Metsä Group is expanding its long-running relationship with Infosys into a broader outsourcing and automation programme covering technology operations across the Finnish forestry company’s mills and offices.

Under the multi-year agreement, Infosys will manage applications, cloud operations, workplace services, service desks, and on-site support at the boundary between information technology and operational technology. The contract spans Metsä’s global operations and is intended to replace a fragmented sourcing model with a more unified structure.

Agentic artificial intelligence will be introduced through Infosys Topaz Fabric, a services suite designed to coordinate AI agents, automation tools, and enterprise systems. The companies expect the technology to support issue resolution, service management, and cost reduction, although they have not disclosed implementation dates, contract value, performance baselines, or quantified savings targets.

Metsä generated sales of €5.8 billion in 2025 and employs about 8,800 people across businesses including pulp, paperboard, tissue products, timber, and forest services. Its industrial footprint makes the agreement more consequential than a conventional office-technology contract, because disruption at the boundary between corporate systems and mill operations can affect production, maintenance, logistics, and worker access.

One contract crosses several operating layers

Application management and workplace support are familiar components of large outsourcing agreements, but the inclusion of cloud systems, service desks, mill locations, and IT/OT interface support places Infosys across several parts of Metsä’s operating model. The resulting service must connect central business technology with industrial environments that have different availability, safety, and change-management requirements.

Operational technology controls machinery and physical processes, while conventional IT manages business applications, data, communications, and employee systems. The two domains have become more closely connected as manufacturers introduce remote monitoring, predictive maintenance, digital production planning, and data-led optimisation, yet they cannot always be operated under the same assumptions.

An office application can often tolerate scheduled downtime or a rapid software update. A mill control environment may require longer testing cycles, local expertise, tightly managed access, and a clearer understanding of how a change could affect production equipment. The value of a consolidated model will therefore depend partly on whether standardisation improves support without flattening the differences between corporate and industrial systems.

“This engagement marks a new chapter in Metsä’s long-standing collaboration with Infosys, as we simplify and transform our IT sourcing model to meet the efficiency demands of today’s business environment,” Kristiina Lammila, chief information officer of Metsä Group, said in the partnership announcement.

The companies have worked together for several years, initially around supply chain software, application management, continuous development, and technology service-management processes. Expanding that relationship reduces the time needed for a new provider to learn Metsä’s technology estate, although it also gives Infosys a larger role across systems on which the group depends.

Agentic services need firm operating boundaries

Infosys says Topaz Fabric will place agentic capabilities at the centre of the agreement, but the announcement does not identify which decisions will be delegated to agents or how performance will be measured. Early uses are likely to sit around service requests, incident analysis, knowledge retrieval, workflow routing, and suggested remedies rather than autonomous control of industrial equipment.

Those applications can reduce repetitive work where service data is well structured and the available actions are constrained. An agent might gather logs, identify similar incidents, recommend a response, or complete a low-risk support task, while a human technician retains authority over changes with wider operational consequences.

The benefits will depend on the quality of Metsä’s service records, asset information, access controls, and system documentation. Agents operating against inconsistent support data can reproduce old mistakes more quickly, while broad permissions could allow a routine task to cross into a sensitive environment.

Metsä and Infosys will therefore need to define which tools agents may use, what data they can retrieve, when human approval is mandatory, how actions are logged, and which systems remain outside the automation boundary. Adding AI to an existing service desk without addressing those controls would simply automate the weaknesses already present in the operating model.

The questions become more demanding around the IT/OT interface. Industrial environments tend to contain equipment with long replacement cycles, specialist protocols, and dependencies that may not be visible from a central service-management platform. Automated diagnosis could improve response times, but an apparently reasonable remediation step could still interrupt production if the system lacks sufficient operational context.

Consolidation changes the risk profile

Moving more services to one provider can simplify accountability, reduce duplicated tools, and create a larger body of operational data from which automation can learn. A unified contract may also make it easier to standardise service levels and compare performance across sites.

The same arrangement concentrates supplier dependence. Where one provider operates applications, cloud services, employee technology, support desks, and connections into production environments, failures in its processes or security controls can reach further across the organisation.

Metsä will consequently need retained internal expertise capable of challenging service decisions, understanding industrial dependencies, and managing a future transition if commercial or operational requirements change. Outsourcing delivery does not remove the need to understand the systems being outsourced.

The commercial test will come through measurable outcomes rather than the presence of agentic software. Resolution times, recurring incidents, plant disruption, employee experience, security performance, and total service costs will show whether the unified model is improving operations.

Until those figures emerge, the agreement indicates where industrial AI adoption is moving. Instead of beginning with an isolated model trial, Metsä is tying automation to a wider redesign of sourcing, service management, and the relationship between corporate technology and physical operations.