There is now hard measurement behind what many business leaders have suspected: the workforce is not ready for AI agents. Not even close.

Workera published its 2026 AI Skills Enterprise Benchmark Report last week, drawing on more than 88,000 individual assessments across major enterprises and the US federal government. The report measured capability across 14 distinct AI and data competencies. Agentic AI came dead last — only 13% of employees tested as “Accomplished” before any training intervention.

That number is not a rounding error. It is the lowest benchmark of any capability in the entire study.

What the Data Shows

The report paints a detailed picture of where enterprise workforces actually stand on AI.

At the top of the skills rankings sit areas where workers can transfer existing strengths into the AI era: Data Storytelling Essentials scored 231 out of 300, AI and Data Communication scored 230, and Responsible AI Essentials scored 229. These capabilities draw on communication, reasoning, and judgment — things experienced professionals already have. The tools are new; the underlying muscles are not.

At the bottom sit the capabilities that are genuinely new to most workers: Deep Learning Fundamentals scored 163 out of 300, and Beyond LLMs: Prompts, Agents, and Retrieval-Augmented Generation (RAG) scored 185. Agentic AI — the ability to understand, configure, and work alongside autonomous AI systems — sits at that same bottom tier.

The pattern is consistent across industries. Workforces are strongest where familiar skills apply to new tools. They are most underprepared where the technology itself is newest.

Why Agentic AI Is the Hardest Gap to Close Unassisted

AI agents are not just chatbots with extra steps. They make decisions, take actions, call external systems, and operate across extended timeframes without a human in the loop at every step. Working effectively with them requires understanding how they reason, where they fail, what kinds of tasks to hand off versus keep, and how to set constraints that prevent costly errors.

None of that is intuitive. Unlike typing a better prompt or using a spreadsheet function powered by AI, working with agents requires a different mental model entirely. It is closer to managing a contractor than using a software tool — and most employees have never been trained to think in those terms.

The 87% of enterprise employees who do not yet have these skills are not falling behind because they are disinterested or resistant. They are falling behind because no one has given them a clear path to get there.

Upskilling Works — When It Is Structured

The Workera data contains an encouraging finding: targeted upskilling moves the needle dramatically. Take Responsible AI as an example. Before training, only 25% of employees score as Accomplished. After a structured upskilling program, that number rises to 81%. A gap that looks almost impossible to close turns out to be entirely closeable with the right approach.

There is every reason to expect similar results for agentic AI skills — but only if organisations invest in structured, measurable development rather than relying on employees to figure it out themselves. Watching a YouTube video or clicking through a vendor demo is not the same as developing genuine capability.

The companies that are moving fastest on AI agents right now are not necessarily the ones with the biggest technology budgets. They are the ones that treated workforce capability as a serious investment alongside tooling.

What This Means for Business

If 87% of your employees lack the skills to work effectively with AI agents, your AI strategy has a ceiling — and it is lower than your tech stack implies.

Most organisations have already spent on licences, platforms, and tools. The productivity gains they expected have been slower to arrive than the sales pitch suggested. This report provides a straightforward explanation for the gap: you cannot extract value from AI agents if your workforce does not know how to use them well.

The implication is not that companies should slow their AI adoption. It is that capability development needs to run in parallel with deployment — not as an afterthought, not as a one-hour onboarding session, but as a real competency-building investment with measurement behind it.

A few practical steps worth taking now:

Assess before you assume. Most leaders dramatically overestimate their team’s AI readiness. Run skills assessments to find out where the gaps actually are, not where you hope they are not.

Prioritise agent literacy over tool familiarity. Knowing how to use Copilot or ChatGPT is not the same as understanding how to work with autonomous agents. The latter is more durable and more valuable.

Build in milestones. The Workera data shows that structured training produces measurable gains. Map out what “accomplished” looks like for your team and track progress toward it.

Connect skills to roles. Generic AI training tends to be forgotten. Training that maps directly to how a specific role interacts with agents — in your actual context, on your actual workflows — sticks.

---

The 13% number will improve. The question is whether it improves because organisations built deliberate capability development programs, or because a handful of self-motivated individuals figured it out on their own while the rest of the workforce fell further behind.

For organisations that want to close this gap faster, Enterprise DNA’s learning platform includes structured courses in AI literacy, data skills, and the applied capabilities that sit just above where most enterprise workforces are today. For teams looking to upskill at scale, the business plan provides structured paths and progress tracking across your organisation.