The matrix, in plain terms: For mixed-model dental groups, safe automation depends on six axes resolving together—site × contract/service type × patient status × clinician × appointment type × policy—before an AI receptionist may book. If any axis is missing or contradictory, the default is do not book: clarify, create a task, or escalate to staff. That six-axis combination is what this article means by the mixed NHS/private matrix at enterprise / DSO scale.
Direct answer: Portfolio automation fails when groups copy one site’s script across brands with different contracts, diaries and PMS systems. Success is governed configuration: global safety and change control, regional/site overrides, clinician and appointment mappings, and honest limits on write-back. Instant network-wide scaling and universal integration are not claimed.
Audience: group ops, clinical directors, regional managers and IT. Differentiation:
| Page | Owns |
|---|---|
| Practice-level NHS/private | Single-practice pathways and flows |
| DSO / dental groups | Group reception models, pilots, portfolio ops |
| Multi-site healthcare | Cross-sector multi-site patterns |
| This page | Enterprise mixed-contract matrix + governance |
| Guardrails | Deterministic booking rule design |
| PMS integration | Write-back levels and connectors |
Non-endorsement. Not legal, clinical or NHS contract advice. Nations and ICBs differ. Clero is not an NHS endorsement.
Configuration model (global → clinician)
| Level | Examples of what belongs here | Who typically owns |
|---|---|---|
| Global (group) | AI disclosure; urgent-language stop; never-do list (no eligibility decisions); retention principles; escalation taxonomy; change-control policy | Clinical governance + group ops |
| Regional / brand | Shared OOH scripts; regional urgent contacts; brand tone; pooled hub reception rules | Regional manager |
| Site | Hours, holidays, bookable types, deposits, local escalation mobiles, clinic slug / PMS clinic id | Practice manager |
| Clinician / resource | Which clinicians appear for which types; durations; chair/operatory where the PMS uses them | Site + clinical lead |
Inheritance rule of thumb: stricter / more specific wins for safety stops; site truth wins for hours and diaries; global wins for never-do clinical/eligibility claims. Document conflicts in a precedence table (guardrails).
Patient status (NHS / private / unclear / plan member) is an input to the matrix—not a model guess. Prefer staff-maintained PMS fields or explicit clarifying questions with escalate-on-ambiguity. Do not claim flawless automatic “tag sync = contract compliance without oversight.”
Worked inheritance sketch (illustrative)
Global policy: “AI never asserts NHS eligibility.”
Regional brand: “OOH uses hub mobile X.”
Site Central Mixed: NHS exam type NHS-EXAM-01 bookable 09:00–12:00 only; private exam PRIV-EXAM-01 all open hours.
Clinician: Dr A only on NHS-EXAM-01; Drs A and B on private exam.
If a caller is unclear on status at Central Mixed, the matrix yields escalate—not a keyword guess into NHS-EXAM-01. If Harbour Private has no NHS types mapped, NHS-framed requests become tasks or transfers, even if North Street would have booked them.
Governance: ownership, approvals, versioning, audit
Enterprise mixed automation is a change-controlled operating system:
- Ownership — RACI for global policy, regional overrides, site mappings, PMS connectors, telephony routing.
- Approvals — widening bookable NHS or private types requires named approvers (clinical + ops). Shrinking scope can be fast.
- Versioning — date, author, sites affected, types added/removed, connector versions.
- Audit — who changed what; sample calls showing pathway chosen; booking corrections linked to rule versions.
- Change windows — avoid Friday 5pm widenings before unsupervised weekends.
- Testing — regression pack per site class (NHS-heavy, private-heavy, mixed, new acquisition).
- Rollback — unmap type / disable AI book for a site within the same day.
- Exceptions — logged human overrides (goodwill, contract disputes) never silently taught as AI defaults.
Exception log fields (illustrative): site, date, call id, matrix axes at time of call, human decision, whether AI should learn a new rule (usually no until a formal version bump).
Regression pack (minimum): one successful NHS-allowed book (where applicable), one successful private book, one unclear-status escalate, one urgent-language stop, one failed write, one wrong-site clarify—for each live site class after a matrix change.
Acquisition and new-site onboarding
Hypothetical sequence (adapt to your PMO):
| Phase | Actions | Exit criteria |
|---|---|---|
| 0. Inventory | Numbers, hours, PMS, NHS/private mix, escalation mobiles, bookable types | Written site card |
| 1. Decide automation posture | Live write-back vs message/task-only | Connector confirmed or deferred |
| 2. Map matrix rows | Site × types × clinicians × policy | Signed rule pack v1 |
| 3. Wire telephony | Overflow/OOH route into AI | Test calls on production-like path |
| 4. Supervised pilot | Limited hours or intents | Correction/escalation rates acceptable to owners |
| 5. Wave expand | Next sites by similarity | No “flip entire brand overnight” |
Acquisition anti-patterns: copying Brand A’s NHS exam mapping onto Brand B because the type “sounds the same”; enabling AI book before the PMS clinic id and treatment routing exist; using a national marketing script that offers NHS appointments at private-only sites.
Do not promise day-one live booking on every acquired PMS. Mixed estates are normal after M&A (DSO page).
Mixed PMS estate considerations
| Reality | Implication for the matrix |
|---|---|
| Brand A on Dentally, Brand B on CareStack | Separate routing maps; no shared “one diary” fantasy |
| Site without verified connector | Message/task or human book until mapped |
| Same type name, different PMS codes | Per-site appointment-type ids in the matrix |
| Location hours in one admin UI ≠ book path hours | Single source of truth for AI writes |
| Dual numbers (NHS line vs private line) | Optional telephony split can reduce ambiguity—still map types explicitly |
Verified Clero dental booking handlers today: Dentally, CareStack, Semble, Aerona and Exact (via Exact Online Booking when enabled), plus calendar-style connectors where configured. Other brands are not assumed—confirm in procurement. Depth: PMS integration.
Cross-sector number architecture patterns (central vs local) remain on the multi-site healthcare page; this article only insists that pathway rules travel with the site, not with a single national bot personality.
Hypothetical multi-site scenario table
Hypothetical only—not customer results. Use the table to pressure-test whether your matrix distinguishes sites or collapses them into one script.
| Site | Mix | AI book (illustrative) | Escalate / staff approval |
|---|---|---|---|
| North Street | NHS-heavy | Hygiene + exam types mapped to NHS diary blocks only when status known | Eligibility disputes; UDA/capacity arguments |
| Harbour Private | Private-only | New exam + hygiene with deposit policy via desk/task if not software-gated | Finance plans; complaint goodwill |
| Central Mixed | Both | Private whitening/exam mapped; NHS exam only if type explicitly allowed | “Am I NHS here?” unclear → escalate |
| New Acquisition | Exact (Online Booking off) or other unsupported PMS | Message/task only until Online Booking/connector + matrix signed | All booking until go-live |
| Ortho brand | Private specialty | No general NHS types; specialist consult → task | Clinical suitability questions |
Safety row (all sites): urgent language → stop book → practice/group safety script (safe-by-design). Automation does not decide clinical urgency or NHS clinical necessity.
Safety boundaries and staff approval
Automation may (when configured): book explicitly allowed appointment types into the correct site diary; answer hours FAQs; create tasks; transfer per escalation policy.
Staff approval / human path required for: ambiguous NHS access; contract or charging disputes; clinical advice; safeguarding; complaints; goodwill fee waivers; any type not in the signed matrix; failed PMS writes; cross-site “send them elsewhere” unless explicitly configured as an offer (never silent load-balancing—see DSO).
Group safety standard: one urgent-language stop script family (nation-aware links to public guidance), with site-specific phone destinations. Local managers may add contacts; they may not weaken the stop-book rule.
Practice-level pathway examples and public NHS.uk orientation remain on the NHS/private practice guide—reuse that thinking per site, then lift shared never-dos to global.
Enterprise metrics and QA (no invented benchmarks)
Track site and portfolio against your baselines:
- Answered vs abandoned on automated paths
- Eligible bookings completed (by pathway tag if you record one)
- Booking corrections / wrong-pathway incidents
- Escalation rate by reason (identity, eligibility, rules, urgent, system error)
- Transfer / task clearance time
- Time-to-onboard a new site to pilot
- Share of sites on live write-back vs message-only (estate honesty metric)
QA cadence: weekly transcript sample per live site class; monthly matrix review after NHS/public guidance or contract changes; incident review when a wrong-pathway book occurs. Owners attend—not only the vendor.
Portfolio review questions: Which sites generate the most eligibility escalations? Which corrections cluster on one type mapping? Which acquisitions are still message-only past the agreed target date—and is that a connector gap or an ownership gap?
Avoid publishing “group-wide abandonment eliminated” or similar absolutes. Report movement vs baseline by wave.
Claims softened from earlier versions of this URL
| Previous | Now |
|---|---|
| Flawless NHS/private tag sync without oversight | Status as matrix input; escalate on ambiguity; staff remain accountable |
| Sensei listed as API integration | Removed; verified set includes Dentally, CareStack, Semble, Aerona and Exact (via Exact Online Booking when enabled) |
| Instant enterprise scale / absolute compliance | Phased onboarding; governance; residual risk |
| Shallow-wrapper rhetoric as proof | Replaced with configuration + governance model |
| Universal deep PMS logic inheritance | Mapping + live availability + practice rules (guardrails) |
Frequently asked questions
What is the matrix?
Site × contract/service × patient status × clinician × appointment × policy—resolved before auto-book.
Vs practice-level page?
That page = one clinic. This page = portfolio governance.
Vs DSO page?
DSO page = group reception broadly. This page = mixed-contract architecture and change control.
Eligibility automation?
No—escalate when unclear.
Every site live day one?
No—mixed PMS and unsigned matrices stay message/task.
Who approves changes?
Named global/regional/site owners; versioned widenings only.
What to measure?
Your baselines for answer, book, correct, escalate—site and portfolio.
Mixed NHS/private groups scale automation by making the matrix explicit, not by hoping a single bot “understands UK dentistry.” Start with two contrasting sites (for example NHS-heavy vs private-only), freeze versioned packs, then wave outward—keeping message-only sites honest until connectors and approvals exist. If you want a facilitated map of sites, contracts, PMS and bookable types before a pilot wave, use the CTA below for group workflow mapping.