THE LATEST OPINIONS
FROM OUR EXPERTS

While big tech firms continue to pour billions into developing advanced models like ChatGPT and Gemini Ultra, some experts argue that the core functionality of large language models (LLMs) is already showing signs of commoditisation. At Appian’s user conference, CTO Mike Beckley noted that the key LLMs offer only minimal differences in performance, with costs of use falling dramatically.For organisations, however, the real challenge is not just which model to choose, but how to apply them safely and effectively. As Silvia Lehnis, Consulting Director for AI and Data at UBDS Digital, explains:“It’s much easier for developers but also for managers to choose something that’s well known. Not least as it’s more defensible to the board if something goes wrong.”Lehnis also highlights that mainstream models often disappoint businesses in practice, particularly when used for specialised tasks. One solution, she says, is building industry- or company-specific models:“If you’re taking a generic model… you’re still getting a lot of the generic content coming through and paying for the cost of the compute of that generic model, which doesn’t really answer that very specialised question.”This raises critical questions for enterprises about trust, defensibility, and the strategic use of AI—issues that go well beyond the hype of the latest chatbot.

As organisations race to adopt AI, the pressure on leaders to balance innovation with security has never been greater. In this piece, Silvia Lehnis, Consulting Director for Data and AI at UBDS Group, explains why AI isn’t plug-and-play, why ownership can’t sit in one team, and how security needs to be baked in from the start. From managing training data and privacy risks, to countering deepfakes and inference attacks, she outlines the practical steps leaders must take to avoid AI creating more risk than value.

Legacy systems are no longer just a burden - they’ve become a strategic threat to agility, efficiency and trust in public-sector operations. This article by Arun Manoharan, Global Head of Strategy Enablement at UBDS Digital, argues that true transformation doesn’t happen through grand AI launches or one-off migration projects, but through lifecycles of incremental change: secure foundations built first, then workflows rewired one at a time.The piece explains how achieving a future of “agentic” digital services - where autonomous systems assist civil servants in planning, executing and learning - requires a phased, outcome-driven strategy. Organisations must map legacy dependencies, prioritise high-impact low-risk processes, design human-AI collaboration, and embed learning feedback loops in every release.This pragmatic approach reframes modernisation from a project to a discipline, and from legacy avoidance to future enablement. It underscores how a civil service transformed workflow by workflow can become more efficient, adaptive, and human-centred.

As artificial intelligence shifts from hype to enterprise-scale reality, many organisations still struggle to move beyond pilots into sustained value creation. According to the UK’s tech industry, 72% of firms had experimented with generative AI by early 2024 - but nearly two-thirds were still unable to scale effectively.In this guest blog for techUK, Edward Howes, UK Consultant at UBDS Digital, presents five actionable tips designed to help organisations make the leap from isolated experiments to enterprise-wide AI adoption. He makes it clear that this journey isn’t just about deploying more tools - it’s about fundamentally transforming how work gets done.For companies ready to scale AI, the key takeaway is clear: aligning tools with strategy, governance, people and metrics will determine whether AI becomes a strategic enabler or a stalled pilot.

This Computer Weekly feature explores why Small Language Models (SLMs) are gaining traction as a practical alternative to large, resource-intensive AI for businesses seeking real-world advantage at the edge. With rising infrastructure costs, security concerns and the limitations of cloud-centric large language models (LLMs), SLMs promise lower cost, enhanced privacy and the ability to run AI intelligence closer to where data is generated - on devices, in factories, or directly within operational systems.Silvia Lehnis, Consulting Director for Data and AI at UBDS Digital, underscores that the biggest strategic benefit of SLMs lies in their ability to operate locally without constant cloud dependency. Silvia highlights that this not only cuts energy and operational costs but also reduces the AI carbon footprint and keeps sensitive data within organisational boundaries - a key advantage for security and privacy-sensitive use cases. She points out that running models locally can shift how enterprises think about AI deployment: from centralised, one-size-fits-all approaches to more focused, cost-efficient and context-relevant intelligence.Read more about Silvia’s opinion and how SLMs could reshape AI strategy in the full article.

In Computer Weekly’s Microscope opinion piece, Silvia Lehnis tackles a critical question for the UK channel in 2025: as AI transforms industries and the government backs ambitious plans for national AI leadership, what role should channel partners play in turning strategy into real business impact? The article outlines how uneven readiness across organisations from data infrastructure gaps to talent shortages means that ambition alone won’t deliver outcomes without practical guidance and capability building.Silvia Lehnis, Consulting Director for Data & AI at UBDS Digital, argues that the channel is uniquely positioned to help organisations bridge the gap between AI potential and measurable value. She emphasises that success requires more than technical investment - it demands AI literacy, clear use cases and structured steps that help organisations adopt AI incrementally and responsibly. Silvia highlights that many businesses either don’t know where to begin or lack the skills and data governance needed to scale AI beyond pilots. By guiding clients through data management, practical use cases, and capability building, channel partners can help ensure AI investments deliver meaningful outcomes rather than isolated productivity gains.Read more about Silvia’s perspective and her views on how the channel can unlock AI leadership in the full article.

Too many organisations find themselves stuck in an endless cycle of AI proof-of-concepts - promising experiments that never make it into real-world use. In this techUK article, Arun Manoharan, Global Head of Strategy Enablement at UBDS Digital, explores why AI initiatives stall and what businesses must change to break free of the PoC trap. He argues that the issue isn’t a lack of innovation, but a lack of strategic alignment, governance, and scalable foundations.Arun outlines how reframing PoC's as value-driven pilots, grounding them in measurable outcomes, and building the right data, platform and operating model capabilities can move AI from experimentation to production - unlocking impact that lasts.

In a world where AI is reshaping how we work, leaders must evolve just as fast as the technology around them. In this guest blog for techUK, Arun Manoharan, Senior Strategy & Innovation Consultant at UBDS Digital, outlines how successful leadership in the AI era blends new operating models, practical technology adoption and continuous learning frameworks.He emphasises the need for leaders to set up structures that prioritise innovation, engage internal and external stakeholders, and adapt beyond legacy systems—so that AI becomes a driver of productivity, wellbeing and customer satisfaction, not just a novelty.