One part of the problem is that employees can learn how a tool works, how to write a prompt, or which policies to follow without experiencing what it is like to use AI in a realistic workplace situation.
As AI becomes part of more day-to-day work, that missing layer can make or break business impact. Employees need opportunities to complete realistic tasks, encounter imperfect outputs, make decisions, and learn from feedback. Otherwise, organizations risk investing in training that does not change how people work and leaving employees to learn through trial and error when inaccurate information can reach a client, shape a decision, or move further into the business before it is caught.
Why AI Training Is More Complicated Than It Looks
The gap between instruction and application is especially consequential with AI. Training can introduce AI tools and policies, but employees still have to apply that information within work that is complex, changing, and difficult to predict.
Three challenges make practice especially important:
1. AI Creates Situations That Training Cannot Anticipate
AI does not always fail in obvious ways. It can produce an answer that appears polished and credible while leaving out important information, misunderstanding the context, or getting the answer wrong. No policy or course can prescribe exactly what an employee should do in every one of these situations. People must be able to interpret what is in front of them and determine the right course of action.
This remains true even when organizations use specialized tools designed for high-stakes work. Stanford researchers evaluated three leading AI legal research products that retrieve information from authoritative legal databases. Despite those safeguards, the tools produced hallucinated information in 17% to 33% of tested responses.
The stakes are growing alongside adoption. Documented AI incidents increased from 233 in 2024 to 362 in 2025—a 55% rise in one year. The database is not a complete record of every AI-related incident, and increased awareness may contribute to higher reporting. Still, the trend shows how quickly exposure can grow as AI becomes embedded in more products, decisions, and workflows.
Handling these situations requires employees to be prepared to respond when an output is imperfect, the available information is incomplete, or the right course of action is not obvious.
2. AI Removes Some of the Work That Builds Expertise
Employees have traditionally developed capability through experience. They complete foundational tasks, observe how others approach the work, make lower-stakes decisions, and gradually take on greater responsibility.
AI can change that process. An employee who once developed expertise by drafting, analyzing, reviewing, and revising work may now begin with a completed first draft from AI. The task happens faster, but the employee may have fewer opportunities to understand how the work was produced and develop the knowledge needed to assess whether the result is any good.
This does not mean organizations should preserve routine work simply for the sake of learning, but it does mean they need to be more deliberate about replacing the experience that automation removes. Otherwise, employees may be asked to evaluate work they have had limited opportunities to perform themselves.
3. AI Can Become Easier to Trust Than to Check
More experience using AI does not necessarily lead to more careful use. A Microsoft Research study found that greater confidence in AI was associated with less critical thinking. By contrast, workers who felt more confident in their own expertise reported evaluating AI outputs more critically.
The researchers did not establish that confidence in AI causes people to think less critically, but the findings point to an important risk: As employees become more comfortable with AI, its answers may begin to feel more reliable even when they still require careful review. Fifty-eight participants also reported barriers to checking AI responses, including insufficient subject-matter knowledge.
Therefore, employees need opportunities to encounter convincing but flawed outputs, test their assumptions, and learn when additional review or expertise is needed.
The Next Phase of AI Training Is Practice, Practice, Practice
Taken together, these challenges create a difficult learning problem. Employees need more experience applying AI, even as they may have fewer opportunities to build that experience through everyday work.
Broader research points to what is missing. A meta-analysis of 225 studies found that failure rates averaged 33.8% under traditional lectures and 21.8% under active learning. Put another way, failure was approximately 35% lower when learners actively engaged with the material.
Research on workplace AI use points in the same direction. Sixty-seven percent of workers who reported both productivity and quality gains from AI said trial and error on real work was a valuable source of learning. Among lower-performing AI users, that figure fell to 33%.
Both findings reinforce the value of applying what people learn. The question is how to give employees that experience without relying on live work alone.
Make Practice Part of the Talent Lifecycle
Forage Workspace Simulations give organizations a risk-free environment where talent can practice and refine new capabilities before applying them to work that affects business outcomes.
Each AI-powered simulation mirrors the flow of work through realistic assignments, workplace interactions, and personalized feedback. Simulations can be tailored to reflect an organization’s roles, workflows, expectations, and best practices, allowing participants to experience AI within the context in which they will actually use it.
Organizations can use Workspace Simulations across the talent lifecycle. Before hiring, candidates and emerging talent can gain exposure to AI-enabled work and better understand what a role involves. During onboarding, new hires can practice the workflows and expectations they will encounter before performing independently. Existing employees can continue building new capabilities as technologies, roles, and business needs evolve.
By bridging AI learning and live application, Workspace Simulations help organizations move talent from awareness to applied performance.
The Bottom Line: AI Training Has to Reach the Work
The first phase of workplace AI focused on access and adoption; the next must focus on application. Giving employees tools and information can encourage AI use, but giving them opportunities to practice is what prepares them to use AI well when it becomes part of the work itself.
If organizations want their AI investments to lead to better work—not simply more AI use—practice can no longer be the missing part of training. Learn how realistic workplace practice can prepare employees for AI-enabled work in our on-demand webinar,Building Workforce Readiness in an AI-Powered World.
Kayla Haskins is an Associate Director, Product Marketing at Seramount. In this role, she supports DEI Practitioners and Talent Leaders in creating more inclusive workplaces by providing valuable insights and resources through webinars, blog posts, guides, infographics, and more.
With nearly a decade of experience in the technology and non-profit sectors,
Kayla Haskins is an Associate Director, Product Marketing at Seramount. In this role, she supports DEI Practitioners and Talent Leaders in creating more inclusive workplaces by providing valuable insights and resources through webinars, blog posts, guides, infographics, and more.
With nearly a decade of experience in the technology and non-profit sectors, Kayla excels in translating complex ideas into clear, actionable concepts. She is passionate about storytelling and is dedicated to addressing today’s most pressing workplace issues to drive meaningful impact.
Kayla holds a degree in English and Creative Writing from Dickinson College. She lives in Silver Spring, MD with her partner, Nick, and their dog, Zero. In her free time, she enjoys hiking, reading, and spending time with family and friends.
