AI Agents · Dental laboratory AI copilot
A copilot for the dental laboratory.
A laboratory system already holds every case, balance and dispatch. A copilot lets authorised staff ask about them in plain language and get an answer from the live data, without opening five screens or asking someone to run a report.
The problem
Where the time goes today.
- “Where is this case?” asked by phone, again, by a clinic.
- Managers waiting for a report to learn which clinic owes what.
- Dispatch and reception answering the same status questions repeatedly.
- No quick daily picture of what happened across the lab.
How it works
What the agent does, step by step.
- An authorised user asksOn the web interface or a registered WhatsApp number.
- Permissions are checkedReception, dispatch, finance and admin see different things.
- An approved query runsFixed, read-only queries against the laboratory database. No free-form database access.
- The answer is built from the rowsCounts and statuses are quoted as returned.
- Optional follow-up is proposedFor example a status message to a clinic, sent only after approval.
Integrations
What it connects to.
“Used in our projects” means we have built and run that kind of connection. Anything else needs a feasibility review before we commit.
| System | Status | Note |
|---|---|---|
| Laboratory management system (Ye-Almaz style) | Used in our projects | The data model is documented: cases, stations, billing, dispatch. |
| In our product plans | For authorised staff and, later, clinics. | |
| Clinic notifications | Used in our projects | Status updates already exist in the lab system; the copilot would propose, not send. |
In practice
Examples of the work.
Illustrative exchanges with fictional data. They show intended behaviour, not recorded client conversations.
Case status
Reception: “Where is case C-4821?”
Copilot: “At Milling since Wednesday 09:15. Next is Quality Control. Due Thursday.” (fictional)
Daily briefing
Admin: “Summarise today.”
Copilot: new cases, cases through quality control, dispatched, delivered and any redo or remake opened, from the day’s records.
Security and control
What keeps it safe.
- Finance questions require the Finance role; a refusal does not reveal figures.
- Only fixed, read-only queries. The copilot cannot change a case or a balance.
- Every question and answer is logged.
- Clinic data is never shown to another clinic.
- Real client records are never used in demonstrations.
Implementation
How we would approach it.
- Confirm the questions that matterThe twenty questions staff ask most.
- Write the approved queriesOne query per question type, reviewed and tested.
- Evaluate on real phrasingCheck the agent picks the right query and answers correctly.
- Pilot with one teamReception or dispatch first, then wider.
Commercial considerations
What it costs, and what it saves.
- This is a concept and roadmap item, not a delivered product. No implementation status is claimed.
- If you already run a laboratory system, the work is writing the approved queries and the agent layer; if not, it starts with the system itself.
- Costs are scoped per laboratory.
Build this agent with Odontofusion.
Tell us the job. We will tell you honestly whether an agent is the right tool for it.