
There is a familiar pattern in enterprise technology. A major technology arrives and organisations are quickly told they need to replace what they have or risk being left behind. We saw versions of this with cloud computing, and we are seeing it again with artificial intelligence. This time, one assumption taking hold is that preparing the workforce for AI means putting AI-capable devices on desks.Â
Some employees and applications will benefit from powerful local AI processing, particularly as more software takes advantage of AI-specific hardware. But that does not mean every employee needs an AI PC today. For many organisations, the bigger risk may be spending heavily on technology they do not yet need, only to find that requirements have changed by the time it’s fully deployed.Â
Start with the workloadÂ
IT leaders first need to understand what employees actually need to do with AI and where the necessary computing resources should reside.Â
Developers, engineers and designers working with demanding applications may benefit from local AI acceleration. Many other employees will access AI through SaaS applications, browsers, virtual applications or centralised infrastructure. A contact centre employee, for example, may spend most of the day using a small set of centrally delivered applications. A significantly more powerful PC offers limited value if those applications cannot use the additional processing capability.Â
A single endpoint strategy is unlikely to make sense for every employee.Â
Look beyond the purchase priceÂ
AI investment is also arriving when IT budgets are under pressure. Parallels’ 2026 State of Cloud Computing Survey of 540 IT professionals across the US, UK and Germany found that 42% say rising costs are driving changes to their IT strategy. Meanwhile, 68% identify IT administration time as the largest hidden cost in their current environment.Â
IT departments are already absorbing subscription licensing changes, operating system transitions, security investments and existing cloud commitments. Infrastructure decisions therefore need to account for ongoing operational costs as well as the initial purchase price.Â
Continuing to use suitable endpoint hardware while delivering applications or computing capabilities centrally can provide breathing room. Devices can then be refreshed when there is a clear requirement rather than in anticipation of workloads that have yet to materialise.Â
Stop trying to predict the futureÂ
There is another reason to be cautious about refreshing an entire estate around today’s definition of an AI PC. We do not know exactly what enterprise AI will look like several years from now.Â
AI workloads are already distributed across endpoints, browsers, SaaS services, public cloud infrastructure and private environments. Security and data sovereignty may push some workloads closer to infrastructure controlled by the organisation, while economics and scalability may favour the cloud for others.Â
Rather than predicting which model will dominate, IT leaders need to support a mixture. Applications will increasingly run locally, through a browser, in a datacentre or in the public cloud. Users should not have to care where an application resides. For IT, retaining the ability to determine where workloads run provides valuable flexibility as requirements evolve.Â
Make AI serve the strategyÂ
The same principle applies to AI investment more broadly. Too often, organisations begin with an instruction to “do AI” and then look for applications that justify it. The better starting point is the operational problem.Â
The survey suggests buyers are becoming more pragmatic. Only 29% say they are willing to pay more for AI features. Instead, 47% prioritise AI for issue detection, 41% want automated application patching and 39% want to reduce administrative overhead.Â
AI should be part of a broader automation strategy, applied where it can solve a defined problem or remove manual work. Adding AI simply because it is available risks creating another technology investment that needs to be supported without delivering enough value in return.Â
Build for the ability to changeÂ
Technology strategies are often influenced by predictions that one architecture will become inevitable. Everything will move to the cloud. Everything will become SaaS. Now, AI processing will supposedly move onto the endpoint.Â
Enterprise technology rarely develops that neatly. Budgets, applications and users move at different speeds. One part of a business may need AI-enabled endpoints today while another can continue using existing hardware for years. Some workloads may belong in public cloud infrastructure while others remain better suited to private or on-premises environments.Â
A sound strategy should accommodate those differences and include a Plan B. If costs rise, a supplier changes direction or new AI architectures emerge, the organisation should have another viable path. The goal is to keep innovating without making decisions today that unnecessarily restrict tomorrow’s choices.Â
AI-capable endpoints will become appropriate for more users as the technology matures. The challenge for IT leaders is not to predict exactly what the AI workplace will look like five years from now. It is to invest according to genuine business requirements today while preserving the ability to adapt as those requirements change. Â
About the AuthorÂ
Chris Marks is a Principal Product Manager at Parallels. As an experienced public speaker with a passion for the power of community engagement, he spends time with customers and channel partners, learning about how their businesses are coping with fundamental shifts in working practices alongside rapid changes in the way services are being consumed. He’s focused on how organizations can remain competitive in today’s environment and how we can work together to deliver simplified, more agile services.Â



