Enterprise AI Readiness Gap 2026: Kyndryl Finds 57% Have AI Embedded but Only 23% Say Their Workforce Is Ready

Kyndryl's 2026 People Readiness Report of 1,100 leaders found AI deployment up sharply while workforce readiness fell six points and only 11% hit both top AI goals. The key figures, what the 9% of Pacesetters do differently, and practical steps to close the gap.

PUBLISHED // UPDATED
DURATION // 06 MIN READ
A stark split-image composition: one side shows a glowing, modern office with digital dashboards and AI interfaces; the other side shows workers looking at screens with confused expressions, the lighting cooler and less polished
FIG.01 // VISUAL DOCUMENTATION A stark split-image composition: one side shows a glowing, modern office with digital dashboards and AI interfaces; the other side shows workers looking at screens with confused expressions, the lighting cooler and less polished

Enterprise AI deployment and workforce readiness are moving in opposite directions. Kyndryl’s second annual People Readiness Report, a June 2026 survey of 1,100 senior business and technology leaders in eight countries, found that 57% of organisations now have AI embedded in core processes or deployed broadly, yet only 23% say their workforce is fully ready for AI, down six points from 2025 (Kyndryl press release). Just 11% have achieved both of their top two AI goals. The organisations getting results are not the ones with the most tools; they are the ones that redesign roles, manage change deliberately and build governance as they go.

The key figures

FindingFigure
AI embedded in core processes or deployed broadly57% (last year, 35% said AI was fully integrated)
Achieved at least one of top two AI goals32%
Achieved both top two AI goals11%
Say workforce is fully ready for AI23%, down 6 points from 2025
Agree AI’s speed will outpace their workforce, governance and operating models79%
Expect AI agents to make impactful decisions within a year81%
Completely trust AI operating without human oversight25%
Have already redesigned roles61%
Creating new AI management roles24%
Find it harder to hire the right skills for AI52%

Source: Kyndryl 2026 People Readiness Report, as summarised in the company’s press release. These are self-reported survey results from a vendor-commissioned study.

Kyndryl’s release also cites Gartner’s forecast that worldwide AI spending will total $2.52 trillion in 2026, a 44% year-over-year increase. Capital is flowing into tools and infrastructure much faster than organisations are preparing people to use them.

The employee view is even weaker

MarketScale’s July 2026 analysis set Kyndryl’s leadership survey beside the Achievers Workforce Institute’s State of Recognition Report, as cited by Inc. That report found only 19% of workers feel confident using AI tools and 18% feel supported in adapting to them. MarketScale also reports from Kyndryl’s data that only about a third of organisations have fully implemented training for working alongside AI, 33% have clear policies on which decisions AI can and cannot make, and 27% use registries and monitoring across all AI systems.

Taken together, the surveys describe a familiar pattern: leadership announces AI initiatives, tools land on desks, and employees are left to work out when to trust them.

What the Pacesetters do differently

Kyndryl identifies a group it calls Pacesetters, about 9% of respondents, defined by three behaviours:

  1. They redesign roles around AI rather than layering AI onto existing jobs.
  2. They implement change management so employees understand the new operating model and the guardrails around it.
  3. They build workforce readiness through upskilling and retraining.

According to Kyndryl, Pacesetters are roughly twice as likely to have fully implemented every governance dimension measured, 1.5 times more likely to achieve AI-related revenue growth and 1.6 times more likely to report better product and service innovation. These are correlations in a survey, not proof that the behaviours cause the results, but they are consistent with what many transformation programmes have found: technology alone rarely changes outcomes.

“The organizations investing in people – whether it’s rethinking roles and workflows, dedicating resources for upskilling and retraining, or guiding employees through change – are experiencing positive outcomes at a much higher rate,” said Kim Basile, Kyndryl’s CIO.

Why readiness falls as deployment rises

Tools arrive before rules

When AI is added to existing processes without redesigning them, employees face new judgment calls with no guidance. Consider a customer service team given an AI call summariser. If an agent does not trust the summary, they re-listen to the call and the tool adds a step. If they trust it blindly, errors flow downstream. The readiness gap lives in that moment of judgment, and generic training on tool features does not address it.

Governance has not caught up with agents

With 81% expecting AI agents to make impactful decisions within a year but only 25% fully trusting unsupervised AI, organisations are heading toward autonomy without the controls to support it. Clear decision rights, meaning what AI may decide, what needs human review and what gets escalated, are an operating model question, not a software setting. Our guide to governed AI coworkers looks at how vendors are building these controls into agent platforms.

Success is measured by deployment, not outcomes

Licence counts and launch announcements are easy to report; achieved business goals are not. With only 11% hitting both top AI goals, the distance between “deployed” and “delivering” is large. We explore the same gap in why near-universal AI adoption hasn’t translated into transformation.

How to close the gap

ActionWhat it looks like in practice
Measure adoption depth, not deploymentTrack active use, task completion and quality, not licences issued
Redesign workflows before rolloutMap the decision chain, from data to AI output to human review, for each process
Define what “ready” meansSet acceptable output quality, healthy override rates and tolerable error rates for each use case
Write decision-rights policiesSpecify which decisions AI may make, which need review and how escalation works
Make AI costs visibleTrack spend per use case before expanding; FinOps practices for AI help (FinOps Foundation)
Build feedback loopsGive front-line users a channel to report errors and workarounds to the team that chose the tool
Invest in role-specific trainingTeach evaluation and override judgment in context, not just features
Register and monitor AI systemsKeep an inventory of AI in use with monitoring and audit logs

A practical sequence is to pick a few high-value workflows, redesign them with the people who do the work, set baselines, deploy with clear decision rights and measure outcomes for a quarter before scaling. For the broader shift toward agent-driven operations, see our overview of the agentic enterprise.

Our analysis

Kyndryl’s figures come from a services company that sells transformation help, so its emphasis on people and change management is not disinterested. Even so, the pattern in its data is credible and matches the employee-side surveys: AI deployment has outrun the redesign of roles, policies and skills needed to use it well. A six-point fall in readiness during a year of rapid rollout suggests that adding tools without changing how work is done erodes confidence rather than building it. The practical lesson is to treat workforce readiness as part of the deployment itself, with redesigned workflows, explicit decision rights, measured outcomes and real feedback from the people using the tools, rather than as training scheduled after launch.