# AI as substance, not as slogan.

"AI-powered" is now table-stakes in energy software. We don't lead on it. We let the features prove it.

Every model in Wia is grounded in your actual consumption curve - not synthetic profiles, not industry averages. Anomalies surface the moment they happen, not in the next monthly report. Savings opportunities arrive ranked by payback, with a confidence band, so your team knows what to chase and in what order.

## How Optimise works

### Learn. Detect. Act.

01

### Learn
Once a meter has 24-72 hours of data, Wia builds an expected-range model for it - informed by weather, working hours, tariff windows and (where relevant) production schedules. The model retrains automatically.

02

### Detect
Anomalies surface the moment a reading falls outside the expected band. Savings opportunities are mined continuously from the data - ghost loads, schedule drift, peak demand, vampire load.

03

### Act
Each finding lands in a ranked queue with annual saving, payback and confidence. Your team triages, assigns, fixes. Measured savings are tracked back to prove the intervention worked.

## Everything Optimise does, in one place.

### Savings Opportunities
Per-technique breakdown - vampire-load, peak-shaving, schedule-drift, HVAC. Each ranked by annual saving, payback and confidence band so your team knows what to chase first. Marked done in-app, with measured savings tracked afterward.

### Solar Yield Optimiser
Solar modelling against your actual load shape, plus an operational queue of actions to capture more on-site generation. Curtailment alerts when production exceeds load. Compares modelled yield to actual every period.

### Charging Optimisation
Schedule batteries and EV fleets to capture cheap tariff windows and on-site generation. Tariff-aware, weather-aware, with manual override for known events. Outputs a daily charge schedule per asset.

### Anomaly Detection
Expected-range bands per meter per period. Deviations surface the moment they happen - pushed to alerts, never waiting for the next report. The bands adapt to your operating pattern over time.

### Forecasting
Bills become forecasts. Consumption projections by site and utility, refreshed daily, with confidence intervals you can plan around. Drives capex models and budget vs actual variance reporting.

### Site Comparison
Like-for-like across the portfolio. Top performers ranked automatically; outliers surface for investigation. Cuts the time from "where do we focus?" to "fix these six sites." Like-for-like respects floor area, opening hours and climate.

### Battery ROI
Battery sizing against your actual consumption curve, not synthetic profiles. Modelled against tariff windows, peak charges and known capacity-market values. Outputs a sizing recommendation and a payback curve.

### Capex Modelling
Solar sizing, retrofit modelling, lighting upgrades - investments against actual data, not assumptions. Each project lands with a payback, IRR and confidence band. Compare scenarios side by side before committing.

### Ghost-load Detection
Always-on equipment drawing power outside operating hours. Surfaced automatically and tracked over time so you can prove the savings after intervention. Tied into the Manage tasks workflow.

## Savings Opportunities, ranked
### A queue, not a dashboard.
Most "AI energy" tools surface insights and stop there. Wia delivers a ranked queue - every opportunity with annual saving, payback and confidence band, scoped to a specific site and a specific technique. Your team works from the top of the queue. Below: real opportunities surfaced across a logistics portfolio.

### Savings · ranked · last 30d

| Opportunity                  | Site                  | Annual € | Payback | Confidence |
|------------------------------|-----------------------|----------|---------|------------|
| HVAC schedule drift          | Tilburg DC · NL       | €184,200 | 4 mo    | ● high     |
| Vampire-load reduction        | Dallas hub · USA      | €142,800 | 2 mo    | ● high     |
| Peak-shaving · battery       | Rotterdam · NL        | €118,600 | 37 mo   | ● med      |
| LED retrofit · warehouse     | Birmingham · UK       | €78,400  | 14 mo   | ● high     |
| Compressor schedule          | Madrid · ES          | €64,100  | 8 mo    | ● med      |

Each closed opportunity is tracked for 12 months to prove the savings actually landed.

## Common questions
### How long until the AI is useful?
Anomaly detection works from day one against simple baselines. Savings opportunities and forecasting need 4-12 weeks of data to reach high confidence - the platform marks confidence on every output.

### Will the AI ever be wrong?
Yes - that's why every output carries a confidence band. Your team triages findings, marks false positives, and the model retrains. We track precision per technique per customer.

### Do you use LLMs?
No. Optimise is built on statistical models, time-series forecasting and physics-aware baselines. LLMs are wrong about energy in ways that matter - we don't use them for any forecasting or anomaly work.

### Can we export the underlying models?
You can export every input, every output and every parameter via the API. The model structures themselves are documented; the trained weights are exportable on enterprise contracts.