the same practice.
- Out-of-Hours and NHS 111 Referrals: Where Remote Triage Fits 1,980 words Argues that remote triage earns its keep by filtering the Monday morning backlog rather than by replacing the urgent care rota.
- AI Notes and FP17 Accuracy: Avoiding a Claims Problem 1,561 words Makes the case that generated notes must be checked against the banded claim, because narrative richness can contradict the treatment actually claimed.
- AI Measurement of Periodontal Bone Loss: Useful or Noise? 1,599 words Argues that automated bone level measurement is most valuable for longitudinal comparison and staging consistency, not for single-visit diagnosis.
- The Data Processing Agreement Clauses That Matter in Dental AI 1,804 words Identifies the handful of clauses — sub-processors, model training rights, data location, deletion on exit — that determine whether a DPA actually protects patients.
- Structured Templates or Ambient Notes? Match the Tool to the Appointment 1,661 words Argues that exams and consultations suit ambient capture while restorative and surgical visits are better served by structured templates, and that mixing them badly is why adoption stalls.
- Showing AI Radiograph Overlays to Patients Without Overselling 1,683 words Contends that overlays raise case acceptance but create a consent risk if presented as certainty, and offers language that keeps the clinician's judgement visible.
- Open APIs in UK Dental Software: Who Actually Lets You Connect 2,002 words Argues that integration depth, not feature lists, decides whether an AI tool gets used daily, and audits what the main UK systems currently expose.
- Getting Usable Patient Photos for Remote Assessment 2,002 words Shows that most remote assessments fail on image quality rather than clinical difficulty, and that simple capture guidance fixes the majority of retakes.