The enterprise AI skills gap is landing on CIO desks with new urgency in 2026, and the headline finding from three major research efforts this cycle is blunt: training budgets are not the fix. Deloitte, Gartner, Thoughtworks, and IDC have all published data in roughly the same window, and each one draws the same structural conclusion - organizations are pouring money into AI literacy while leaving the work itself largely unchanged.
Deloitte AI Institute, State of AI in the Enterprise 2026, 3,235 global leaders across 24 countries
| Value (%) | Share of enterprises |
|---|---|
| Increased AI spending | 85% |
| Targeting revenue growth | 74% |
| Actually achieving revenue growth | 20% |
| Have NOT redesigned jobs | 84% |
| Training-only talent strategy | 53% |
| Redesigned career paths and workflows | 30% |
What the data actually shows about AI investment and returns
Deloitte's AI Institute published its eighth annual State of AI in the Enterprise report in January 2026, drawing on 3,235 business and IT leaders across 24 countries. The headline tension is stark: 85% of enterprises in the survey increased AI spending, and 74% list revenue growth as a primary goal. Only 20% have actually achieved it. Insufficient worker skills ranked as the single biggest barrier to integrating AI into existing workflows - ahead of data quality, infrastructure, and cost.
The finding echoes a May 2025 Gartner survey of 506 CIOs, which found that 72% report their organizations are breaking even or losing money on AI investments. Gartner separately projects that all IT work will involve AI by 2030, meaning the pressure on returns is not easing anytime soon. IDC research cited by CIO.com adds a readiness dimension: 94% of leaders consider AI skills essential for this year, yet only about one third believe their organization is actually ready to deliver on that.
53% of enterprises are adjusting talent strategy through AI literacy training only - while just 30% have redesigned the career paths and workflows those employees work inside.
The workflow redesign gap that training cannot bridge
Here is where the Deloitte data gets particularly damaging for the training-first orthodoxy. Despite ranking skills as the top barrier, 84% of enterprises surveyed have not redesigned roles or workflows around AI capabilities. At the same time, 82% expect at least 10% of jobs to be automated within three years. That combination - expecting automation while leaving processes and roles unchanged - is the structural failure that training courses cannot resolve on their own.
Deloitte found that 53% of enterprises are adjusting their talent strategy through AI literacy training only, while just 30% have redesigned career paths and workflows. The sequencing problem is clear: employees trained on AI tools are being dropped back into approval chains, handoff points, and role definitions that predate the technology. Under those conditions, even skilled employees cannot generate the returns leadership expects.
CIOs are accountable for a problem they cannot fully control
The accountability dimension is where the Thoughtworks data adds something the other reports do not. Its October 2026 global CIO survey found that 89% of CIOs now say they are more responsible for workforce redesign than for core IT infrastructure - a significant shift in how the role is being defined. More pointedly, 35% say they feel personally accountable for workforce disruption from AI even though they cannot fully influence the outcome.
That combination of accountability without authority is a familiar management failure mode, and it maps directly onto what CIO.com has been calling out in its recent commentary on the Deloitte findings. CIOs are being handed ownership of an AI talent problem that was generated by business-unit decisions and HR policy constraints that IT does not control. The result is that IT leaders absorb reputational risk from an ROI shortfall that is structurally baked in from the start.
Reframing AI skills as an operating-model problem
The practical mandate that emerges from this research cycle is a reframing exercise. The AI skills conversation needs to move from an L&D line item to a workflow and operating-model redesign initiative, owned jointly between IT, business units, and HR rather than delegated to any one function. That means auditing the processes that AI tools are being deployed into before expanding training programs, and asking whether the approval chains, data flows, and role handoffs were built to accommodate AI-generated outputs at all.
CompTIA's State of the Tech Workforce 2026 reinforces the workforce planning dimension: technology roles are being redefined faster than hiring pipelines or internal reskilling programs can track. CIOs who treat this as a training throughput problem will keep producing the same 20% ROI ceiling. Those who push for process redesign authority alongside skills budgets have a materially different set of levers to pull.
