Cyzag Intelligence
Finds what is costing you money, and finds the cause.
It prices what is drifting and proposes the most likely cause, with the evidence. Your people decide. What they confirm is kept against the equipment, so a repeat starts from the answer.
What you get
A ranked list of what is costing you money.
Problems the engine found on its own, priced, each with its likely cause and evidence.
- EV-201 Steam per tonne 8% above normal Likely: Cleaning interval stretched from four to six weeks on 14 March Action approved €180,000 a year €21 / h
- C-210 Cooling water 12% above normal Likely: Temperature transmitter TT-212 reading high since its May calibration, so the loop over-cools Under review €120,000 a year €14 / h
- P-301 Vibration rising Likely: Alignment after coupling work; realignment cured the same symptom in 2024 Evidence assembled €78,000 a year €9 / h
- FIC-220 Control loop oscillating Likely: Valve stiction, with the controller fighting it New €52,000 a year €6 / h
How it works
From a drift to a verified fix, in four steps.
- 1
Find the opportunity
It learns each unit's normal from your history, watches for drift and prices it from your cost data.
- 2
Understand the evidence
It proposes the most likely cause, shows what else it checked, and attaches the records behind it.
- 3
Take action
Your people confirm or correct the cause in a root-cause workflow, and a named owner approves the fix.
- 4
Verify the result
It measures the unit against its baseline after the fix, and keeps the outcome for next time.
What a finding is, at each stage.
- 1 · The engine Anomaly A priced departure from normal. Nothing more is claimed.
- 2 · The engine Likely cause A ranked hypothesis, with the alternatives it checked.
- 3 · Your people Confirmed cause Only now is it treated as the cause.
- 4 · The engine, after the fix Verified improvement Measured against the baseline. Only this counts as a result.
Every number comes from the engine, never the AI.
One finding, opened
EV-201: steam per tonne, from flag to verified fix.
One flag, from first signal to measured result.
- What's happening
- Steam use per tonne about 8% above what this evaporator normally needs: roughly €180,000 a year at 8,760 operating hours if left alone.
- When it started
- 14 March, the same week the chemical-clean interval was stretched from four weeks to six.
- Most likely cause
- The cleaning-interval change on 14 March. A hypothesis until your people confirm it.
- Checked and less likely: Feed rate, steam pressure.
Evidence: every claim clicks through to its source
- Steam trend, Mar–Jul
- Shift log · clean intervals
- Shift entry, 14 Mar
- Investigation · 2024
Drafted for approval
Return the chemical-clean interval to four weeks. Based on the 2024 record, steam use should recover within one cycle, and the saving outweighs the extra cleaning cost by roughly five to one.
- Engine · week 28 Likely cause proposed Cleaning interval stretched to six weeks. Feed rate and steam pressure checked and less likely.
- Agents · minutes later Evidence assembled The 14 March shift entry, the cleaning records and two similar flags from 2024.
- M. Keller, process engineer Cause confirmed “We went to six weeks to save a stop. It fouls faster than we thought.”
- Area manager Action approved Back to a four-week clean, with owner and date.
- Engine · week 32 Improvement verified Steam per tonne back within its normal.
What it finds
More than a threshold on a trend.
Each objective compares the plant with what it should itself do, not a limit typed in once.
- Equipment off its design curve
- Pumps, compressors and exchangers watched against the curve they were bought to.
- Drift from a good period
- Normal learned from when the unit ran well, by grade, season and rate.
- Loops and instruments
- Oscillating loops, sticking valves, frozen or drifting instruments.
- Energy and efficiency losses
- More steam, power or reagent for the same output, priced early.
- Alarm floods, grouped
- Forty alarms become a few incidents, each with its first alarm named.
- Mass and energy balances
- A drifting meter or a leak shows up as a gap in the balance.
Ask the plant
Questions no manual can answer.
Answers from your own records, with sources, and credit to the people who knew.
Ask the plant
“Why do nights run the compressor off-spec in winter?”
Below −5°C the suction line ices unless flow is kept above the winter minimum. The setting nights use has held every winter since 2023, and the 2024 investigation recorded why. The formal setpoint was never updated. Suggested next step: raise a change for the process engineer to review.
- Handover · J. Müller, Dec 2023
- Investigation · Jan 2024
- Trend · 3 winters
For the technically curious
How the agents work.
Each agent is a role plus the read-only tools it may use. None can change anything in the plant.
- Process agent Reads: Trends, the plant model, design curves Cannot: Change a setpoint
- Maintenance agent Reads: Work orders, equipment history, past investigations Cannot: Raise or close a work order
- Quality agent Reads: Lab results and specifications Cannot: Hold or release product
Start here
A ten-week pilot. It carries on only if the numbers hold.
Start with a working session: your engineers and ours, on your plant's own problems.