The United Kingdom doesn't have an artificial intelligence adoption problem; it has a training crisis. As companies rush to deploy AI agents across operations, a critical gap has emerged: the people managing these systems lack the skills to do so effectively. New research from Salesforce and Focaldata surveyed 1,430 UK workers and found that 61% of employees have received no formal AI training from their employer, even as autonomous agents begin handling customer service, marketing, and business operations.

This training deficit comes at a pivotal moment. Dreamforce 2026, Salesforce's annual conference, is showcasing what executives are calling the "Agentic Enterprise," where AI agents handle increasingly complex work across every business function. Yet the infrastructure to prepare workforces for this shift remains largely absent. The gap between AI deployment speed and employee readiness is widening, creating a hidden risk for organizations betting on agent-driven transformation.

Why Is AI Training Lagging So Far Behind Deployment?

The disconnect between AI adoption and workforce preparation reflects a broader pattern in enterprise technology. Companies are investing heavily in AI infrastructure, tools, and platforms, but treating training as an afterthought. This mirrors earlier waves of digital transformation, where organizations purchased new systems without adequately preparing teams to use them effectively.

The challenge is compounded by the novelty of agentic AI. Unlike traditional software training, which focuses on how to use a specific tool, AI agent training requires employees to understand how autonomous systems make decisions, when to trust them, and how to intervene when necessary. This demands a different skill set entirely. Marketing leaders, for instance, are being asked to become "A-Shaped" executives who combine deep domain expertise with AI fluency and data mastery, according to Salesforce research showing that 77% of marketing leaders are redesigning their departments for AI agents.

What Skills Do Workers Actually Need to Manage AI Agents?

As AI agents take on more autonomous work, the skill requirements for managing them have evolved significantly. The Agentic Enterprise Index data cited by Salesforce shows that AI skill sets are tripling as brands shift from basic task automation to multi-step workflows. This shift demands more than technical knowledge; it requires strategic thinking about how agents integrate with existing business processes.

The training gap extends beyond technical skills. Employees need to understand how to evaluate agent outputs, recognize when an agent has made an error, and know when human judgment should override automation. They also need to grasp data governance and security implications, since AI agents are only as reliable as the data feeding them. Without real-time, clean data, autonomous agents cannot deliver personalized customer experiences or make sound business decisions.

How to Build an Effective AI Training Program for Your Organization

• Establish a formal AI strategy with executive ownership: Organizations with a formal AI strategy are 3 times more likely to report measurable impact, according to Info-Tech Research Group's June 2026 survey of 551 senior leaders. This means designating clear accountability for training outcomes and linking them to business objectives.
• Create role-specific training pathways: Different departments need different AI competencies. Marketing teams need to understand how agents personalize customer conversations; operations teams need to know how agents optimize workflows; customer service teams need to learn when to escalate to human agents. Tailoring training to specific roles increases relevance and adoption.
• Invest in data literacy as a foundation: Since AI agents depend on clean, real-time data to function, employees need to understand data quality, governance, and how to identify data problems. Without this foundation, even well-designed agents will fail to deliver value.

Salesforce UKI CEO Zahra Bahrololoumi CBE acknowledged the urgency of this challenge, noting that while the UK continues to lead globally in AI innovation, the country faces a critical skills bottleneck. The research underscores that formal training programs are not yet widespread, leaving most organizations to rely on informal learning or vendor-provided resources.

What Happens When Companies Skip the Training Step?

The consequences of inadequate AI training are already visible across enterprises. Research from Gartner, Mavvrik, and Deloitte shows that AI budgets are shifting from training to operations, while surprise costs derail approximately one quarter of AI projects. This pattern suggests that companies are discovering hidden expenses only after deployment, often because teams lack the expertise to manage systems efficiently.

Additionally, 74% of enterprises run AI in production, but half cannot prove it pays off, according to analysis cited in MarketScale's coverage of enterprise AI trends. This ROI blindness is partly rooted in the training gap; without employees who understand how agents work and how to measure their impact, organizations struggle to quantify business value.

The training deficit also creates security and governance risks. Employees unfamiliar with AI agent behavior may inadvertently expose sensitive data or fail to recognize when an agent has been compromised. As enterprises deploy agents across customer-facing channels like email, WhatsApp, SMS, voice, and Slack, the stakes for proper training increase significantly.

Is the UK Alone in This Challenge?

While the Salesforce-Focaldata research focuses on the UK, the training gap appears to be a global phenomenon. Enterprise AI spending is maturing rapidly, with organizations shifting resources from experimentation to accountability. However, this transition is exposing gaps in governance, data readiness, and workforce preparation across markets. The difference is that the UK's research explicitly quantifies the problem, making it harder for organizations to ignore.

The timing of this research is significant. As Dreamforce 2026 showcases real-world examples of AI agents solving business problems, the conference also highlights what's required to make those solutions work at scale. Formula 1's fan-facing agent built for 831 million global fans, Xero's Agentforce resolving 60% of customer inquiries, and IBM's Slackbot-powered go-to-market operating system all represent success stories, but they also represent organizations with the resources and expertise to implement AI effectively.

For most companies, the path to that level of maturity requires closing the training gap first. Without a workforce equipped to manage, monitor, and optimize AI agents, even the most sophisticated technology will underperform. The UK's research serves as a wake-up call: adoption without preparation is a recipe for disappointment.