Hexagon's approach to AI: two things customers need to know

Every business is on an AI-enablement journey. Our customers are looking for answers from AI and, as a technology organisation ourselves, we are looking at how we can transform. At Hexagon, our goal with AI is twofold: create the best possible products for real industrial outcomes and - as importantly - continuously improve the way that we work.

As a company that uniquely develops solutions for industries from public safety to manufacturing to aesthetic medicine, I wanted to talk about our hybrid strategy for using AI both in product development, and as a tool for our own use. We don’t think you can have one without the other.

While there are challenges inherent in applying artificial intelligence, as experienced AI optimists, we believe it plays a crucial role in democratising access to our technology. Empowering more people and enabling more industries by making physical data actionable back in the real world.

Developing domain-specific AI products

We have been working seriously to incorporate AI into our solutions since 2012, building on earlier work with machine learning. Our platforms are built for sharing data between sensors, software, and ecosystems, to turn that data into insights with our algorithms and AI working in harmony.

We have embedded AI capabilities directly into our hardware and software. From AI at the edge, literally in the chips and graphic processing units in systems like scanners and total stations; to deep learning in our processing software that turns large point clouds and images into meaningful semantic models, by identifying, segmenting, removing and much more.

We take a pragmatic, use-case driven approach to AI, leveraging partnerships, our own data, and a mix of custom and pre-trained models to drive innovation across our solutions. Hexagon operates at the intersections of multiple industrial domains, so open-source AI models from academia and hyperscalers - like Microsoft and Google - informed our strategy but could only take us so far.

As technologies like deep-learning emerged, we knew that serving our customers in an AI-driven future necessitated training models with our unique datasets to develop domain-specific solutions, purpose-built to solve the challenges of our own business and those of our customers.

Training AI models on reality

We are in a unique position to develop digital reality AI because of our access to extensive reality datasets from Hexagon sensors, which are invaluable for training AI models in areas like point cloud classification. The unmatched blend of precision sensors and software enable us to focus on the value within that data directly. For example, in Building Information Modelling (BIM), AI can drastically reduce the time to get to a BIM model - a digital twin - from data collected with our mobile handheld scanners like the Leica BLK2GO.

Our newly launched solution, Aura, exemplifies this best-of-both-worlds approach, applying our AI expertise in reality capture to the world of aesthetic medicine. The Aura product includes 13 cameras which can capture a digital twin of a person’s face with a single shot. This digital twin is then analysed by an AI model trained on our data to give dimensionally accurate insight into all kinds of features. The domain-specific AI embedded into our products is a key differentiator for us and is due to the hybrid nature of our approach.

This hybrid approach further drives us to fuse generative AI models, such as language and multi-modal foundation models, into our industry specific use-cases as diverse as manufacturing, construction and public safety.

Building an AI-enabled company

Building a truly AI-enabled enterprise requires far more than simply using AI. It demands a comprehensive strategy coupled with an ‘AI mindset’, developed across our workforce by piloting new technologies and assessing the results before implementing broader rollout. Many new AI based capabilities arise at a fast pace. To properly assess their value, we further extend our data-centric decision-making processes.

To embed AI into our work at Hexagon, we examined how it could enhance internal functions like sales, marketing, operations and more. We explored how our service engineers, for example, could use AI to rapidly find and retrieve the documentation they needed, or how marketers could locate the perfect images for campaigns from databases through AI’s powers of classification, or how our developers can write AI-assisted code.

Our strategy encompasses spreading out AI throughout our company in multiple divisions. Hexagon operates in a hub and spoke model, where the increased use of AI is not bound to one team and entity. It rather focuses on having low to high-risk AI projects taking place within the different functions of our business - the spokes - and project managers coming together to discuss their learnings at a hub level.

AI research and AI engineering specialists within Hexagon ensure AI is optimally integrated into our solutions as part of our AI Hub. Our AI Hub comprises a corporate team of experts that stay ahead of the latest AI trends and research, to develop an AI mindset across our entire workforce. The Hub is at the centre of a global network of universities and companies developing the cutting edge of industrial AI, including NVIDIA and Microsoft.