How Metalúrgica Fey cut WIP inventory by 35% and compressed its planning cycle from 3 days to 3 hours, then added AI agents on top
Fey is one of Brazil's largest fastener manufacturers, with over 5,000 MTS items, 3,000 MTO items, 200 production resources and around 400 weekly production orders. The bottleneck was the planning process: slow, disconnected from execution, and unable to simulate what came next.
Highlight results
Measured after NPLAN stabilization and the subsequent AI agent layer rollout.
About Metalúrgica Fey
Founded in Indaial (SC), Fey is one of Brazil's largest fastener manufacturers. Its 47,000 m² industrial complex houses around 700 employees and a monthly production capacity of 60 million parts. The portfolio includes nuts, bolts, clamps, tube and hose accessories and special parts, serving the automotive, agricultural, tractor and motorcycle sectors domestically and across Latin America.
The operational complexity is substantial: more than 5,000 MTS items, more than 3,000 MTO items, over 200 production resources, and a weekly demand of roughly 400 production orders, equivalent to around 500 tons.
The challenge
Before NPLAN, Fey's planning operation had structural limitations that constrained both the speed and the quality of decisions:
- The planning cycle took up to 3 days to execute
- S&OP and production scheduling were disconnected, with no synchronized model between them
- Scenario simulation was not viable in practice
- Inventory health was compromised, with excess and shortage coexisting across the same portfolio
- WIP levels were high, signaling misalignment between plan and execution
The result was a slow review cycle that left the operation exposed to demand shifts and production changes it could not anticipate or respond to quickly.
Integrating planning and execution
The NPLAN deployment structured an integrated planning layer connecting S&OP, MPS, materials planning and production capacity, all integrated with SAP S/4HANA. The integration covered materials, resources, routings, bills of materials, production orders, sales orders and sales plans.
The project ran over four months in iterative cycles with frequent validation between Fey's Logistics, Planning, Production and Technology teams and NPLAN's implementation team. Decisions were made jointly through operational analyses, indicators and comparative simulations.
Key technical and methodological elements included:
- Integration with SAP S/4HANA covering all relevant planning objects, with no manual data bridges and no parallel spreadsheets.
- Inventory policy parametrization with prioritization criteria by channel and product family, replacing ad hoc decisions with structured logic.
- Scenario structuring enabling multivariable simulations that consider demand, production capacity, inventory policies and operational parameters simultaneously.
- Iterative testing rounds (unit, advanced and integrated) with continuous parameter revision, operational homologation and KPI tracking throughout the project.
What changed in practice
The planning cycle stopped being the bottleneck
A process that previously took three days now runs in three hours. The team moved from spending most of its time executing the plan to spending it analyzing and deciding.
S&OP and production scheduling became connected
Before, the tactical plan and the shop-floor schedule lived in different realities. After NPLAN, the production plan became more realistic and directly connected to scheduling, reducing the gap between what was planned and what was executed.
Scenario simulation became viable
The ability to simulate multivariable scenarios across demand, capacity, inventory policy and operational parameters changed how Fey's planning team makes decisions. The team now compares alternatives and chooses the best path instead of reacting to events.
Inventory health improved significantly
With structured policies and a connected planning model, Fey eliminated the previous coexistence of excess and shortage. WIP inventory dropped 35%.
The 2026 evolution: AI agents
With the planning platform stabilized, Fey and NPLAN moved to the next phase and introduced AI agents into the planning routine.
The architecture operates across three layers. At the base, SAP S/4HANA remains the transactional system and source of data. In the middle layer, NPLAN's calculation engine runs optimization, MRP explosion and plan validation. At the top, language models interpret requests, explain results and orchestrate interactions between agents.
AI Fundamentals in NPLAN
Why AI Agents need a Supply Chain Engine to deliver real results.
Read the next articleThe configuration includes specialist agents, such as a dedicated inventory agent, and an orchestrator agent that coordinates interactions between them to compose analyses and respond to operational requests. One concrete use case is multi-scenario comparison, where the orchestrator interacts with the inventory specialist agent to present options, explain trade-offs and support the decision.
Fey's complete operational dataset is not sent to AI models. Interactions use data samples only, preserving the confidentiality of the operation's base.
Agent Harness: the architecture behind real AI in Supply Chain Planning
Why the competitive edge in enterprise AI moved from the model to the harness around it. Guardrails, orchestration, validated tools and governance for Supply Chain Planning.
Read the next articleConcrete results
- -35% in WIP inventory
- Planning cycle compressed from 3 days to 3 hours
- Significant expansion of simulation capacity for decision support
- More team time available for analysis and opportunity identification
- Evolution of analytical and decision-making maturity across teams
- Production plan more realistic and connected with scheduling
- Greater integration between planning and execution
- Increased reliability of planning data
- Improved Supply Chain governance
The main takeaway
Fey's transformation came from integrating technology, processes and cross-functional collaboration. Once that foundation was solid, AI agents were added on top to amplify analytical capacity and response speed while keeping the planning engine at the center of every calculation.
The sequence is what made it work: a viable, connected plan first, and analytical intelligence built on top of it.
Want to transform planning in your operation?
Talk to the NPLAN team about SAP S/4HANA integration, structured S&OP and AI agent architecture for industrial planning.