In association with Dell Technologies Ireland
Agentic AI is beginning to change the workplace conversation. For years, AI was mainly seen as a data centre or cloud challenge: where to find enough compute, how to store enough data and how to manage cost at scale. That still matters, but it is no longer the full story.
Across EMEA, AI is moving from experimentation to execution. IDC expects AI spending in EMEA to reach $319 billion in 2026, growing 19.2% year-on-year, while the latest State of the Digital Decade report from the European Commission found that Ireland ranks fifth in the EU for generative AI adoption, with 44% of businesses embracing AI to enhance productivity and empower their employees.
That is why the desk is becoming part of the AI strategy being developed by businesses in Ireland. As AI agents move into real workflows, they are moving closer to the people and teams they support. Developers need to test agents quickly. Analysts need to work with sensitive data. Engineers want local performance for complex models. Business users want assistants that can retrieve information, call tools, and complete multi-step workflows without sending every interaction to a remote service.
A major device refresh cycle reinforces this timing. Windows 10 support ended on 14 October 2025, creating a practical trigger for organisations to modernise fleets rather than extend the life of older devices. This refresh is not only an operating system migration. It is an opportunity to rethink the endpoint as a platform for AI productivity, privacy, and security.
The AI PC ecosystem is also maturing, with more software designed to use on-device AI capabilities such as NPUs. That matters because local AI will only scale when hardware capability becomes useful in everyday productivity, creativity, security, and collaboration. Dell’s Modern Enterprise Readiness Index also found that 94% of organisations in Ireland agree it is critical that modern, AI PCs and devices support AI-driven applications and workloads and that they will play a key role in their organisation’s overall future strategy, reinforcing that the endpoint is becoming part of AI readiness, not a side issue.
For Irish CIOs and business leaders, the question is not whether every agentic AI workload should run locally. It should not. The real question is which workloads benefit from being closer to users, data, and development teams, and how those local systems connect safely to the broader enterprise AI environment.
Treat the AI PC as part of the edge AI strategy
AI at the edge is not only about factories, retail sites, hospitals, or field operations. It also includes the desk, where knowledge workers, developers, analysts, and designers interact with data and applications every day.
AI PCs can support local inference, small language models, content creation, translation, research, and productivity workflows without relying on the cloud for every task. That can improve responsiveness, reduce unnecessary data movement, and give users more immediate access to AI-supported work.
For CIOs, the key is to stop treating the PC as a mature endpoint category and start treating it as a local AI execution layer. The device becomes a productivity engine, a security control point and a bridge between users and enterprise AI workflows.
Use local compute to manage cost and iteration risk
Agentic AI changes AI economics because a single workflow can involve many steps: planning, retrieval, tool use, validation and retries. Each step can create additional inference calls and token usage.
Running selected workloads locally can help teams manage cost, especially during development, testing, and workgroup use. Local compute does not replace cloud or data centre AI. It gives enterprises another control point in a hybrid AI architecture.
Keep sensitive work closer to the user when it makes sense
Agentic AI often needs context. It may retrieve documents, query systems, summarise records, generate code or automate steps across internal tools. That makes data movement a central risk.
In Ireland, this matters because AI adoption is shaped by data sovereignty, privacy, security, and regulated-sector requirements. These considerations are locally relevant as Irish businesses increase their use of AI and the Government continues to emphasise on responsible adoption, along with continued trust and appropriate governance. Ireland’s AI strategy sets out an approach to balance innovation with proportionate regulation and building trust.
But local does not automatically mean secure. CIOs still need access controls, logging, encryption, patching, identity management, and clear rules for what data agents can use. The benefit of local AI is greater control, but only when it is managed as part of the enterprise security model.
Give developers a governed place to build and test agents
Agentic AI needs experimentation, but experimentation without guardrails creates risk. Developers need controlled environments where they can test models, retrieval systems, tools, and agent behaviour without affecting production systems or creating unmanaged cloud usage.
Workstations and workgroup systems can provide a practical bridge. They give teams local performance for experimentation while allowing IT to apply policies around data access, logging, approved toolchains, and governance.
The governance question should be simple: can teams see what the agent accessed, what tools it called, what actions it attempted and why it behaved the way it did? That requires sandboxing, audit trails, approved datasets, access boundaries, and a clear route from prototype to production.
Build a path from workstation to data centre
The most important design principle is continuity. Agentic AI will not live in one place. A workflow may begin on a workstation, be refined by a workgroup, connect to enterprise data, move into a private or sovereign environment and scale in the data centre.
If every stage uses different controls, monitoring and governance, the organisation will struggle to scale. CIOs should design for a path from desk to data centre: consistent policy enforcement, repeatable deployment patterns, common security expectations, and clear rules for when a workload should move from local experimentation to centralised scale.
The future of enterprise AI is distributed. Some AI will run in large clusters. Some will run in sovereign or private environments. Some will run at the edge. And some will run directly where employees work.
The PC is not going away. It is becoming a smarter productivity device for the AI era. Businesses in Ireland that recognise this early will be better prepared to give teams local AI capability without losing governance, security, or a path to scale. Agentic AI is moving to the desk. The question is whether the enterprise is ready to manage it when it gets there.
Tara Gale is clients solutions country lead at Dell Technologies Ireland
Ltd
