Occupancy reports from lighting sensors
FW Thorpe (Thorlux) · AI discovery and technical specification
OpenKit built a proof of concept that turned Thorlux’s live SmartScan sensor data into space-utilisation reports for three kinds of reader. Thorlux’s engineering team took OpenKit’s specification forward.
Overview
FW Thorpe’s Thorlux business runs SmartScan, a connected lighting platform covering roughly a million luminaires across 661 buildings. Every luminaire carries an occupancy sensor, so the data on how each space is used already existed. A facilities director facing a capital decision still had to read a dashboard to get an answer. Universities and large corporates make decisions about their buildings from manual space-utilisation analysis, which does not scale across a customer base this size.
Thorlux asked OpenKit to find out whether that sensor data could become automated reports a director can act on. OpenKit built a proof of concept that generated three sample reports from live Thorlux data and wrote the strategy report and technical specification for the service that would produce them.
~6.5M
Occupancy records queried and matched against the live dashboard
3
Reports, one per reader, all from one dataset
30 days
Of data needed before a report will generate
7
Models compared on cost and output quality, so the model can be swapped
Approach
Three reports from one occupancy dataset
Every report draws on the same query and the same pre-calculated metrics. Fixed thresholds set every action flag, so a space marked critical is critical by the same rule in every report and every building.
- The board gets one page covering only the spaces that need a decision. Everything performing normally moves down to the other two reports.
- Facilities managers get every space in the category, so each exception sits beside ordinary spaces they can compare it with.
- Department heads and finance get the long report. It is keyed to teams, so they read their own footprint without translating physical zones first.
Design
Accuracy and UK hosting
OpenKit’s query against SmartScan’s own data matched the live dashboard for the same site. The model never sees raw data. The service calculates every figure first, and the model writes the wording around numbers that have already been checked, which makes every report auditable. A report will not generate from less than 30 days of data, so a thin dataset cannot produce a confident wrong answer.
Customer occupancy data cannot leave the UK, so the service runs in the UK.
Result
The engineering team took the specification forward
The specification is written so an in-house team can build from it without a second round of discovery. It covers the service design, the queries, the prompt schemas and the report structures. The service runs inside SmartScan, whose existing screens display and export each report, so it sits in a platform Thorlux’s team already supports. The same design holds for one building or the whole installed base.
The strategy report sets out the tiered recommendation, the model comparison behind it and the full risk register.
Other engagements
What could your business do with AI?
Find out with experts who become part of your team. We uncover opportunities, get ideas working, and help your people build on the results. We reply within one working day.
Get in touch AI Audit