Nesta’s sustainable future mission aims to reduce UK household carbon emissions by 30% by 2030, and one approach we are pursuing is accelerating the adoption of heat pumps. In the UK, this is still a relatively niche heating method for family homes, so an important part of our work involves helping gas engineers become more productive by using new technologies, such as generative AI. We have developed a generative AI chatbot prototype that aims to support heat pump installers in becoming more productive by providing instant access to necessary support materials. Our prototype, which utilises a WhatsApp interface for ease of use, has access to a database containing heat pump installation manuals. It has already undergone preliminary user testing with heat pump installer trainees and initial feedback has been promising.

In the coming phase, we plan to scale and evaluate the chatbot while partnering with a company to start testing the chatbot in a real-world setting. 

We want to identify impactful use cases to deploy generative AI chatbots for heat pump installers and advance their use within the sector. We believe that this support will enable heat pump installers to grow in confidence, be more productive and increase the quality of their installations.

We believe that the transition to heat pump technology is essential for reducing carbon emissions, yet the adoption rate among UK households remains low. One of the cited reasons for this is a lack of installer availability.

Often, newly trained heat pump installers lack confidence in taking on their first job after training and this can discourage them from developing their work in the sector. To address this, we are developing a supportive AI chatbot intended to help installers with their on-the-job queries. While the initial scope is limited, the long-term aim is to boost installer confidence and productivity to support the transition to clean home heating. 

The project involves developing a chatbot that integrates a comprehensive database of heat pump installation manuals and educational materials. We aim to test the chatbot in real-world settings, evaluate its outputs, refine its database and functionalities, and explore the feasibility of features like schematic image retrieval from manuals. Our approach combines data science with user-centred design to ensure the chatbot meets the needs of its users effectively.