ChileManufacturing2025 – ongoingEmpresas CMPC
Generative AI agents for maintenance notes, maintenance planning and process monitoring at pulp mills
CMPC cut maintenance planning from up to two hours to about ten minutes with AI agents on its own data platform
Problem
Maintenance staff at CMPC's mills wrote up inspection notes by hand, and planners spent up to two hours per note searching past records to build a plan. The Guaiba mill in Brazil alone produced about 44,000 notes a year.
Approach
CMPC built AI agents on a data platform it had developed in-house over six years, which pulls together sensor data, manuals, operating standards, engineering reports and maintenance records. At Guaiba, inspectors dictate observations by voice and the agent fills in the digital form. A planning agent repeats the search that specialists used to do and proposes a plan in about 40 seconds, and the planner keeps the final decision. At the Santa Fe mill in Chile, agents watch key performance indicators, diagnose deviations and send recommendations to shift heads by WhatsApp or email.
Results
Takeaway
The agents drew on a data platform CMPC had spent six years building, and planners still make the final call on each maintenance plan.
Categories
- Size
- Large enterprise
- Industry
- Manufacturing
- Business function
- Production and quality
- Innovation type
- Process innovation
- AI technology
- Generative AIAI agents
- Adoption
- Built in-house
- Evidence
- Company's own claim
Sources
Source summaries
The article describes three AI uses at CMPC mills: a voice agent for maintenance notes at Guaiba, a maintenance planning agent, and KPI-monitoring agents at Santa Fe. It gives 13 minutes saved per note, about 44,000 notes a year at Guaiba, and planning time cut from up to two hours to about ten minutes. It gives no result for the Santa Fe agents and does not say when the agents went live.
Full details are in the original sources.
