Example agent
Work orders from field notes: an AI agent for IBM Maximo
An example Sigma agent for maintenance teams: it reads IBM Maximo work orders and asset history, cites procedures, and drafts work orders a supervisor approves.
· 1 min read · By Sigma editorial team
- Function
- Energy and utilities
Anatomy of this agent
An agent is instructions, skills, tools and knowledge. For each request it decides which of them to use; there is no fixed flow. Sensitive actions wait for a person’s approval.
Instructions (excerpt)
You help maintenance planners, supervisors and technicians at our sites. Answer questions about work orders and assets from Maximo, and always say which records you used. When someone describes a fault, draft a corrective work order using our work order conventions and propose creating it; never guess an asset number. For procedures, search the maintenance manuals and cite the document and page. If Maximo and a manual disagree, say so and do not choose for the person.
Skills
- work-order-conventions: our failure codes, priority matrix and description format (SKILL.md)
Tools
- search_work_orders (REST tool, IBM Maximo Manage API)
- get_asset_history (REST tool, IBM Maximo Manage API)
- create_work_order (REST tool, IBM Maximo Manage API)
- Company Search (native tool, over Spaces)
Knowledge (Spaces)
- Maintenance manuals and site procedures (Space, folder per site)
Needs a person’s approval
- create_work_order: always
- search_work_orders and get_asset_history: off after the agent's first review
What this agent does
The people who need the maintenance system most are often furthest from a desk. A technician who finds a leaking pump wants a correct work order, with the right asset and failure code, without fighting the screens on a tablet.
The agent turns the technician's note into a complete work order and checks the asset's history first. Nothing is written to Maximo until a supervisor approves.