Artificial Intelligence (AI) has rapidly moved from experimentation into everyday business operations. Yet for many organisations, the next challenge is no longer whether to adopt AI—but how to deploy it economically, responsibly, and at scale.

As enterprises prepare for Windows 11 refresh cycles, rising AI infrastructure costs, and the emergence of autonomous AI agents, the traditional PC looks to become an increasingly strategic endpoint rather than simply another employee device. Organisations must now balance cloud AI with on-device intelligence, while addressing security, governance, compliance, productivity and long-term ROI.

As such, the role of the Agentic PC may prove pivotal.

This Tech Monitor executive roundtable—convened in partnership with AMD—brought together senior IT and business leaders for a candid, peer-level discussion exploring how Agentic PCs, hybrid AI architectures, and the next generation of enterprise computing might help organisations accelerate innovation while maintaining control over cost, data, and user experience.

Here are some of the highlights the mid-September event which took place in Amsterdam.

The ‘who’ and ‘where’ of Agentic PCs

Three big topics of conversation emerged when roundtable attendees were introduced to the concept of agentic PCs: how best to deploy them, how to evaluate relative costs, and their impact on existing AI architecture. We’ll consider the latter two topics below but first use cases. Where might agentic PCs be deployed and who should use them? For some, extensive rollout across the enterprise might prove attractive. But for others (at least for now) discrete deployment for power users is likely to prove most effective.

Software engineers are a good example of early adopters. As power users with a “level of expertise to set it up effectively”, software developers have an immediate need and evident proficiency. Greater simplicity will open up the use of agentic PCs elsewhere in the organisation. For example, some are experimenting with deployment in customer support, a function that tends to be less mobile than others—hence applicable for a fixed device—and one that relies of curated datasets.

Other tasks that might benefit from PC-based agent automation include workflow automation, data summarisation, document creation and revision, system upkeep, meeting scheduling, and project status reporting. Attendees were asked to imagine, as an example, a PC-based AI assistant as chief-of-staff to the CEO. Additional research by IDC suggests that the departments that organisations expect to benefit most from from agentic AI on PCs are IT and security (73%); operations/supply chain (42%); marketing (38%); HR and talent management (31%); and finance (31%).

The cost equation

A number of senior leaders around the table—representing a range of organisations from financial services to chemicals to utilities—were considering agentic PCs for the first time. While most could see the potential benefits of adoption for particular functions within their company, a number were concerned about upfront costs. When considering an enterprise-scale PC refresh, even a 50% cost increase is likely to prove unaffordable for most organisations. As one attendee put it in the context of financial commitment: “how do I deploy Agentic PCs responsibly?

Any cost equation must be performed in the context of existing architectural frameworks and the demand of the users. As explored above, for those with explicit need—or a requirement to access discrete datasets in specific environments—the total cost of agentic PC ownership is likely to make sense. Similarly, as explored below, set against existing costs of AI execution—in the cloud, most notably—a user-centric device can offer a more cost-efficient alternative. It was suggested that an agentic PC can deliver a return on investment within six to seven months.

Another attendee asked about agentic PC refresh cycles and whether aging devices would still be capable of holding AI models several years into their lifetime? In response, another speaker noted that while it was unwise to try and predict the future, existing models are already getting smaller from a compute point of view. That trend will continue which will result in greater efficacy and, in turn, help extend the life of existing agentic PCs, possibly beyond conventional life cycles.

It’s a hybrid world

If the PC takes on an increasingly important role within enterprise IT, not only will the footprint of AI deployment expand but organisations may wish to review the role of other parts of that architectural framework. The role of cloud computing is likely to come under scrutiny. It remains an effective place to provide model access, especially to users who are mobile and likely to use shared devices. But for other use cases, the cost of tokens is forcing a rethink. In short, noted a number of attendees, the cost is becoming prohibitive. The future of AI will be hybrid.

Offering an alternative to both the PC and cloud, one reference use case shared during the evening featured a regional hospital and illustrated the potential role of edge computing. The hospital is deploying edge servers at key locations across its estate to provide access to models where clinicians and other workers have the most pressing, and location-based, need for access. More broadly, one way of rethinking hybrid infrastructure is to use cloud for scale, edge for immediacy, and devices for context. As always best practice dictates that organisations assess on an individual basis how to place the right workload in the right place.

Getting the most out of AI

Artificial Intelligence has rapidly moved from experimentation into everyday business operations. Yet for many organisations, the next challenge is no longer whether to adopt AI—but how to deploy it economically, responsibly, and at scale.

As enterprises prepare for Windows 11 refresh cycles, rising AI infrastructure costs, and the emergence of autonomous AI agents, the traditional PC looks to become an increasingly strategic endpoint rather than simply another employee device. Organisations must now balance cloud AI with on-device intelligence, while addressing security, governance, compliance, productivity and long-term ROI.

As such, the role of the Agentic PC may prove pivotal.

This Tech Monitor executive roundtable—convened in partnership with AMD—brought together senior IT and business leaders for a candid, peer-level discussion exploring how Agentic PCs, hybrid AI architectures, and the next generation of enterprise computing might help organisations accelerate innovation while maintaining control over cost, data, and user experience.

Here are some of the highlights of the mid-September event which took place in Amsterdam.

The ‘who’ and ‘where’ of Agentic PCs

Three big topics of conversation emerged when roundtable attendees were introduced to the concept of agentic PCs: how best to deploy them, how to evaluate relative costs, and their impact on existing AI architecture. We’ll consider the latter two topics below but first, let’s consider use case. Where might agentic PCs be deployed and who should use them? For some, extensive rollout across the enterprise might prove attractive. But for others (at least for now) discrete deployment for power users is likely to prove most effective. Software engineers are a good example of early adopters. As power users with a “level of expertise to set it up effectively”, software developers have an immediate need and evident proficiency. Greater simplicity will open up the use of agentic PCs elsewhere in the organisation. For example, some are experimenting with deployment in customer support, a function that tends to be less mobile than others—hence applicable for a fixed device—and one that relies on curated datasets.

Other tasks that might benefit from PC-based agent automation include workflow automation, data summarisation, document creation and revision, system upkeep, meeting scheduling, and project status reporting. Attendees were asked to imagine, as an example, a PC-based AI assistant as chief of staff to the CEO. Additional research by IDC1 suggests that the departments that organizations expect to benefit most from agentic AI on PCs are IT and security (73%); operations/supply chain (42%); marketing (38%); HR and talent management (31%); and finance (31%).

The cost equation

A number of senior leaders around the table—representing a range of organisations from financial services to chemicals to utilities—were considering agentic PCs for the first time. While most could see the potential benefits of adoption for particular functions within their company, a number were concerned about upfront costs. When considering an enterprise-scale PC refresh, attendees noted that a significant increase in device cost could prove unaffordable for most organisations. As one attendee put it in the context of financial commitment: “how do I deploy Agentic PCs responsibly?”

Any cost equation must be performed in the context of existing architectural frameworks and the demand of the users. As explored above, for those with explicit need—or a requirement to access discrete datasets in specific environments—the total cost of agentic PC ownership is likely to make sense. Similarly, as explored below, set against existing costs of AI execution—in the cloud, most notably—a user-centric device can offer a more cost-efficient alternative.

Another attendee asked about agentic PC refresh cycles and whether aging devices would still be capable of holding AI models several years into their lifetime. In response, another speaker noted that while it was unwise to try and predict the future, existing models are already getting smaller from a compute point of view. That trend is expected to continue, resulting in greater efficacy and, in turn, help extend the life of existing agentic PCs, possibly beyond conventional life cycles.

It’s a hybrid world

If the PC takes on an increasingly important role within enterprise IT, not only will the footprint of AI deployment expand but organisations may wish to review the role of other parts of that architectural framework. The role of cloud computing is likely to come under scrutiny. It remains an effective place to provide model access, especially to users who are mobile and likely to use shared devices. But for other use cases, the cost of tokens is forcing a rethink. In short, noted a number of attendees, the cost is becoming prohibitive. The future of AI will be hybrid.

Offering an alternative to both the PC and cloud, one reference use case shared during the evening featured a regional hospital and illustrated the potential role of edge computing. The hospital is deploying edge servers at key locations across its estate to provide access to models where clinicians and other workers have the most pressing, and location-based, need for access. More broadly, one way of rethinking hybrid infrastructure is to use cloud for scale, edge for immediacy, and devices for context. As always, best practice dictates that organisations assess on an individual basis how to place the right workload in the right place.

Getting the most out of AI

To feel the full benefit of AI within an organisation requires guidance, not just the adoption of the technology itself. As one attendee observed, there’s no point rolling out a co-pilot tool to 7,000 staff without direction, assistance, or support. Most are likely to ignore it and those that do use it are unlikely to employ it to its full potential. Another voice around the table acknowledged that their company had been guilty of this approach in the past. It has since increased the availability and visibility of training, and appointed AI champions in key departments. A technology’s potential must be understood before it can be deployed.

‘Hybrid Workloads and Agentic PCs: Building the Modern Enterprise’ – a Tech Monitor / AMD  executive roundtable discussion – took place on Thursday 17 September at the Waldorf Astoria, Amsterdam.