UK Races Ahead with AI Agents, Leaving Governance Behind
UK organisations are rapidly deploying agentic AI, but the frameworks to manage them are not keeping up. A new report from Salesforce highlights a growing gap between the speed of AI adoption and the governance needed to control it, creating significant operational risks.
This creates a siloed and poorly orchestrated system that demands immediate attention from leadership.
The Scale of Adoption Reveals a Deeper Problem
Salesforce's 2026 Connectivity benchmark report found that 89% of UK and Ireland organisations already use AI agents. The average organisation has 12 agents, a number projected to climb by 67% over the next two years.
However, half of these agents are isolated in departmental "silos." They aren't connected to the wider enterprise, which limits their value. This points to a tactical, rather than strategic, approach to AI deployment.
The Governance and Infrastructure Gap
The core issue is a lack of central oversight. Only 54% of organisations have a formal governance framework for their AI agents. This lack of control is a major concern, with 75% of IT professionals worried that agents will introduce more complexity than business value.
Outdated IT is a primary barrier. Some 97% of organisations face obstacles in using their data for AI, with 35% identifying disconnected systems and data silos as the top blocker. Almost all respondents (94%) agree that agentic AI requires new, more API-driven IT architectures to work effectively.
The Danger of "Shadow AI"
The ease of creating AI agents introduces a new risk: shadow AI. When individual teams build their own tools without oversight, it can lead to serious security flaws, poor reliability, and uncontrolled costs. Kurt Anderson of Deloitte Consulting warns against letting contributors build systems that lack security or run up large bills.
The old top-down governance model is too slow for AI. Instead, the focus must be on creating "guard rails" that allow for innovation while ensuring transparency and security. For more on official guidelines, the OECD AI Policy Observatory provides a global perspective on principles and practices.
A New Foundation for AI Success
Success with agentic AI isn't about the number of agents, but how they are discovered, governed, and orchestrated to work together. Andrew Comstock of MuleSoft suggests IT's role must evolve from managing silos to building a unified foundation that acts as a central control plane.
This is supported by research from Celonis, which found 81% of executives believe AI projects will fail without clear visibility into the business processes they are meant to improve. You cannot automate a process you do not understand.
Focus on Value, Not Just Technology
When governed correctly, the benefits are clear. Another Salesforce report shows that top salespeople are 1.7 times more likely to use AI agents to hit their targets. They expect agents to cut time spent on research and email drafting by 38%.
"We want to kill the busy work so our teams can focus on what moves deals forward," said Paul O'Sullivan, CTO at Salesforce UK&I. For the public sector, this means freeing up staff to focus on mission-critical work and citizen services, not paperwork.
Building this new operational model requires a deep understanding of process and technology. Developing skills in AI automation is no longer optional; it's a core competency for any modern organisation.
Your membership also unlocks:
