Skip to content
be-human.ai
← Back to the journal

AI by industry

AI for Dental Labs: Case Tickets, Shade Notes and Doctor Approvals

An AI agent for a dental lab can keep case tickets and shade notes current and route doctor approvals, with steps it must never skip.

Be Human typographic cover with the title: AI for Dental Labs: Case Tickets, Shade Notes and Doctor Approvals
Conceptual illustration created with AI · Be Human

An AI agent for a dental lab follows each case from the prescription to the returned restoration. It keeps the case ticket and shade notes current, and it drafts or routes the doctor approvals the work depends on. It never moves a case past a doctor decision on its own, and it never treats silence from a practice as consent.

Key takeaways

  • The case ticket is the unit of memory. Prescriptions, files, shade notes and messages attach to one ticket with a single current status and a named owner.
  • Shade notes belong to the case first. A doctor's past preferences can be suggested for a new case, but they are never treated as this case's approved shade.
  • Doctor approvals are gates. Design, shade, material changes, remakes and credits wait for a documented decision from the prescribing dentist.
  • An agent can check, draft and remind. It should not answer clinical questions, change a prescription or release files to a milling partner without a human step.
  • Start with one measurable criterion read directly from the tickets, such as how many cases reach fabrication with a complete prescription, and review it on a fixed schedule.

What the agent handles in a dental lab

Dental labs receive prescriptions from many practices, through portals, forms, email, faxed pages and phone calls. Each arrives in a different shape. The work then moves through design or model work, milling or pouring, layering or pressing, glazing, finishing, quality checks and shipping, and at several points it waits on a doctor. Most delays are not technical. They come from a missing shade, a prescription that changed in a message nobody attached to the case, or a question that sat in one person's inbox.

The agent's job is the connective tissue around the craft. It checks incoming prescriptions against a completeness list, keeps status and ownership accurate, writes shade and preference notes to the right record, prepares approval requests and reminds the right person when a decision is overdue. It does not replace the technician's judgment on margins, contacts or esthetics, and its instructions should say so plainly.

Case tickets: the record everything else hangs on

A case ticket is the single place where a restoration's history lives. If the ticket is complete, anyone can ask where a case is and what it is waiting for without searching three inboxes. A useful ticket holds:

  • The practice, the prescribing doctor and the approval contact for that doctor
  • The patient reference the practice uses, with clinical detail limited to what the lab's agreement and local data rules allow
  • Restoration type, site identifiers, material and the current prescription revision
  • Every file received, with the date it arrived and the version it replaces
  • Current stage, named owner and the next expected action
  • Open questions and the doctor's most recent answer to each one
  • The due date, and the reason behind any change to it

This is the same discipline that other service businesses apply to order records. The dry cleaner ticket model and the shoe repair ticket workflow rest on one rule: a status is only as reliable as the dated entry that supports it. A dental lab needs that rule with more clinical files and more approval steps.

Shade notes: memory that travels with the case

Shade notes capture what a technician needs to match a restoration: the base shade and guide used, the lighting in the photographs, characterization and stain instructions, translucency, the doctor's written preference and any photo the practice sent. These notes are fragile because they often live in the margin of an email or in a comment made during a try-in.

Keep two kinds of memory apart. The first is case memory: what this doctor asked for on this patient's case, with its source and date. The second is preference memory: what this doctor has chosen before. The agent may surface past choices as a starting suggestion, labeled as a suggestion, but it must never write them into the approved shade field of a new case.

A shade mentioned in a reply is not an approved shade until it is written on the ticket with its source.

Illustrative scenario: A practice sends a crown prescription with the shade written as a guess in the body of an email, plus one photograph taken under warm office light. The agent attaches the email and the photograph to the ticket, flags the shade as unconfirmed and drafts a question for the doctor that names the guide options the lab uses. Fabrication stays on hold until the doctor replies on the ticket, so the technician does not discover the gap at the try-in.

Doctor approvals an agent must never skip

These are decisions an agent can prepare but never complete. Each one needs a yes from the doctor or from a person the lab has authorized in writing, recorded on the ticket.

  • Confirm the prescription is complete before fabrication begins: patient reference, site identifiers, restoration type, material, and margin or contact notes where the lab requires them.
  • Obtain written shade approval when the shade is missing, contradictory or taken only from a photograph.
  • Obtain design approval before any file goes to a milling partner, a printer or an outside vendor.
  • Obtain doctor approval before any material substitution or change to the prescribed design.
  • Obtain a documented decision before a remake, rework or credit is offered, accepted or booked.
  • Route every clinical question to the doctor. The agent may record the question, but it may not answer it.
  • Store each approval with who gave it, when, which file or prescription version it covers and the channel used.

Silence is not approval. A reminder that goes unanswered keeps the case on hold and escalates to the lab's named contact after the wait the lab has agreed. A casual reply counts only if the lab's written process accepts that channel and the reply is stored on the ticket against the correct version. Where the process is unclear, the agent asks the lab owner rather than deciding for them. Rules on patient data and dental practice requirements should be confirmed with the lab's own advisors, since they depend on jurisdiction and contracts.

Permissions, tools and what stays with people

Tool access defines what an agent can actually do, so it should be scoped as tightly as any staff login. A reasonable starting set lets the agent read tickets and files, write status and shade notes, send templated reminders to approved contacts and draft replies for a person to send. It should not change billing, release files to a vendor, alter a prescription or send shipping instructions without a human step. The principle of least privilege in the OWASP authorization guidance applies directly: every permission should be needed for a named task and reviewed when that task changes.

Inbound text is untrusted. A prescription email can contain sentences that try to instruct the agent to skip an approval. Treat incoming content as material to read, never as instructions to obey, and keep the approval rules in the agent's configuration rather than in the content it processes. The OWASP guidance on prompt injection prevention describes this risk and the controls that reduce it.

Be Human designs and operates this kind of agent as managed infrastructure. A first engagement defines one useful role, the memory and tools it needs, the approval boundaries and an observable success criterion. Managed operations can then cover monitoring, recurring evaluations, integration maintenance, permission review, incident response and controlled changes when the lab's process changes. Service is remote, in English and Portuguese.

How to tell whether the agent is working

Pick one criterion you can read from the tickets without a special report. Good candidates are the number of cases that reach fabrication with a complete prescription, the number of approval requests that were answered and stored, or the number of cases reopened for a shade question. One criterion is enough for the first month.

Build a small test set from past cases, including the ones that became remakes, because those show where the process failed. Run the agent against that set after every change to its instructions, tools or the lab's process. Keep a log of every approval request, its outcome and every escalation, so a person can trace what happened on any ticket. The OpenTelemetry observability concepts describe the general pattern for this kind of record. Remakes and rework fit the same habit of tracking exceptions that quality management agents use for nonconformances.

  • Name one criterion and the ticket field that proves it
  • Assemble a test set from completed cases and remakes
  • Confirm every gated step produced a stored approval record
  • Review escalations and refusals on a fixed weekly schedule
  • Re-run the test set after any change, before the change goes live

The NIST AI Risk Management Framework offers a useful vocabulary for deciding who owns these reviews and what counts as an acceptable failure, which is worth settling before the agent goes live rather than after the first remake.

Where a dental lab should start

Start with the narrowest role that removes real waiting. For most labs that is intake completeness and shade capture, because both feed every later step and both are easy to check. Keep design approvals, vendor release and any clinical question with people from the first day. Expand only after the ticket record has been reliable long enough that you trust it during a busy week, not just a quiet one.

If you want to find the first task in your lab worth handing to an agent, take the free Be Human AI diagnostic.

Frequently asked questions

Can an AI agent approve a case on a dentist's behalf? No. The agent can check that a prescription is complete, draft the approval request and remind the doctor, but the approval itself must come from the prescribing dentist or from a person the lab has authorized in writing. The approval should be stored on the case ticket with the version it covers, and anything less should be treated as pending.

What belongs on a dental lab case ticket? A ticket should hold the practice and doctor, the patient reference the practice uses, the restoration type, site identifiers, material, the current prescription revision, every file with its version, the owner and stage, open questions with their answers, and the due date with the reason for any change. Clinical detail should be limited to what the lab's agreement and local data rules allow.

How is a shade note different from a shade reading? A shade reading is one measurement taken under specific conditions. A shade note is the wider record: the guide used, the lighting, photographs, characterization instructions, translucency and the doctor's written preference, each with its source and date. The agent should keep the approved shade separate from past preferences, which it may suggest but never treat as approved.

Does an agent need full access to patient records? No. A lab's agent needs only the fields required to fabricate and track the case. Access should be scoped to the tickets and files in the lab's own workflow, reviewed on a schedule and removed when a role changes, and the lab should confirm its data handling duties with its own advisors.

Your next step

See where an AI agent fits in your company first.

Take the free Be Human AI diagnostic. A few questions about your work show which task to hand to an agent first, what it needs to know and where a person approves.

Be Human editorial guidance. Examples are illustrative unless explicitly attributed. Sources provide background; implementation choices depend on your workflow.

Less busywork. More room for you.

Bring us the work you wish you had time for.

Book a call to try a personalized AI agent free for 7 days ↗Or take the free AI diagnostic →