A cloud subscription that sits on top of your unified data layer. It benchmarks, forecasts, flags anomalies and produces the compliance evidence your reporting teams need. Advisory only, human-in-the-loop, by design.
Where it fits
AI Optimization is the intelligence layer of the puniq stack. It does not touch your controllers directly and it does not need new hardware. It runs on the clean, vendor-neutral data layer that Data Auditor builds over your existing systems, or that Infrastructure Blueprint lays in from scratch.
Most building analytics tools fail because they sit on top of noisy, mislabeled, half-connected data. Garbage in stays garbage out. Because puniq has already unified and documented every point through the Documented Data Map, the models read a building that actually makes sense.
The output is trustworthy enough to act on, because the building underneath it was made readable first.
What you see
Time-series models read the unified layer and project where load, consumption and cost are heading, then flag the moment a pattern breaks, before it becomes a fault or a runaway bill.
What it does
Each one reads the same unified layer and turns it into something a facilities team can defend in a budget meeting.
Every asset, floor and system is scored against its own history and against comparable buildings in the portfolio. You see where a chiller, an AHU or a whole wing is drifting from where it should be, in numbers you can defend.
Time-series models predict load, consumption and peak demand days ahead. The heritage is real: LSTM and SARIMA methods from our founder's microgrid and energy-storage research at TU Dresden, applied to whole buildings instead of single cells.
The system learns each asset's normal signature and raises a flag when a pattern breaks, before it becomes a fault or a runaway bill. A pump cycling too often or a setpoint quietly fighting itself surfaces as an alert, not a surprise.
One click turns operational data into the evidence your sustainability and finance teams already need: structured reporting aligned to EU Taxonomy, CSRD and GEG §71a, drawn from real metered data rather than annual estimates.
By design
AI Optimization recommends actions and surfaces anomalies, but a person on your team always approves and executes. It does not take autonomous control of your building. That keeps the product outside BSI C5 and NIS2 audit scope by design, and keeps your team in command.
The foundation
AI Optimization needs one clean, vendor-neutral source of truth. The two products that build it are designed to be the foundation this sits on.
Unifies the five or more vendor systems you already run into one readable, on-premises data layer, closing with the Documented Data Map.
See the service →
Designs the open, vendor-neutral automation layer from scratch for new builds and deep retrofits, so the data is clean from day one.
See the service →
Client testimonials
puniq brought real structure to a project we expected to stay messy. They designed the system around our goals, not a product catalogue.
Questions
Tell us what you run today. We will map the path from unified data to daily, defensible optimization decisions.