Van pilots naar prestaties op industriële schaal
AI is no longer the constraint. Execution with clear ROI is.
Across industrial organisations, AI pilots have multiplied, tools have matured, and investment continues to grow. Yet many leaders see a familiar pattern: promising experiments that don’t translate into sustained operational performance, because they never scale into the processes, roles, and routines that run the business.
In 2026, the winners will treat AI like any other performance investment: with clear owners, measurable outcomes, and a disciplined path from use case to rollout. The benchmark is improved operations performance: Reduced downtime, higher throughput, better quality, lower cost-to-serve, improved service levels, lower working capital, and reduced operational risk. And delivered with payback you can explain.
But there’s a non-negotiable foundation beneath all of this: clean, contextual, real-time data that is agent-ready. Without it, even the most advanced AI systems or agents are working blind. Data quality, latency, and accessibility determine how effectively insights can be generated, decisions made, and actions executed.
Agentic automation is the next step once that data foundation, discipline, and governance are in place. It shifts AI from generating insights to coordinating and executing work across end-to-end workflows with defined guardrails and human control where it matters.
Companies that achieve real impact in 2026 will be those that move beyond experimentation, prove ROI from AI at scale, and then embed agentic automation into their operating model to accelerate what works.
Content
1. From AI tools to agentic operating models
2. Reinventing the operating model
3. From experimentation to measuable ROI
4. The use of pre-built agentic solutions accelerate
5. Multi-agent systems become the norm
7. The rise of the agentic command center
6. Trust must be engineered
7. Data becomes a performance multiplier
8. What leaders should do now
1. From AI Tools to Agentic Operating Models
Early AI deployments focused on insight generation: forecasts, predictions, recommendations. These tools supported decisions but rarely executed them.
Agentic automation changes that dynamic. AI agents can now:
- Interpret context
- Coordinate multiple steps
- Interact with systems and other agents
- Execute actions toward defined goals
This capability exposes a structural limitation: operating models designed purely around human workflows do not scale in agentic environments. Informal handoffs, manual approvals, and siloed responsibilities become bottlenecks when autonomous systems are involved.
However, data readiness poses an equally critical constraint. Many processes still depend on fragmented, inconsistent, or delayed data. For agents to operate effectively, information must be both reliable and instantly available: structured, governed, and aligned to operational semantics.
Organisations with strong process orientation and data discipline are at a clear advantage here. They already manage work through defined processes, standards, roles, and continuous improvement routines. That maturity becomes a speed multiplier: it accelerates redesign, clarifies decision rights, and makes it far easier to translate an agentic concept into a governed, repeatable way of working.
2. Reinventing the Operating Model
Agentic systems can outperform traditional ways of working only when organisations rethink fundamentals.
Incremental automation layered onto legacy processes often increases complexity faster than value. In contrast, redesigning workflows around agentic capabilities and working from clean, connected data enables:
- Faster decision cycles
- Fewer handoff errors
- Continuous optimisation instead of periodic improvement
This reinvention goes beyond technology. It reshapes:
- How work is visualised and assigned
- How performance is measured
- How humans supervise and intervene
Reinventing the operating model is the minimum requirement for success.
3. From Experimentation to Measurable ROI
The era of open-ended pilots is over. Boards and executive teams are demanding clear answers:
- Where does (agentic) AI create value?
- How fast can it scale?
- How is performance measured?
Organisations that succeed focus on high-pain, high-impact processes. These typically share common traits:
- High exception rates
- Cross-functional coordination
- Material impact on cost, throughput, risk, or service
Just as importantly, they update how value is measured. Traditional automation metrics miss much of agentic AI’s impact. Leaders are also monitoring indicators such as:
- Speed of exception resolution
- Reduction in rework and escalation
- Improved resilience and decision quality
As processes span networks of suppliers, partners and customers, value will also be created across the entire network.
And above all, people must be able to trust the outcomes.
That trust is achieved through transparency and explainability: agents’ actions and decisions must be traceable, auditable, and aligned with organisational policy. Human supervisors need clear visibility of why an agent acted, what data it relied on, and what alternatives it considered. Continuous monitoring, validation routines, and human-in-the-loop reviews make trust an engineered property of the system.
4. The use of pre-built agentic solutions accelerates
Domain-specific, preconfigured agentic solutions can be launched faster and deliver results sooner because they come with the pre-built components and capabilities needed for deployment, integration, and ongoing operation.
However, this acceleration must remain controlled.
An organisation’s policies, ethics, and guardrails must be explicitly programmed into the agents to ensure decisions align with business standards, compliance requirements, and safety obligations.
People readiness is just as critical as system readiness. Employees must be prepared for new modes of collaboration with AI agents as active participants in operations. That requires investment in knowledge, skills, attitudes, and behaviours: from understanding how to supervise and escalate, to developing confidence in using agent insights.
Effective change management focuses on building digital literacy, clarifying decision boundaries, reinforcing accountability, and above all ownership of the mixed human–agent environment.
Vertical solutions are best used as accelerators, not substitutes for operating-model change.
5. Multi-Agent Systems Become the Norm
As complexity increases, single-agent approaches reach their limits. Organisations are increasingly deploying multi-agent systems, where specialised agents collaborate to execute end-to-end workflows.
This approach mirrors how human teams operate:
- Different roles
- Clear responsibilities
- Coordinated execution
In industrial settings, this model is particularly powerful for workflows that require parallel reasoning, trade-offs, and ongoing coordination, as in production scheduling across multiple lines, predictive maintenance planning across asset fleets, or supply-chain rebalancing in response to disruptions.
However, it also raises the stakes. Without orchestration, oversight, and reliable shared data, multi-agent environments quickly become opaque and risky. Scale demands structure and a unified data backbone that all agents can trust.
6. The Rise of the Agentic Command Centre
As agentic automation expands, organisations are establishing centralised control layers to manage it.
These “command centres” bring together:
- Orchestration of agents, automation, and humans
- Embedded governance and access control
- End-to-end observability and auditability
- Lifecycle management for agent changes
This is what enables autonomy at scale. Without it, organisations face agent sprawl, inconsistent controls, and growing operational risk.
7. Trust Must Be Engineered
As agents gain autonomy, trust must be designed into the system.
Effective agentic environments incorporate:
- Explicit rules encoded into workflows
- Least-privilege access to systems and data
- Human-in-the-loop controls for high-impact decisions
- Continuous monitoring and traceability
In regulated or safety-critical environments, these controls must meet the same standards as quality systems and operational technology governance.
Just as importantly, people within the organisation must trust what agents produce.
That trust is built through transparency: by showing the reasoning behind recommendations, providing audit trails for every action, and ensuring data lineage is visible and verifiable. Training and communication play a major role: teams need to understand how agents reach conclusions and where their oversight adds value. When people trust both the data and the process, they trust the outcomes.
8. Data Becomes a Performance Multiplier
Agents need data they can act on confidently.
Organisations that succeed invest in:
- Consistente definities en semantiek over systemen heen
- Governed, real-time data pipelines that eliminate latency
- Data models that encode context, not just values
- Policies and lineage that travel with data and decisions
Without this foundation, agents make technically correct but operationally poor decisions. With it, they become reliable contributors to performance, i.e. able to reason, prioritise, and act in real time.
In practice, the path to agentic automation runs directly through data modernisation. Clean, real-time, agent-ready data is crucial to enable the transformation.
9. What Leaders Should Do Now
Across industries, a clear pattern is emerging. Organisations that scale agentic automation successfully:
- Start from measurable value (operational pain + clear economics), then scale what works.
- Redesign workflows end to end, rather than automating fragments.
- Build orchestration, governance, and data readiness early.
- Redefine human roles towards supervision, judgement, and improvement.
- Measure outcomes, not activity—and link them to trusted data quality.
