Companies are investing large amounts of money in AI upskilling, with workshops, bootcamps, internal certifications, "AI literacy" programs, and others;
the goal is rational: if AI will change how we work, employees need new skills;
However, the issue is not with the training itself.
It's the underlying assumption about it.
Most organizations assume that developing the necessary skillset directly leads to improved job performance. It doesn't.
The Transfer Problem
A well-known body of research exists within the field of organizational psychology around the topic of "training transfer."
This body of research has consistently demonstrated over several decades that a very low percentage (typically estimated to be in the 10-20% range) of training results in actual on-the-job behavior change unless the surrounding systems in which the employee works remain the same.
Employees attend classes.
- They learn the material.
- They may even be motivated.
- Then they go back to their roles.
- Their approvals.
- Their KPIs.
- Their incentive plans.
- Their reporting requirements.
And their job performance returns to what it was before.
Not because employees are unwilling to learn.
Because the system in which they operate is.
AI Skills Without Structural Landing Zones
AI tools can greatly improve drafting, analysis, ideation, modeling.
However, if the employee role:
- Does not grant the employee authority to implement the results of using an AI tool.
- Does not provide incentives for speed or experimentation.
- Does not change the employee KPIs to account for automation gains.
Then the employee's AI skills sit idle.
The organization views this as "low adoption," but in fact, it is simply a matter of misaligned systems.
Structural landing zones for skills are required.
Unless the workflow is changed, the skill cannot be compounded.
Why Leadership Misreads the Signal
From the executive level, the reasoning appears logical:
We trained our employees.
Low usage of the AI tools.
Unclear ROI.
Therefore, we conclude that AI impact is being overestimated.
That conclusion omits a key piece of information.
AI capabilities inside a stagnant structure create resistance.
Accelerating through a slow governance model creates tension.
Tension often is referred to as failure.
It is not failure.
It is architectural misalignment.
Capability vs. Skill
An important difference is overlooked by many organizations.
Skill is an individual characteristic.
Capability is a systemic feature.
You can train 500 employees in prompt engineering.
But if your approval processes continue to presume that each review cycle takes weeks to complete, your pricing continues to include labor costs from previous manual efforts, and your leadership culture continues to punish employees for taking risks, then you have not created AI capability.
You've created awareness.
Capability requires:
- Role changes
- Workflow adjustments
- Recalibrations of incentives
- Clarified authority boundaries
Until those changes occur, training is merely symbolic.
The Real Risk
The risk is not that AI training doesn't work.
The risk is that organizations will believe they have successfully "done AI" based on the number of sessions they held.
That creates a false sense of security.
While competitors are redesigning their systems to accommodate the shift toward AI-enabled workflows, the gap grows.
It is not due to the competitor having better-trained employees.
It is due to the competitor integrating capability into its operational model.
What Actually Works
When employees' roles are defined around enhanced workflows that incorporate AI tools, KPIs reflect the new productivity standards, decision rights are defined to enable rapid iterations, managers are trained prior to training their employees, and when the organizational structure is redesigned before training occurs, employee adoption of AI tools is successful.
In other words, when structural changes occur first and training follows, employee adoption of AI tools is successful.
Training without structural redesign appears innovative.
Structural redesign without training appears chaotic.
Both need to happen simultaneously.
The Quiet Conclusion
Most AI initiatives do not fail because the technology is immature.
They fail because organizations attempt to embed AI into organizational structures that were designed for slower intelligence cycles.
AI does not fit into legacy models cleanly.
It compels organizations to consider issues of authority, tempo, and responsibility. That is unsettling.
So, instead of redesigning systems, companies opt for training employees.
It is easier to teach employees new skills than it is to redesign organizational systems.
Skills developed by employees that lack structural permission will not scale.
And that is why most AI training programs produce disappointing results.
—
#skillsonic #aiarchitecture #futureofcompanies #aitransition #futureofwork
by Eugene Baiste, AI Capability Architect at Skill Sonic
.avif)
.avif)
.avif)
.avif)
.avif)