Read -> Match Window -> Rank -> Review -> Schedule

Changeover Maintenance Readiness Agent

For the next product changeover window, the agent tells maintenance which equipment must be serviced then, ranked by criticality, so a job deferred to that window does not turn into a breakdown. It reads open work orders, preventive-maintenance due dates, the defect log, equipment criticality, and the changeover calendar as an input only. It returns a ranked do-this-during-the-window checklist per equipment with the reason and the risk if the job is skipped, and it does not plan production.

The Problem

The window came. The right jobs were not staged. The pump failed two days later.

A changeover window opens on Friday night, the only time the C-train is down and the agitator can be opened. Maintenance uses the window for the jobs that happened to be top of the list, and a preventive task on that agitator gets pushed again because nobody flagged it against the window. The train restarts, and two days later the agitator seal fails, forcing an unplanned shutdown that costs far more than the window would have.

What Changes

The window matched to the jobs that can only be done then

The agent looks at the next changeover window and pulls the open work orders, due preventive tasks, and logged defects for the equipment that window makes accessible. It ranks them by equipment criticality and how overdue they are, and returns a checklist of what to service during the window with the reason and the risk if the job is skipped. Maintenance walks into the window knowing which jobs matter most.

Humans stay in control: the agent ranks the candidate jobs, and your maintenance planner decides what is scheduled into the window.

HOW THE AGENT WORKS

TRIGGERLLMAPIDECISIONACTIONWAIT

TRIGGERChangeover window approachesFires a set number of days before the next window on the changeover calendar.

APIRead the changeover calendarReads the next window and which equipment it makes accessible, as an input only.

APIRead work orders and PMsPulls open work orders, preventive-maintenance due dates, the defect log, and equipment criticality from the CMMS.

LLMMatch jobs to the windowSelects the jobs that need this equipment down and can only be done in the window.

LLMRank by criticality and riskRanks the matched jobs by equipment criticality, overdue status, and the risk if skipped.

DECISIONAny must-do jobs?

ACTIONSend ranked checklist to plannerDelivers the ranked do-this-during-the-window checklist to the maintenance planner.

The agent does not schedule work, assign crews, or plan production. Your maintenance planner does. The agent removes the manual cross-check of work orders against the changeover window.

HOW TO BUILD IT

Three levels of depth. Pick the one that matches where you are.

For the ops person who needs to understand what's involved and brief their team.

Your Maintenance DataWork orders, PM due dates, defect log, criticality, changeover calendar

The AgentWorkflow engine + LLM

Your Maintenance DeskPlanner reviews the ranked checklist

WHAT YOUR TEAM PROVIDES ONCE

CMMS export - your open work orders, preventive-maintenance due dates, and defect log
Equipment criticality - the ranking that says which assets hurt most when they fail
Changeover calendar - the upcoming windows and which equipment each one takes offline, read only
An equipment-to-window map - which assets a given window actually makes accessible
A notification channel - email, Teams, or Slack where the checklist lands

What happens day-to-day: a set number of days before a changeover window, the agent reads the window, pulls the open and due jobs for the equipment that window frees up, and ranks them by criticality and risk. The planner gets a checklist of what to service during the window with the reason for each job. The planner decides what goes on the schedule.

For the IT person or technical lead who needs to scope the integration.

INQUIRY SOURCES

CMMS export (work orders, PM due dates, defect log)

Equipment criticality ranking

Changeover calendar (read only)

WORKFLOW ENGINE (N8N)

Days-before-window schedule

Fetch changeover window

Fetch work orders and PMs

Select window-eligible jobs

Rank by criticality and risk

Must-do jobs present?

Post checklist to planner

CONNECTS TO

Maintenance system (SAP PM / Maximo / Fiix)Read: work orders, PM due dates, defects, criticality. Write: none

Changeover calendar (production scheduler / spreadsheet)Read: window dates and equipment offline. Write: none

Asset register (CMMS / spreadsheet)Read: equipment-to-window mapping. Write: none

Notification (Email / Microsoft Teams / Slack)Read: none. Write: the ranked window checklist

WITHOUT PMS API

Without API access, the CMMS export and the window list are dropped into a shared folder and the agent reads those on a schedule. The ranked checklist is the same, it just reflects the last export.

WITH PMS API

With API access to the CMMS and the changeover calendar, the agent rebuilds the ranked checklist as work orders and windows change, so the planner always sees the current picture before the window opens. Best experience.

For the person who wants to build this agent step by step. Each node below maps to an n8n workflow node.

TRIGGERNode 1

Trigger - window approaching

n8n Schedule node that fires a set number of days before each changeover window.

Output: The next window and the date it opens.

APINode 2

HTTP Request - read jobs and window

GET open work orders, PM due dates, defects, criticality, and the equipment the window frees.

Output: A dataset of candidate jobs keyed by equipment tag.

LLMNode 3

LLM - pick jobs that need the window

Send the candidate jobs and the window's offline equipment, ask which jobs can only be done while that equipment is down.

You are a maintenance planning assistant. From the candidate jobs, select those that require the equipment the changeover window takes offline and that cannot be done while the plant runs.

Jobs: {jobs}
Equipment offline in window: {window_equipment}

Respond ONLY with valid JSON. No explanation.

INPUT

jobs on P-102 agitator (needs shutdown) and P-105 filter (can do online); window takes C-train offline including P-102

OUTPUT

{"eligible": [{"work_order": "WO-771", "equipment": "P-102", "task": "agitator seal PM"}], "excluded": [{"work_order": "WO-780", "equipment": "P-105", "reason": "can be done online"}]}

LLMNode 4

LLM - rank the eligible jobs

Send eligible jobs with criticality and overdue status, ask for a ranking with reason and risk if skipped.

Rank these window-eligible jobs by equipment criticality and how overdue each is. For each, give the reason it belongs in the window and the risk if it is skipped.

Eligible jobs: {eligible_jobs}
Criticality: {criticality}
Today: {today}

Respond ONLY with valid JSON. No explanation.

INPUT

WO-771 agitator seal PM on P-102, criticality high, 45 days overdue

OUTPUT

{"ranked": [{"rank": 1, "work_order": "WO-771", "equipment": "P-102", "reason": "high-criticality seal PM 45 days overdue, needs shutdown", "risk_if_skipped": "seal failure and unplanned C-train shutdown"}]}

DECISIONNode 5

IF - any eligible jobs

n8n IF node on ranked list not empty.

Output: Routes the ranked list to the notifier.

ACTIONNode 6

Send - notify planner

n8n Email or Teams node sending the ranked checklist before the window.

Output: A message with the ranked window checklist per equipment.

MVP vs Advanced — Where To Start

MVP — START HERE

Start with one train and its next window so the equipment list is small.
Export the CMMS jobs and the criticality ranking to a spreadsheet the agent can read.
Map by hand which equipment the window takes offline before the agent uses it.
Have the planner confirm the first ranked lists to check the eligibility logic.
Send the checklist to one planner and compare it against what they would have chosen.
Only then extend to all trains and their windows.

ADVANCED — ADD LATER

=+ Connect the CMMS API so the checklist rebuilds as work orders and PM dates change.
=+ Pull the changeover calendar live so a moved window updates the checklist automatically.
=+ Add defect-log severity so a repeat defect lifts a job's rank.
=+ Track jobs that were skipped in a window and later failed to tune the ranking.
=+ Add spare-parts availability so a job with no parts on hand is flagged, not ranked to do.
=+ Produce a post-window report on what was done versus what was recommended.

Full Video Walkthrough Coming Soon

Watch the agent rank an overdue agitator task into the next changeover window.

Everything on this page is free to learn.

When you're ready to build — do it yourself or bring us in.

Book a call to scope this

OTHER AGENTS IN LEASING & MARKETING

Critical Spares Stockout-Risk Agent

Flags critical spares one failure from a stockout, plus stale levels and dead stock

Frequently asked questions

Changeover Maintenance Readiness Agent is an AI agent that ranks the maintenance to do in the next changeover window by risk.

The window came. The right jobs were not staged. The pump failed two days later. The Changeover Maintenance Readiness Agent addresses this: ranks the maintenance to do in the next changeover window by risk.

No. Your team stays in control. The agent does not schedule work, assign crews, or plan production. Your maintenance planner does. The agent removes the manual cross-check of work orders against the changeover window.

This is 1 of 2 agents in Engineering & Maintenance. → · Explore all 15 departments →