AI wastewater treatment — a discharge breach is seen two hours too late; with iLEAN, two hours early.

The problem

The plant sees everything — and it sees it in three different places.

A treatment plant (municipal or industrial) has more data than it looks. The problem is not a lack of information; it is that the information lives on three islands that rarely meet in time:

1. The LIMS — lab analyses, internal and external. COD, SS, TN, TP, sludge SVI. They arrive hours or days late, in a spreadsheet or in a PDF from the external lab.
2. The SCADA and the online probes — dissolved oxygen, redox, pH, conductivity, flow, temperature. Real time, but read shift by shift by the operator.
3. What happens outside — the industrial customer that dumps a special discharge without warning, the storm that spikes inlet flow, the sludge truck running late, the email from the external lab with a bad result.

The plant manager knows all of it, but only once they are at their desk. The night-shift operator has their screen and their instructions. The integrated environmental permit does not wait until Monday: if the discharge goes out of limits, the record stands. Sludge bulking does not come out of nowhere — it had been signaling three shifts earlier, in different places.

How it fits the IRIS system

iLEAN does not replace the SCADA — it seals the cracks between the SCADA, the LIMS and what happens outside.

The problem with the biological stage is not a lack of data: it is information living on islands that, at the critical moment (the first sign of bulking, the rogue discharge that came in upstream), does not reach the decision-maker in time. iLEAN acts as the putty that fills those gaps, without asking you to change your SCADA, your LIMS or your probes.

Connect reads the LIMS, the probes and the SCADA even though each speaks its own language. The agent cross-references with history and, if the trend points to a breach or to bulking, it warns with hours of margin. The person decides and signs.

The three iLEAN pieces applied to wastewater treatment:

• Connect — captures the biological SCADA, the online probes, the LIMS analyses and the formats the system ignores (the external lab's spreadsheet, the email from an industrial customer warning about a discharge, a photo of the sludge in the clarifier). And it captures what comes from outside through its native channels too.
• Agent — cross-references incoming load, reactor parameters, sludge age, F/M ratio, SVI and weather. It anticipates bulking, COD/SS/N/P deviations and opportunities to cut aeration with variable setpoints. It does not send an email at 10 pm: it alerts the plant manager with a concrete proposal. The person validates.
• Edge (optional) — a vision terminal over the clarifier or the reactor outlet to detect loose floc, abnormal sludge color, foam or a high Mohlman index. It works without a network — as long as it has power, it keeps seeing.

See the full IRIS architecture →

Before and after

A classic treatment plant vs. cross-referenced operation with iLEAN

Aspect	Classic operation	With iLEAN Connect + Agent (+ Edge)
LIMS analysis	Arrives hours later, stays in a spreadsheet	Captured at second zero, cross-referenced with SCADA
Bulking detection	When you can already see it at the weir	Anticipated from weak signals in earlier shifts
Aeration setpoints	Conservative, over-aeration	Variable with real load, proposed by the agent
Rogue discharge upstream	The lab finds out days later	Connect captures the customer's warning / conductivity drift
Environmental dossier (ISO 14001 / permit)	Rebuilt by hand for every audit	Generated by the agent with the real data for the period
Weekend / night-shift operation	Operator alone, deciding without context	Agent assists 24/7, the person decides and signs

Impact estimate

Impact estimate for your plant — to be validated with your numbers.

The block below is an estimate to be validated with the specific data of your plant. We put it forward so the committee has an order of magnitude; we refine it during the diagnostic.

• Municipal or industrial treatment plant with an activated sludge stage, clarifier and sludge line. SCADA and online probes in service, LIMS with internal + external analyses.
• Connect + Agent pilot on the SCADA-LIMS integration and on anticipating bulking / breaches. First value expected within a few weeks: the agent running on the last year of history.
• Indicative payback between 4 and 9 months, depending on the weight of aeration in your energy bill and the real cost of past breaches (fines, discharge suspension, sludge deterioration).
• Reasonable reduction to expect: ≥30% in documented bulking episodes and a measurable improvement in aeration cost per m³ treated, once the model is calibrated to your plant's load.

The plant manager's reasonable doubt

“What if the AI proposes lowering oxygen and the biology destabilizes?” — the AI does not decide in the biological stage; the person does. iLEAN proposes the setpoint together with the data that justifies it; the operator validates; if in doubt, they do not apply it. And in the tasks the AI does handle (summarizing, cross-referencing, recontextualizing SCADA data against the LIMS), the best models brought error below 1.5% ^[1]. Operational safety is not delegated; analytical horsepower is.

^[1] OpenAI paper “Why Language Models Hallucinate”, 2025 — on the reliability of AI in anchored tasks.