The Skill Set Learning Teams Need for an Agile Organization
Edmundo Barrientos · 2026-09-02 · 3 min read

TL;DR: McKinsey finds agentic AI stalls for organizational reasons, not technical ones. That makes it a learning problem first. It takes four new skills inside the learning team, and none is producing more courses.
The blocker is the organization, not the LLM
If agentic AI stalls for organizational reasons, the group whose job is changing how people behave at work owns the problem. Most learning teams are not set up to take it.
That premise is McKinsey's. Its August 2026 analysis of agentic AI change management finds technically ready programs still stalling, blocked by siloed teams and unmanaged change rather than technology, and prescribes change leadership. Ours is narrower; capability adoption, not technology adoption, is what decides an AI program, and capability is our work.
Agility is a cycle time, not an attitude
"Agile organization" gets described as a culture. In practice it is a number: how long it takes people to get good at a process after it changes.
Standard development cycles take months: identifying needs in January, designing curriculum in March, launching the first cohort in May. Then an agent starts drafting every claim summary, and the adjuster's job changes the week it ships. A four-month cycle is not just slow, it is obsolete.
Four skills that shorten the loop
Each deletes a phase of the slow version.
Read a process, not a request. "We need AI training for the claims team" is a request. Walk the process instead to find the step a person still owns: the adjuster who now approves or overrides the agent's draft. That step is where the ROI lives.
Diagnose from observed work, not surveys. Read twenty recorded calls or agent-assisted tickets today and note what broke. A survey round tells you how confident people feel; the work tells you where they struggle.
Design a repetition, not a module. The new hire who aced the objection-handling module and froze on their first live call did not need more content. They needed ten tries at that opening, under real pressure. A repetition ships in days; a module takes months.
Instrument the step. Escalation rate, rework rate, days to ramp. Each already sits on the process, so there is nothing to build. You cannot plan a capability you do not measure.
Why this is not a bigger catalogue
All four point away from producing content, which is fortunate; content has never been the bottleneck. Your docs, calls, and recordings already contain the training; the work is turning them into practice on the steps that just changed, in leadership, compliance, onboarding or customer success. The catalogue still does its job, it is just not where agility comes from.
A learning team that runs that loop in weeks is what makes the rest of the organization agile. One that can only publish courses will keep arriving after the process has moved again.
Related reading
- Augmentation Over Automation: Why AI Transformation Succeeds When Organizations Build Human Capability
- No Process, No ROI: Aligning Skill Development to the Work It Serves
- You Cannot Plan Capabilities You Do Not Measure
References
- McKinsey & Company. Agentic AI change management: Closing the adoption gap, August 2026.
About the publisher: this article is published by Grasply (grasply.ai), a capability builder. Grasply turns what a company already knows, its docs, calls and recordings, into skills people use at work, for employees, partners and customers. Practice turns knowing into doing, and Grasply shows proof that the work changed.