It definitely can! Full disclosure, I'm a DevRel at FlowFuse, so I'm definitely biased here - but there's quite a lot of support for AI workflows, from lightweight automation to fairly complex pipelines.
For example, I'm currently building a flow that connects device monitoring data to a central reporting structure, compares the incoming values against internal spec docs, and uses OpenAI to summarise overall operational status and any drift/out-of-spec issues. The summaries (along with the raw JSON objects) are then sent over MQTT for multi-site compliance and operations. Once you map it out, it's surprisingly straightforward to build.
What makes Node-RED and FlowFuse stand out on this particular use case IMO is that AI flows get treated just like any other message payload. That means your AI output becomes a first-class data object in the system - you can remix, transform, mutate, or extract from it the same way you would any other JSON payload, making it pretty portable and easy to integrate.
For example, I'm currently building a flow that connects device monitoring data to a central reporting structure, compares the incoming values against internal spec docs, and uses OpenAI to summarise overall operational status and any drift/out-of-spec issues. The summaries (along with the raw JSON objects) are then sent over MQTT for multi-site compliance and operations. Once you map it out, it's surprisingly straightforward to build.
What makes Node-RED and FlowFuse stand out on this particular use case IMO is that AI flows get treated just like any other message payload. That means your AI output becomes a first-class data object in the system - you can remix, transform, mutate, or extract from it the same way you would any other JSON payload, making it pretty portable and easy to integrate.