How to Automate iTero to 3Shape Workflow in 2026

If you're reading this at 7:30 AM with your coffee, clicking through the iTero portal, downloading STLs one at a time, renaming them, dragging them into 3Shape case folders, and wishing you had an extra pair of hands — this post is for you.

I've been running a solo dental lab in St. Catharines for years. For the first two of those, I did the iTero → 3Shape handoff manually, every single morning. Somewhere around case number 400, I realized I was spending about 35 minutes a day on pure clicking and file-shuffling before I'd even touched a margin line.

That's 140 hours a year. Gone. Before design work. Before anything billable.

Here's what I figured out, why the "AI" solutions most labs are being sold don't quite work, and the actual architecture a modern lab can use in 2026 to automate this in a way that's cheap, reliable, and doesn't break the first time iTero updates their portal.

What the manual workflow actually looks like

Let's be specific, because "automating iTero" means different things depending on what part of your morning you're trying to eliminate:

1. Log into the iTero portal (bff.cloud.myitero.com/labs/home)
2. Scroll through the case list, eyeballing which ones are new overnight
3. Open each case, download the Rx PDF, download the STL(s), sometimes download photos
4. Create a new folder in your 3Shape cases directory, name it by patient or case number
5. Rename each STL to what 3Shape expects (upperjaw.stl, lowerjaw.stl, preop.stl, etc.)
6. Open 3Shape Dental System, create a new order, manually type in patient name, dentist, due date, tooth number, material, shade — all things already written on the Rx you just downloaded
7. Import the scans through the 3Shape scan import dialog
8. Save, and now finally you can start designing.

If you have three cases, that's fifteen to twenty minutes. If you have eight cases, that's an hour. Every single day.

Why most "solutions" don't actually work

A few products market themselves as "iTero automation" or "scan downloaders." Having tried most of them, here's what they actually do — and where they fall short for small labs:

1. Generic downloaders

They pull files from the iTero portal on a schedule and dump them into a local folder. Great, you've saved 10 minutes of clicking. But you still manually create the 3Shape order, still manually type the Rx data, still manually rename files. You're automating a small piece of the problem.

2. Enterprise LMS integrations

Big platforms like DLMS claim iTero integration. In practice, their integrations require IT setup, cost $200+/month on top of your base plan, and break any time iTero changes their API. They're also built for labs large enough to have a dedicated IT person — which most labs aren't.

3. Off-the-shelf "AI" tools

Some newer services promise to read the iTero Rx with AI and build 3Shape orders. The read part works reasonably well — modern vision models have no trouble extracting tooth numbers and shades from an Rx image. The write part is where most of them fail, because 3Shape doesn't have an open API for order creation. Third-party tools work around this by either building XML manifests that 3Shape imports, or by scripting the 3Shape desktop app itself. The XML approach is more reliable.

The actual architecture that works

Here's the pattern I arrived at after a lot of trial and error. It has four parts:

1. A pull agent that runs on your lab PC

A small Windows service that checks the iTero portal every 5 minutes, logs in with your credentials, and downloads any new cases to a local folder. This has to run on a computer that's on during lab hours — typically the workstation that runs 3Shape.

The agent has to be stateful: it needs to remember which cases it has already downloaded so it doesn't re-pull the same thing every five minutes. A small JSON file works fine for this.

2. An AI parser for the Rx

Once the Rx PDF is downloaded, a vision model reads it and extracts the structured data: patient name, dentist, due date, tooth number, material, shade, special instructions. Anthropic's Claude and OpenAI's GPT-4o can both do this reliably on clean iTero Rx forms. Handwritten addenda take a bit more care.

3. A 3Shape XML generator

3Shape Dental System can import orders from an XML file. The format is documented (somewhat) and once you have a template, generating it from the AI-extracted Rx data is straightforward. You'll need to know:

• The 3Shape-native tooth numbering scheme (FDI)
• The material codes your lab uses (these are often lab-specific)
• The directory structure 3Shape expects for scans (scans go in a sub-folder with specific filenames)

Once you have this generator written, the workflow becomes: AI reads Rx → JSON → XML file + scan folder → drop into 3Shape watch directory → 3Shape picks it up on next scan.

4. A visual board so you actually know what's happening

This is the part most labs forget. If you automate the ingestion but still don't have a visual of what's on your plate today, you've just moved the chaos upstream. You need a scheduling board that shows, at minimum: what came in today, what's due when, what's at which production stage, and what's ready to ship.

The build-vs-buy decision

Everything I just described can be built. I know, because I built it. Over about eight months of nights and weekends.

If you're a software-inclined lab tech with a lot of patience, you can absolutely do it yourself. You'll need:

• A place to run the Windows pull agent (your lab PC works)
• An AI API account (Anthropic or OpenAI — expect around $5-15/month in usage fees for a 1-tech lab)
• Some way to host your server-side logic (Netlify, Vercel, Cloudflare all have free tiers that work)
• A database for case state (Firebase's free tier is generous)
• A front-end framework — whatever you know
• Time. A lot of time.

If that list made you want to close this tab, there are done-for-you options. My Lab Board bundles all of this — the iTero agent, the AI Rx parser, the 3Shape XML generator, and the scheduling board — into one subscription. (Full disclosure: I built it. I was getting tired of explaining my DIY setup to other lab owners, so I packaged it up.) There are also enterprise options like DLMS and LabStar that do parts of this, though they tend to be priced and scoped for 20+ person operations.

What this actually changes about your day

When this is working properly, your morning goes like this:

1. You walk in at 8 AM, pour coffee.
2. Your board already shows 6 new cases that came in overnight, sorted by due date.
3. Each case has the Rx parsed, patient data populated, scans already imported into 3Shape.
4. You pick the top case, open it in 3Shape, start designing.

The 35 minutes of clicking is gone. You're designing at 8:02 instead of 8:37.

Over a year, that's close to 150 hours of extra bench time. For a one-tech lab that's probably 30-50 extra cases of capacity, just from not doing data entry.

Common gotchas

iTero portal changes

The iTero web portal UI changes every few months. If your automation scrapes the UI (instead of using a proper API), it will break. Build it so it fails loudly rather than silently producing bad data. An alert in Slack or email when a case fails to parse is much better than silent corruption.

Handwritten notes on the Rx

Dentists love to scribble a "PLEASE RUSH" or "shade matches #8 — check photo" note on the Rx. AI reads these fine, but your data model needs a free-text "notes" field to capture them, or they'll get dropped.

Non-iTero scans

If you also accept cases from Medit, TRIOS, or other scanners, each one has its own portal and file formats. The same architecture applies, but you'll need a parser per scanner family. Start with iTero since it's usually the highest volume.

Shade and material mapping

"A2" and "A 2" and "a2" are all the same shade. Your data pipeline needs a normalization layer. Same for materials — "zirconia," "zir," "zr" all need to map to one canonical value before your XML generator sees it.

Bottom line

iTero-to-3Shape automation is not a nice-to-have anymore. For a 1-5 tech lab, it's the single highest-ROI workflow change you can make. The tech to do it has been commoditized (AI models are cheap, cloud hosting is cheap), and labs that don't adopt it over the next two years will find themselves at a real disadvantage against labs that have.

Whether you build it, buy it, or cobble it together — just stop doing 35 minutes of clicking every morning. Your bench time is worth more than that.