AI Fault Diagnostics: Honeywell's View on Maintenance
In an interview published by Plant Engineering on November 12, 2026, Honeywell's Greg Turner argues that AI diagnostics already deliver results and that integrating existing systems pays off more than replacing them. These are vendor claims, but they offer practical takeaways.
On November 12, 2026, Plant Engineering published an interview by Tyler Wall with Greg Turner, chief solutions officer for building automation at Honeywell Technologies, recorded at Honeywell Technologies Accelerate 2026. The central message: AI-based fault diagnostics are already useful in building maintenance, and to get reliable data it is better to connect existing systems than to replace them.
Keep in mind that this is an interview with a vendor executive, so the claims should be read as the vendor's point of view. The article also gives no figures or application cases that would allow the benefits to be quantified.
The context: aging infrastructure, rising costs, fewer people
According to the interview, facilities are dealing with aging infrastructure, rising energy costs and a shortage of experienced maintenance staff. Turner believes operational efficiency will remain the defining trait of the best-performing facilities over the next five years, and he breaks it down into two components: workforce productivity and energy consumption.
Where AI works today
The areas described as mature or promising are:
- Fault diagnostics: AI filters out irrelevant alarms and anomalies, helping to spot problems developing in equipment and reducing the time spent reviewing alarms.
- Variable processes: when many variables change, traditional control strategies struggle. AI can look for ways to improve process quality, reduce energy intensity and limit waste.
- Technician time: automating repetitive alarm review frees skilled staff to focus on equipment performance and operational improvements.
Turner himself acknowledges an important limit: not every AI application is yet delivering measurable results.
Integrate rather than replace
Many sites run several generations of building management and control systems. A complete replacement is costly and disruptive. The advice is to connect the existing infrastructure and extract the useful data, without changing systems just to gain access to them. Open protocols and integration technologies can make equipment from different manufacturers communicate, although older systems call for a more careful approach to data collection.
More data does not mean better decisions
Turner notes that early connected-building projects tried to collect data from every device at very high frequencies, ending up with oversized datasets. His recommendation is to first establish which measurements are needed, how often to collect them and which operational questions they answer.
Start with the operational question, then the measurement and the collection frequency.
Takeaways for engineers and integrators
For those working in automation, the article suggests a pragmatic path, consistent with the experience of many brownfield projects:
- start with alarm rationalization and diagnostics, where the benefit is easiest to tie to time saved;
- favor integrating existing systems over complete replacement;
- define data requirements from the question you want to answer, not from the amount of data available;
- ask vendors for measurable evidence, since the interview provides none.
The page I reviewed was truncated after this point, so any later sections, for example on energy or workforce, are not considered here. Overall, the interview confirms a sound trend: the value of AI in maintenance depends less on the algorithm than on the quality and relevance of the upstream data.