Why AI Transformations Fail Without Learning and Development
Edmundo Barrientos · 2026-11-04 · 3 min read
TL;DR: AI initiatives often succeed in testing but fail to transform organizations. The missing link is enabling employees to master new workflows, a Learning & Development challenge rooted in practice.
The Monday Announcement
Dana manages a six-person team that drafts client reports manually. On Friday, leadership hands her a slide detailing a new process: starting next month, an AI application will generate report drafts, and her team will transition to reviewing them.
Tasked with breaking the news on Monday, Dana has the presentation deck but lacks the right words.
When she reads the slide to her team, they take away one message: their jobs are disappearing. When an employee asks who is accountable for potential data errors, Dana has no answer.
Fast forward three weeks: half the team manually rewrites every AI draft from scratch, while the other half approves reports without reviewing them. Despite a successful pilot, the team's actual work practices remained completely unchanged.
The last mile is people, not software
Karim R. Lakhani, Jared Spataro, and Jen Stave identified this exact issue in Harvard Business Review. As they put it: "The primary obstacle to progress is rarely model quality or data availability, but rather the 'last mile' of transformation where technical capability must meet organizational design."
While they outline seven specific frictions, we argue that three key actions they recommend are fundamentally Learning & Development initiatives:
- Redesigning work from scratch: Evaluating whether existing workflows would even be built under an AI-first framework today.
- Documenting expert intuition: Codifying and sharing institutional knowledge rather than hoarding it.
- Transitioning to oversight roles: Enabling teams to evaluate and guide AI output rather than executing every task manually.
Simply detailing a process change on a slide does not build new habits; habits require active practice before launch.
Core Needs of Dana's Team
1. Practice for the Redesign: A process redesign is only real when people execute it. Prior to Monday, Dana needed a low-stakes environment to rehearse navigating the difficult conversation with her team.
2. Scaling Expert Judgment: Critical institutional knowledge was isolated in a single individual. Ravi, a senior analyst, could identify errors instantly. While his insights served as informal training, the team had never turned them into structured practice.
3. Building Oversight Skills: Shifting into an oversight role demands a brand-new skill set, yet no one had been trained on how to effectively interrogate and review AI-generated drafts.
To bridge these gaps, Dana's L&D partner designed realistic, practical exercises grounded in the company's own context. Dana practiced delivering the announcement against an AI role-playing a skeptical team member who pushed back with tough questions. Meanwhile, the team trained on draft reports embedded with errors Ravi had previously flagged. These targeted simulations focused on real-world application. They are not virtual reality: no headset needed.
While the AI tool generated the baseline scenarios, the L&D partner refined them and determined the key skills to emphasize in practice.
What it proves
Dana's second attempt yielded a completely different result because she was already prepared for the tough questions. Her team also identified and corrected Ravi's errors during practice, ensuring the subsequent drafts contained no surprises.
This same pattern applies to other organizational shifts, such as a hospital updating its shift handover process: technical tools are consistently ready long before the workforce.
This dynamic redefines the core value of L&D. While AI can generate content, producing content is no longer the primary objective; building capabilities is. Practice is where knowing turns into doing. Translating Ravi's expertise into active habits across his team represents true capability building. It directly addresses the knowing-doing gap and serves as the essential final phase of any successful AI implementation.
Related reading
References
- Harvard Business Review. The "Last Mile" Problem Slowing AI Transformation, 2026. By Karim R. Lakhani, Jared Spataro, and Jen Stave.
- Harvard Business School AI Institute. Why Your AI Strategy May Be Failing, 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.