Domains
Automation: say what the plant should do, and prove it was done.
Plants run on controllers that are slow to change and risky to touch. Senua AI turns an action in plain language into a control sequence, runs it against the plant's own controllers and checks every write by reading the device back. On site and offline, with no cloud behind it.
What has been shown
Actions such as "start pump one" and "stop all fans" were drafted into a visual control sequence and run against a real OpenPLC controller over Modbus TCP, the protocol most plant equipment speaks. Each write was confirmed by reading the value back from the controller. Watch the run.
Every write proven
A command only counts as done when the controller's own readback shows it. The operator sees what was asked, what was written and what the device reports.
The plant's normal, learned
The same engine learns how each machine normally behaves and raises the change that matters, in place of a flood of fixed-limit alarms. This is the method; results for plant equipment are not yet published.
We do not use models. No LLMs. No SLMs. Of any kind. There is no GPU, no data centre and no AI token cost, ever, and the engine runs on low-end CPUs, so it fits an air-gapped plant. If you run a plant and want to try it against a test controller, talk to us.