The push to evaluate agentic state tools on performance rather than marketing labels is gaining traction among UK public‑sector buyers. While the term ‘agentic AI’ is now common in policy circles, the practical question for procurement teams is how to compare a supplier’s method and results with the glossy claims on a brochure.
What the regulators say
The UK Information Commissioner’s Office (ICO) released a Tech Futures report that maps the technical evolution of agentic AI and flags data‑protection risks for organisations planning deployments. The report does not prescribe procurement check‑lists, but it stresses that “understanding its capabilities and the associated risks is essential” for any public body considering adoption. The ICO’s focus on privacy and security implies that buyers need evidence of robust data‑handling practices, not just a vendor’s marketing narrative.
At the same time, the Competition and Markets Authority (CMA) published a research paper on 9 March 2026 that examines how agentic AI may affect consumers. The paper notes that existing UK consumer‑protection law applies to AI‑driven services and that breaches could attract fines of up to 10 % of global annual turnover. While the CMA does not issue a procurement‑specific guideline, its emphasis on “misleading practices delivered through an AI agent” signals that buyers should demand demonstrable outcomes and transparent methodology.
Industry commentary
Legal‑tech consultancy ReedSmith highlighted the uptick in regulatory activity in a March 2026 blog, pointing out that both UK and EU regulators are interpreting existing frameworks – competition, data protection, cybersecurity – for autonomous AI systems. ReedSmith’s analysis suggests that public‑sector buyers will soon need to provide evidence that an agentic system’s decision‑making process complies with these frameworks, moving the focus from brand reputation to measurable performance.
Elsewhen’s own report frames the broader transformation: agentic AI can shift public‑service workflows from manual, hours‑long tasks to seconds‑level operations, freeing civil servants for judgment‑heavy work. The report gives the example of “a hundred planning records per day, versus the usual average of five,” illustrating the scale of efficiency gains. However, it stops short of prescribing how buyers should vet the underlying methods that deliver such gains.
Implications for procurement
For procurement officers in ministries, local councils and NHS trusts, the emerging guidance translates into a few concrete steps. First, request detailed technical documentation that explains the AI’s decision‑making pipeline – data sources, model architecture, validation procedures – rather than accepting a high‑level capability claim. Second, ask for pilot‑phase performance metrics that can be benchmarked against existing service levels. Third, verify that the supplier’s data‑protection impact assessment aligns with ICO expectations.
These steps echo the broader regulatory thrust: accountability and transparency over brand promises. By demanding method‑level evidence, buyers can better assess whether the promised “seconds‑level” processing will hold up under real‑world data loads and legal scrutiny.
What remains unknown
Despite the regulatory focus, no official UK procurement guideline currently mandates a specific comparison framework for agentic AI. The ICO, CMA and other bodies have highlighted risks and expectations, but the exact criteria that a public‑sector tender should include are still being debated. Buyers therefore face a knowledge gap: they must balance the urgency of digital transformation with the lack of a standardized evaluation rubric.
Stakeholders are watching for forthcoming updates from the Government Digital Service and the Department for Business, Energy & Industrial Strategy, which are expected to publish more detailed procurement advice later in 2026. Until then, public‑sector buyers are advised to treat the regulator‑driven risk signals as a de‑facto requirement to move beyond brochure branding.
For further reading on the regulatory landscape, see the ICO’s Tech Futures report (ico.org.uk) and ReedSmith’s analysis of the CMA paper (reedsmith.com). The UK government’s overview of agentic AI and consumer implications can be found on GOV.UK (gov.uk). For practical guidance on drafting AI‑focused procurement specifications, see our internal guide to AI procurement (newdispatch.com/articles/ai-procurement-guide) and the recent briefing on digital transformation in public services (newdispatch.com/insights/agentic-ai-regulation).