Your CMMS isn’t lying — it’s incomplete. iLEAN completes it with voice, photo and an agent.

The problem

The CMMS is not a system that lies — it is a system nobody talks to.

Almost every plant has a CMMS in use, paid for and configured. And in almost every plant, a surprising half of the real maintenance work never goes in there:

1. The urgent incident. Something breaks, the operator doesn’t stop to open a laptop — they send a WhatsApp message to the shift leader with a photo and carry on. The report gets filled in “when there’s time”: sometimes on Friday, sometimes never.
2. The veteran’s knowledge. The operator with 30 years on the floor knows that when the line 4 motor makes a certain noise, it has two days left. That lives in their head and in a notebook, not in the CMMS.
3. The warning from outside. The supplier emails to say “the pump we shipped carries an old seal, we recommend replacing that batch”. It lands in the purchasing inbox, not in maintenance. And the day the pump fails, nobody cross-references that email with the work order.

The result: the CMMS ends up as a queue of pending orders, a pile of badly closed reports and an unrealistic preventive-compliance percentage. Operational reality lives in a spreadsheet, in WhatsApp and in the heads of three people. And when the audit comes, or the manager changes, that parallel maintenance system disappears.

How it fits the IRIS system

iLEAN doesn’t replace your CMMS — it seals the cracks reality escapes through.

The CMMS does well what it does: it plans preventive work, records orders, manages spare parts. The problem is what lives between the CMMS and the plant floor. iLEAN acts as the filler in those gaps: it captures by voice and photo what the operator used to send by WhatsApp, captures by email what landed in the wrong inbox, and the agent formats it against the CMMS schema before pushing it inside.

Connect carries the voice, the photo and the email. The agent formats, prioritizes and proposes. The maintenance manager decides and signs. The CMMS matches what is happening on the floor again.

The iLEAN pieces applied to an augmented CMMS:

• Connect (inbound) — phone + headset, online, full duplex. The operator dictates “the line 7 centrifuge is vibrating oddly when it ramps to 3,000 rpm” and, without stopping, keeps working. The agent takes the dictation, formats it as a CMMS work order with the right asset, a proposed priority and a technical description, and leaves it in the approval queue.
• Connect (outbound) — the supplier email about the defective seal enters the system at second zero and is cross-referenced with the pumps installed. So does the customer’s WhatsApp message about a shipping delay — because it can change the priority of a critical maintenance job.
• Agents — they read pending preventive work, open orders and asset history from the CMMS. They cross that with what Connect has just captured and propose a priority queue with reasons to the maintenance manager. If a FIFO order from 10 days ago should wait because an incident has just come in on a bottleneck asset, the agent says so — with an argument. The manager signs or adjusts.

See the full IRIS architecture →

Before and after

CMMS alone vs. CMMS augmented with iLEAN

Aspect	Classical CMMS	CMMS + iLEAN Connect + Agents
Capturing the urgent incident	WhatsApp to the shift leader; into the CMMS if anyone remembers	Operator photo/voice → order formatted in seconds
External information (supplier, customer)	Purchasing or sales inbox; never reaches maintenance	Cross-referenced with assets and orders at second zero
Work-order prioritization	FIFO or a corridor meeting	Queue proposed with criticality, context and reason
Closing the report	Free text whenever possible, often incomplete	Template pre-filled by voice; the operator corrects
The veteran’s knowledge	In their head and in a notebook	Captured by voice and turned into an asset pattern
Real preventive compliance	Reported and actual figures diverge	They match — the CMMS reflects what was done

Impact estimate

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

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

• Mid-sized industrial plant, CMMS in use, maintenance team of 10-30 people, a high proportion of incidents arriving by WhatsApp or voice that never reach the CMMS.
• Pilot: Connect in one zone (line or area), integration with your CMMS, an agent for formatting and prioritization. First value expected within a few weeks: a measurable rise in the quality and completeness of work reports, and fewer “zombie” orders left open without a real close.
• Indicative payback between 4 and 9 months. Hard lever: a single unplanned stoppage avoided because the information now reaches the CMMS in time.
• Expected reduction in the maintainer’s administrative time of around ≥30% when dictating by voice instead of typing reports.

And the maintenance manager’s reasonable doubt

“What if the agent formats the order badly and pushes garbage into the CMMS?” — hallucination is a problem of free generation, not of anchored tasks. Here the agent doesn’t invent the order: it recontextualizes a piece of data (voice, photo, email) into the CMMS schema. On anchored tasks of this kind, the best models brought error below 1.5% ^[1]. And the order doesn’t go in on its own: it waits in the approval queue so the maintenance manager can correct whatever is needed before signing. The three safety rings exist precisely for this.

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

Related solutions: RCM with AI · APM (Asset Performance Management) · Digital operator rounds