Dental Lab Communication: AI Replies on Turnaround and Remakes

Give the AI only two sources: the case record and your written policy sheet. Have it draft replies on status, due date and remake terms from those alone, with a rule to say "I will check with the bench" when the data is silent. A person approves anything about remakes, credits or late cases.

Most advice on dental lab communication talks about being friendlier, calling more, or sending photos. That is fine, but it does not address the real drag: the same five questions arriving by email and phone all day, each needing someone to open the case, read the stage and type an answer. The fix is a process that drafts those answers from your data, keeps the risky ones with a person and checks each reply before it reaches a dentist, whether you run a shared inbox, a case-management system or a spreadsheet.

Follow me on Instagram@sagnikteaches

Sort the enquiries first

Before writing a prompt, spend an hour with last month's emails and call notes. Group every dentist enquiry into types. A lab usually finds six or seven, and the first three account for most of the volume.

Connect on LinkedInSagnik Bhattacharya

Types with a "nobody decides" in the last column are the ones AI can draft safely, because the answer is a fact in your records. The ones with a person's name need a human before the dentist sees anything. That split is the backbone of the design.

Subscribe on YouTube@sagnikteaches

Build the two sources the AI is allowed to use

The case data export

Decide which fields the AI may see: case ID, product type, stage, date received, due date, courier status and any note a technician has marked as shareable. Leave out the patient's name or initials unless the tool you use is covered by a processor contract, no training on your content, and you have told your dentists in your terms. A case ID is enough for the reply. If you work from a spreadsheet, export those columns to a separate sheet for the AI so it never has the full record.

The policy sheet

This is a one-to-two page document in plain language: what your lab does about turnaround, rushes, remakes, adjustments and returns. Write it once, carefully, because everything the AI says about policy comes from here. A filled-in illustrative remake section:

• Remakes or adjustments for lab error are free; we collect the original and return a corrected case within [number] working days.
• A remake for a change of prescription, shade or design is charged at [percentage] of the original fee.
• To assess a remake we need: case ID, a clear photograph of the problem, the opposing model if relevant, and the dentist's description.
• Remake requests should be made within [number] days of delivery.
• Credits are agreed by the lab owner only.

The square brackets are for you to fill in. Do not let AI invent your terms. If the policy is not written, writing it is the first job, and the model is a good partner for turning your rough bullet list into plain sentences. The quality-control checklist and remake log covers the record side of this.

Write the status prompt and test it

For status and due-date questions, the prompt is short and strict. Give it the case row and the dentist's message.

You write replies for a dental lab to dentists. Use ONLY the case data below.
Never state a date, stage or price that is not in the data. If the case is past
its due date or the data is missing, say "I will check with the bench and come
back to you by [time]" and do not guess. Tone: brief, courteous, no jargon.
Always include the tracking link if there is one.

Case data: ID 4412 | Crown, zirconia | Stage: sintering | Received: 3 Nov |
Due: 10 Nov | Courier: not yet booked | Tracking link: [link]
Dentist message: "Hi, any update on 4412? Patient is booked for the 11th."

An illustrative reply: "Hello, case 4412 (zirconia crown) is at the sintering stage and is due with you on 10 November, the day before your patient's appointment. You can follow it here: [link]. I will let you know if anything changes." That is correct and brief. Check it for three things: the due date matches the data, it did not add a promise about courier timing (the data says not yet booked), and the link is present. A good habit is to add one more line to the prompt: "If the due date is within one working day of the patient date the dentist mentions, flag it for a person." Then the tight case is escalated rather than reassured.

Draft remake replies from your policy, with a person deciding

Remake emails are the ones that cost money and goodwill. The AI should not make the decision. Its job is to collect the facts and write a clear reply once you have decided. Two prompts do this.

The first asks for what is missing:

Using the policy sheet below, read the dentist's message and list what we still
need before a technician can assess the remake. Then draft a reply asking for
only those items, politely. Do not say whether the remake will be free.
[policy sheet]
Message: "Case 4398, the bridge doesn't seat properly on the lower left. Please
remake."

An illustrative result asks for a photograph of the bridge on the model and in the mouth if possible, the opposing model, and the date of delivery, and says a technician will review within the stated time. The "do not say whether it will be free" line is essential, because a model asked to be helpful will tend to reassure.

The second drafts the reply after a person has decided. You give it the decision: "Technician reviewed: open margin on the distal abutment, lab error. Remake free, collect original, return in 5 working days." It writes the email in the same courteous tone. You approve it. This keeps the decision with a person while saving the typing and keeping the wording consistent.

A worked example: a lab with a flood of status emails

Say an illustrative lab has four technicians and a front-desk person who handles about 60 dentist emails and calls a week. A fortnight's tally shows 24 status or due-date questions, 8 delivery and tracking, 7 pricing, 8 prescription queries, 6 rush requests and 7 remake or adjustment messages. The status, delivery and pricing groups, 39 in all, read straight from records. If each takes 4 minutes by hand, that is 156 minutes a week; drafting with the prompt and a skim takes about 1.5 minutes each, so about 59 minutes. The saving is about 97 minutes a week from these three groups alone. The remaining 21 enquiries need people, and the AI helps there only with drafting. These numbers are an invented illustration, and your own tally will differ, but doing the tally is the part that matters: it tells you which groups are worth building first.

Hand-off rules, written down

A reply process is only as safe as the rules for passing to a person. A short table pinned above the front desk:

If you use a shared inbox, label the categories and have the AI suggest the label as well as the draft. A wrong label costs little; a wrong remake promise costs a lot.

Tracking links and the self-serve answer

The cheapest enquiry is the one that never arrives. If your case system can produce a status link, put it in every dispatch and status email, and in the email footer: "Check any case: [link]". Where it cannot, a simple shared sheet or web page showing case ID, stage and due date, with dentists' access limited to their own cases, does the job. The AI helps with the wording of the footer and with a short "how to read the stages" note: received, design, milling or pressing, finishing, quality check, dispatched. When dentists can see the stage themselves, status emails drop and the remaining ones are the unusual ones. Related is the case tracking template with AI status updates, which builds the sheet itself.

Turnaround answers that are honest about capacity

The most common unforced error is promising a due date the bench cannot meet. Teach the process to use two dates: the standard turnaround for the product, written in your policy sheet, and the actual due date on the case. If a dentist asks "what's your usual turnaround for a PFM crown?", the AI answers from the policy sheet and says "typically [number] working days from receipt of a complete case". If they ask about their case, it uses the record. Add to the prompt: "Never combine the two. A usual turnaround is not a promise." For rushes, the reply template says the request is under review and gives a time for the answer, never a yes. The price list tutorial shows how to word rush fees and tiers so those answers are consistent too.

An illustrative rush exchange shows why the template matters. A dentist writes: "Can you turn round a two-unit zirconia bridge for case 4430 by Thursday? The patient travels on Friday." The draft from a rush prompt reads: "Thank you. I have passed case 4430 to our production lead, who will confirm by 2pm today whether Thursday is possible and what the rush fee would be." Check two things before it goes. The 2pm promise must match the production lead's real day, so a person supplies that time, not the model. And the draft must not quote a fee the price list does not hold: if the model adds "a 50% surcharge applies", delete it and quote the list. Once the lead has decided, the second prompt writes the yes or the no from that decision, with the confirmed date and fee filled in by hand.

Testing before a dentist sees anything

1. Collect 20 past emails of mixed types, with the answers you actually sent.
2. Run them through the prompts with the matching case rows.
3. Score each draft: correct facts, no invention, right tone, right escalation. A pass needs all four.
4. Look for the dangerous failures first: a promised date, a stated price not on the list, a remake commitment.
5. Fix the prompt, not the output, then re-run the failures.
6. Go live with approval for two weeks, with every draft read before sending.
7. Review the first fifty sent replies and count edits. If staff change more than one in five, the prompt or the data needs work.

The tool you use is less important than the discipline. A shared inbox with an AI draft button, a chat assistant with a Project holding your policy sheet, or a short automation: any can work if the sources are controlled.

Prescription and shade queries: let the slip answer, not the model

When a dentist asks "which shade did I specify?", the answer is on the prescription, and the risk is that the AI reads a handwritten slip badly. Do not ask a chat model to interpret a photographed slip and then send the answer on. Instead, have the technician confirm the field from the original slip and type it in the case record, and let the template quote it: "Our record of the prescription for case [ID] shows shade [shade] with [margin type]. If that is not what you intended, please tell us before we proceed." The sentence invites correction at the right moment, which prevents a remake. For turning slips into structured cases in the first place, see the tutorial on turning handwritten Rx slips into cases.

A small template library you can load once

Keep a handful of approved replies as the style the AI copies. Here are three, filled in with illustrative content. Load them into the same Project or prompt as the policy sheet.

• On track: "Hello, case [ID] is at the [stage] stage and due with you on [date]. Follow it here: [link]."
• Running behind: "Hello, case [ID] has taken longer than planned at the [stage] stage. I have asked the bench for a firm date and will confirm it by [time today]. I am sorry for the worry this causes for your patient's appointment."
• More information needed: "Thank you for letting us know. To assess this properly, please send [items]. We will come back within [number] working days of receiving them."

Three templates cover most of the status traffic. Notice that none states anything the data would have to supply except the bracketed parts, and that the "running behind" version never offers a new date it has not got. When a template fires wrongly, edit the template once rather than correcting each reply.

Onboarding a new dentist and a weekly digest

New customers ask the most questions, so give them the answers early. A one-page "how we work" sheet, drafted with AI from your policy sheet, covers: how to send a case, what the stages mean, standard turnaround, how to ask for a rush, what to include when asking for a remake and how to see case status. Send it with the first case acknowledgement. A second idea is a weekly digest for busy practices: a short email listing their open cases with stage and due date, built from the same data. Ask the AI to turn a table into a readable list, but let the spreadsheet or system generate the table so there is no chance of a transcription slip. Dentists who receive a digest on Monday morning stop emailing on Wednesday afternoon.

After-hours and the first reply

An auto-acknowledgement that says when a person will reply is worth more than a clever draft. Something plain: "Thank you for your message about case [ID]. We open at [time]; a member of the team will reply by [time]. You can check status at [link]." The AI can write variants for different hours, but the commitment inside them must be one you keep. If the first working-hour reply is routinely later than promised, change the promise, not the wording.

The internal note that reached a dentist

An illustrative lab let its assistant read a notes column along with the status fields. One technician had typed "might slip, check sintering load" against a case. The assistant, helpfully, told the dentist that the case "might slip". The dentist rang the owner, annoyed that the lab was predicting its own lateness to a client rather than fixing it. The note was internal and should never have been in the export. The fix was a rule: only fields on an approved list go to the AI, and internal notes never do. A related slip is stale data: a stage that had not been updated for two days led to a reply saying the case was still at design when it had been dispatched. The tell was a dentist replying "it arrived yesterday". The check is a daily export, not a weekly one, and a rule that any case with no stage change for 48 hours gets flagged for a person.

Quick checks that it is working

• Time to first reply. Track the average on status emails before and after.
• Edit rate. The share of drafts changed by a person.
• Repeat emails. If dentists ask twice about the same case, the first answer was unclear or the link is not being used.
• Remake disputes. If these rise, the policy sheet is ambiguous. Rewrite the unclear line.
• Dentist feedback. Ask two or three regular clients whether the replies are clearer than before.

If you want to map your own enquiry types and decide which to automate first, that is what my AI implementation consultation covers. Start with the tally, write the policy sheet, and let the AI draft only what your data and your written rules can support.

Questions dental lab owners ask about AI replies

Should AI send replies to dentists automatically?

Start with drafts a person approves. Status answers built from a case record are low risk and can be automated later; anything touching remakes, credits, rush dates or a disputed fit should stay with a person who can decide.

What case data can I put into an AI tool?

Use case IDs, dates, product type and stage, not patient names. If your system stores patient names or initials on the case, either leave that field out of the export or choose a tool with a processor contract and no training on your content.

How do I stop AI promising a date we cannot meet?

Give it only the due date from the case record and a rule: never state a date that is not in the data, and say 'I will check with the bench' if the case is behind. Test it with cases you know are late.

Can AI decide whether a remake is our fault?

No. It can lay out your policy and ask for the facts you need, such as photos, the original prescription and the model. A technician or the lab owner makes the decision and the reply confirms it.

Do tracking links help?

Yes. A link that shows case stage answers most 'where is my case' emails before they are written. Pair it with a one-line reply template that includes the link and the due date.

Further reads

• Dental Lab Invoice Template: AI-Drafted Statements and Chasers — Statements and payment chasers drafted with AI.
• Dental Lab Marketing: AI-Assisted Outreach to Local Dentists — Outreach to dentists once communication is tight.
• AI Automation Agency for Dentists: What to Check Before You Hire — What to check before hiring an automation agency.
• Dental Practice AI Mistakes: Consent, Data, and Over-Automation — Mistakes on the dentist's side of the relationship.
• Which Dental Front Desk Tasks Can AI Take Over? — What the other end of your emails is doing with AI.
• Explaining Dental Treatment Plans to Patients With AI — How dentists explain treatment, useful for shaping your updates.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: general process and spreadsheet practice; no vendor pricing is quoted. Remake terms and turnaround targets are the lab's own and are not taken from any third party.