The release of GPT-5 has cemented large language models (LLMs) as the centrepiece of how we talk about and evaluate AI.
But while it’s right that OpenAI’s latest creation is grabbing the headlines, it’s easy to forget that there is a whole world of AI out there that doesn’t revolve around language but instead focuses on sensing, decision-making, optimisation and prediction.
That’s what we do at Passion Labs. As a recently launched UK-based research and development lab, we’re exploring the possibilities of AI beyond language models to include areas like reinforcement learning, sensor data and embodied agents, as well as complex optimisation problems in areas like logistics, engineering and design.
For us, that is where the real power of AI lies. After all, complex AI systems have been running behind the scenes in areas such as logistics and finance for years. For example, if you order something from Amazon, there’s a huge amount of AI in place to coordinate and fulfil deliveries.
It’s the same with social media. People often say, “I love watching TikTok because there’s so much interesting stuff, and I’m hooked for hours.” But the reason they’re hooked is because, out of the millions of videos available, the algorithm gives them exactly what it knows they want to see. It learns from their behaviour.
AI is more than talk
For me, these are some of the clearest and most accessible examples of non-linguistic AI around. It’s an important distinction to make – especially with all the noise currently being made about GPT-5 – because in many cases, the companies and organisations we speak to think they need a GenAI-based solution.
But once we start looking under the hood – once we start to understand their workflows, data and actual pain points – it quickly becomes clear that LLMs aren’t always the best fit. In many cases, the real opportunity lies elsewhere in areas such as overlooked processes, non-text data and problems that require a different kind of intelligence altogether.
For instance, one UK-based independent record label came to us because their talent spotters were spending hours manually scrolling through TikTok and Spotify to find emerging artists. By their own admission, it was hit-and-miss. So, we built an AI-powered A&R assistant that continuously scans different platforms, analysing thousands of videos every hour, cross-referencing trends with streaming data and social metrics to spot the next big thing. Within three months, we helped them discover 15 new artists they said they would likely have missed.
Supporting structural engineers with AI-driven design
In another project, we worked with a structural engineering firm to streamline how engineers validate and optimise building designs. Traditionally, this involves interpreting 3D models and running multiple simulations to test different support layouts, a process that’s slow, repetitive and expensive.
Instead, we built a bespoke neural network capable of reading point cloud and line-based geometry from architectural files and generating multiple viable structural solutions. The output allows engineers to choose the best option based on factors like safety, aesthetics or material cost.
While in the agriculture sector, we’re working on a government-funded project to help farmers make better use of their own data. We developed an AI assistant that brings together this field data with external sources like satellite imagery, weather forecasts and regulatory updates to make their farms more productive and efficient.
The importance of an AI audit
Of course, none of these projects started with a model, let alone an LLM. They all start with the same question: what’s really going on inside the business? That’s why, before we write a single line of code, we carry out a structured audit to understand the data a company holds, how teams operate, where the pain points lie, and how existing tools are (or aren’t) being used.
It’s this strategic process – not the tech – that is the starting point for any organisation serious about using AI. What’s more, you’ll be surprised at what you might unearth.
For instance, it’s not unusual to find a business with fewer than 1,000 employees running over 100 overlapping tools, many of which could be replaced with a simple workflow or a well-structured prompt. While in other cases, businesses may already have strong systems in place but lack the know-how to achieve their goals.
In other words, in their enthusiasm to embrace the latest technology, many organisations have accumulated layers of complexity without realising it. Which means the next step isn’t necessarily to adopt the latest breakthrough technology but to pause, take stock and understand what’s working, what’s not and where existing tools are being underused or misunderstood. That might not be as eye-catching as the latest headline-grabbing product launch, but it’s the kind of strategic reset that is essential if organisations are to maximise their AI investment. If you want to find out more, then contact me here or via LinkedIn.
