Can AI really save you money on your heating bill? I spent two years testing to find out

Every smart thermostat manufacturer claims their device can cut heating bills by an average of 25% or around £180 a year, but these promises are often heavily caveated. To find out if the technology can truly help, Samuel Gibbs, a technology journalist at The Guardian, spent two years comparing his own miserly manual control of a gas boiler with an AI-governed system in his 1930s solid-brick home.

The test began in the winter of 2024 with a 10-year-old Worcester Bosch combination gas boiler, a Hive thermostat with smart features disabled, and 13 radiators with standard thermostatic radiator valves (TRVs). Gibbs set a basic schedule to heat the house for a few hours in the morning and evening, manually adjusted valves in rooms as needed, and turned the system off when leaving the house.

This manual regime kept bills low, consuming 14,316.7 kilowatt hours of gas over 213 days from November to May, about 10 to 15% lower than previous years. However, the laborious effort and the frequent arguments with family members over the temperature led Gibbs to seek a more automated solution.

In the summer of 2025, Gibbs replaced the Hive thermostat with a full Tado X system, which included 13 smart radiator thermostats, wireless thermostats in the two largest rooms, and a programmer wired to the boiler. The system manages heating room by room, turning radiators off in unused spaces and, if compatible with the OpenTherm protocol, adjusting the boiler's flow temperature to reduce energy consumption.

The Tado system costs approximately £930 for the full setup, with an additional £3.99 monthly or £29.99 yearly subscription required for the AI-powered features. The AI is built using anonymised data from over 1 million homes and learns how quickly individual rooms heat up and cool down to control temperature and energy consumption more precisely.

The AI features include automatic hydronic balancing, which takes the guesswork out of radiator efficiency, and automatic actions such as turning off radiators when an open window is detected, shutting off heating when the last person with the Tado app leaves home, and preheating the house before the owner is expected to return.

Gibbs set a schedule for the AI-controlled system and recorded monthly meter readings throughout the winter of 2025 and into 2026. The AI's energy IQ feature initially highlighted that a radiator in an uninsulated porch was consuming a lot of energy, prompting Gibbs to adjust the schedule to turn it off most of the time.

After the initial adjustment, the system ran in the background without Gibbs' interference. The house felt warmer and more comfortable, but Gibbs resisted the urge to manually adjust settings to avoid biasing the results. When the final numbers were in, the results were surprising.

Across the same 213-day period with similar outside temperatures, the AI-controlled heating used 11,735.7 kWh of energy, which was 2,581 kWh less than his manual control. This 18% reduction in gas usage equated to a saving of £205.70 at the time of the test.

Beyond the financial savings, Gibbs found multiple side-benefits to the AI system. The ability to control rooms individually allowed for a warm living room in the evening without overheating the bedroom, and the system could preheat a bathroom before a shower or dry out wet coats in a porch.

The temperature sensors also measured humidity, helping Gibbs manage dehumidifier use and improve air quality. The system significantly increased comfort levels and reduced the stress of managing the heating, proving that AI can indeed save money on heating bills while making a home more pleasant to live in.