Hawa Sliding Solutions: From customer service to experience with AI from Microsoft

In numerous buildings around the world, high-quality and durable fittings from Hawa Sliding Solutions AG (Hawa) move sliding doors, sliding walls and sliding shutters almost silently, whether on the facade, in rooms or on furniture. Headquartered in Mettmenstetten, Switzerland, the globally active company is regarded as a market and technology leader in sliding solutions. Hawa employs around 240 people at its two Swiss manufacturing sites and at three subsidiaries in the USA, Singapore and Dubai.

Today's dynamic business landscape presents the company with various challenges. Future economic success and further growth depend mostly on proactive management and the ability to make swift, well-informed decisions. In addition, many employees will retire within the next few years and important internal know-how will be lost. In technical service, for example, this includes knowing whether certain spare parts, especially older ones, are still in stock.

Hawa is meeting these challenges with a long-term digitalization strategy. The aim is to automate and optimize processes across departments and make them interoperable across systems. Sales and Customer Service have already laid the groundwork by rolling out the Dynamics 365 Customer Engagement (CE) cloud platform. An important component of this strategy is the use of artificial intelligence (AI).

The initiative starts with a part-finder app based on Microsoft Power Platform and the Azure AI service Custom Vision for machine vision. With this app, technical service staff can provide customers within minutes whether the spare parts they have requested are available and in stock or have been discontinued. These parts, which often cost just a few Swiss francs, such as a hinge, a running rail, or a complete hardware set, formerly required a labor-intensive manual search through product catalogs that could take 20 minutes or more. As Hawa products are very durable, it is quite common for older components to be out of production and therefore lack a current article number.

The Part Finder app identifies requested spare parts in less than a minute with the help of AI-supported automatic image recognition. Its average hit rate is well above 90 percent and rising, because the AI is constantly learning. "Overall, the app speeds up the process enormously, by around 20 to 30 times. This significantly reduces the workload for service staff, leaving them more time to deal with customer concerns and other important tasks. This proactive service, combined with the high response quality enabled by AI, creates a positive customer experience and supports digital customer-experience management," says André Saiu, Head of Global Customer Experience at Hawa Sliding Solutions.

AI-powered image recognition still has plenty of untapped potential. It lets Hawa capture expertise about spare parts that was previously held in the heads of individual employees available centrally in the Microsoft Dataverse cloud data platform. It now remains within the company, even if an employee leaves.

Despite all this, employee acceptance of the app is very high. On the one hand, the idea of identifying spare parts digitally and automatically with the help of AI was welcomed. This option had not previously been considered. On the other hand, employees really enjoy working with the AI-supported Power App, as it has a modern, intuitive interface that makes it quick and easy to learn.

In practice, AI-powered image recognition works like this: when a customer (a carpenter, fabricator, wholesaler, or architect) asks by email whether a certain spare part is available and sends along a photo, the employee responsible loads the image into the Power App using drag and drop. In future, customers will also be able to upload images via the contact form on the Hawa website, which the service team will then transfer to the app.

The photo is stored in Azure Blob Storage, Microsoft’s cloud object-storage service, for safekeeping and quick access. Azure Custom Vision then identifies the spare part shown in the picture, so the result is available within seconds. The employee sees the match in the app, together with a recommended product. To keep raising the recognition rate, the AI model is retrained with images whose accuracy has first been verified by experts.

To kick-start the part-finder app, an internal project team of service and IT employees carried out a proof of concept (PoC). Even though the AI model in this feasibility study was trained on only a handful of images and therefore a very small data set, the resulting image-recognition accuracy already approached 90 percent.

The PoC formed the starting point for building and launching the Power App. To move from prototype to production, Hawa was looking for a cooperation partner with deep expertise in Microsoft Azure AI and the Microsoft Power Platform. They found that partner in ORBIS Schweiz AG, which had already rolled out Dynamics 365 Sales and Dynamics 365 Service for Hawa and is very familiar with the company's processes.

The app was developed in just ten days, much faster than planned. This was feasible thanks to a tight Hawa–ORBIS co-creation approach that used fixed time-boxed work packages and the fact that there was no need to start from scratch. During the rollout, ORBIS also created the training loop for the AI model as part of the introduction and transferred the necessary know-how to Hawa’s in-house team so they could train and further develop the app independently. All in all, the use of AI has already sped up and improved the spare-part identification process dramatically.

But this is just the beginning. In the future the roadmap calls for matching customer-submitted photos against existing virtual images. AI-supported parts recognition is a central building block on the way to an automated process - from the spare parts inquiry by email to the availability check in the ERP system.