AI opens new horizons for service delivery

As a new industrial era dawns, generative artificial intelligence opens up a vast array of opportunities, from training to maintenance and repairs, to support operational users.

For day/night target detection and identification, the latest version of Thales’s SOPHIE Ultima handheld thermal camera for land forces (part of the SOPHIE 4 family) was on show in a whole new way at Eurosatory 2024, the global defence and security trade show. Visitors to the Thales stand had a chance to see how cutting-edge generative AI capabilities can be applied to a wide range of support services for the thermal imager and accessed directly by users on a tablet computer.

Thanks to their natural language processing capability, generative AI algorithms like those on show at Eurosatory – developed in collaboration with the Thales Digital Factory – can leverage large volumes of detailed information, such as Thales product data, to generate new content in response to prompts and questions from human users.

Designed as a software building block that can be used with all Thales products, the new generative AI solution is secure, trusted and specialised, having been trained by Thales’s product and support experts. These qualities make it ideal for use in military environments. In addition, ongoing supervision by Thales experts ensures continuous improvement of the responses provided by the AI tool.

Support for learning and troubleshooting 

Generative AI can be used to deliver training in the use of equipment, for example by creating questions on equipment characteristics. It can also be used in maintenance, thanks to chatbots that will provide guidance to help maintainers to troubleshoot faults. Until now, searching through maintenance manuals to find the right fix has been a tiresome and time-consuming exercise. But generative AI can search and summarise information from document libraries at unprecedented speed, while allowing maintenance technicians to tailor questions to their specific issue. For example, if they ask the chatbot how to replace a particular part, the generative AI solution will quickly provide a summary of the exact procedure – and eventually they could even generate how-to videos on demand!

When a product cannot be repaired by a customer’s own maintenance teams and has to be returned to the manufacturer, generative AI will also be able to support Thales’s in-house repair specialists by searching through document libraries, particularly for information about older parts, where relevant skills are more difficult to keep up to date. Repair manuals, lifetime maintenance logs and spare parts data can all be leveraged to propose potential solutions to technicians in minutes rather than days, as is sometimes the case with the techniques available today.

Technological advances driving future developments

Thales’s generative AI solution, trained using existing documents as well as specific input from experts, still has to pass a number of milestones before making the step up from proof-of-concept to a fully fledged product offering. One of the key challenges is to find the right business model, and designing the optimum user interface will be another crucial step. Using the human voice to ask questions will not always be possible, especially in a noisy environment such as a maintenance shop on an airbase. It will sometimes be easier to use a photo of a defective part, for example, to interrogate the software.

Tomorrow, technological advances will make it possible to respond to customers’ queries in a whole range of different formats. Thales’s AI solution can currently generate text and answer questions in English, even though document libraries are written in French. However, it is not yet capable of creating images or videos, although these functionalities would have clear benefits and should be available at some point in the future. 

Setting boundaries

The increased computing power needed for these developments will doubtless also unlock other potential applications of AI, enabling it to be incorporated into advanced maintenance and diagnostics tools, for example. But boundaries must be set to ensure that generative AI never exceeds its mandate and always provides pertinent responses without revealing confidential information.