Sandvik Manufacturing Solutions
Sandvik builds an Azure OpenAI copilot into its CAM software to support customers and sales staff
Sandvik Manufacturing Solutions built a Manufacturing Copilot on Azure OpenAI Service and put it inside its Cimatron, GibbsCAM and SigmaNEST CAM software. Sandvik staff report time savings of close to 20% to 30% on average, and a new salesperson now takes about half the time to onboard.
Key figures
Copied exactly from the sources. Copy any figure with its citation.
close to 20% to 30%
Average time saved by Sandvik employees using the Manufacturing Copilot, as stated by Sandvik's Head of Collective Intelligence Engineering1
up to 30%
Productivity improvement by Sandvik employees, as worded in the story's executive summary1
about half the time
Time to onboard a new salesperson with the Copilot, as stated by a SigmaNEST sales manager1
19
Languages the Copilot supports1
What it took
The problem
Sandvik Manufacturing Solutions sells CAM software, the programs that plan how machines cut and shape parts. These products are complex. Customers want support at all hours, and Sandvik wanted to shorten the learning curve for customers and staff. Sandvik wanted to use smart technology worldwide while still looking after customers.
What they built
A Manufacturing Copilot, an AI assistant that sits inside the software and answers questions about it. It first appeared in three products: Cimatron for dies and molds, GibbsCAM for advanced machining and SigmaNEST for sheet metal. It runs on Microsoft's Azure OpenAI Service and Azure AI Search, which lets it look up Sandvik's own product documents before answering. Sandvik says it supports 19 languages.
How they did it
Sandvik built a shared service first, then a product-specific Copilot on top for each brand. Each one learned from that brand's own knowledge base. Sandvik says the training and prompt tuning took 12 months before launch. The company announced the Copilot in July 2024 for the September releases of the three products. Staff such as sales managers and application engineers use it too, and customer feedback from it flows back to the teams.
Results
Sandvik's Head of Collective Intelligence Engineering says the Copilot saves employees close to 20% to 30% of their time on average. A SigmaNEST regional sales manager says onboarding a new salesperson now takes about half the time. The Microsoft story also says employees improved productivity by up to 30%. Sandvik plans to extend the Copilot to more products and to give it to 400,000 users.
All reported figures
| Measure | Figure | Citation |
|---|---|---|
| Average time saved by Sandvik employees using the Manufacturing Copilot, as stated by Sandvik's Head of Collective Intelligence Engineering | close to 20% to 30%1 | |
| Productivity improvement by Sandvik employees, as worded in the story's executive summary | up to 30%1 | |
| Time to onboard a new salesperson with the Copilot, as stated by a SigmaNEST sales manager | about half the time1 | |
| Languages the Copilot supports | 191 | |
| Months of training and prompt tuning before launch, per Sandvik | 12 months2 |
Lessons for your project
Build one shared AI service first, then thin product-specific assistants on top. Sandvik did this instead of building a separate solution for each software brand.
Ground each assistant in that product's own knowledge base, as Sandvik did for each brand, so answers reflect that product's documentation.
Serve two audiences with one tool. The same Copilot helps customers get support and helps new salespeople learn products faster.
Measure the gain the same way each time. This story gives several different savings figures, which weakens the business case.
Caveats
- Every figure comes from Microsoft's customer story and Sandvik's own announcement. No independent source confirms them, and none says how time savings were measured.
- The Microsoft story is inconsistent on the headline: its summary says up to 30% productivity improvement, a callout says 30% average, and a Sandvik manager says close to 20% to 30%. All three are shown as worded.
- The model is not named. The story names Azure OpenAI Service only, so the model name is shown as undisclosed.
- No source gives a go-live date. The deployed date is the July 2024 announcement, which said the Copilot would arrive with the September releases.
- The story says the Copilot 'will soon be accessible' to 400,000 users. That is a plan, not a result, and is not counted as one.
- No cost or team size is disclosed.
Sources
- Sandvik revolutionizes manufacturing with Microsoft AI solutions
Microsoft Customer Stories · 2025-03-19 · Vendor case study · microsoft.com
- Sandvik, Microsoft Launches AI in CAM Software
Digital Engineering 24/7 · 2024-07-05 · News · digitalengineering247.com
Summarized in our own words from the sources above. Spotted an error? See corrections.