Employers Train Employees to Close the AI Skills Gap

When employees in the talent development group at consulting firm Booz Allen Hamilton began using generative AI (GenAI) tools in their work, they weren’t left to their own devices to learn how to master the technology. Learning and development practitioners in the group, who use GenAI for tasks such as creating training videos and summarizing notes from staff meetings, first had to complete their own comprehensive instruction and have use of the technology approved before applying GenAI on the job.

“Using GenAI has reduced our content production time by hundreds of hours as well as brought down production costs,” says Jim Hemgen, director of talent development for Booz Allen Hamilton. “But everyone who uses GenAI first has to [be] approved and certified to use it by completing internal training programs.”

The company’s multitiered AI Ready training initiative includes courses designed to teach all of its 33,000 employees the basics of using GenAI, as well as more in-depth, custom content for specific roles such as engineers and consultants. Baked into all of these courses is a focus on the responsible and ethical use of the tool, covering topics such as preventing the introduction of bias, ensuring the accuracy of GenAI outputs and avoiding the use of confidential company data in models.

“Our goal is AI readiness, which means ensuring everyone in our workforce is conversant in understanding ­GenAI’s capabilities, which includes using GenAI ethically and safely,” Hemgen says.

A Daunting Training Challenge

The meteoric rise of GenAI has created a vast need for reskilling across job roles and industries that has challenged HR and learning leaders to respond with agile training strategies that can keep pace with the rapidly evolving technology.

Many organizations have moved from merely dabbling in tools such as OpenAI’s ChatGPT, Google’s Bard and Microsoft Bing Chat to widespread workplace adoption. GenAI tools are being used in technical roles such as software development but also in jobs within HR, marketing and customer service departments. In response, organizations have had to teach employees not only how to master quickly changing GenAI applications, but also how to use the technology responsibly and ethically.

Large swaths of the workforce stand to be impacted by GenAI—incrementally at first, but in wholesale fashion soon after. The reality is that many workers have yet to receive any substantive training in how the technology works, how it will change the way they perform their jobs and how to mitigate the still-significant risks of its use in the workplace.

Experts say that many roles will be transformed by GenAI. For example, employees in traditional customer service jobs may find that they spend more time overseeing chatbots or automated processes than interacting with clients.

A 2023 international survey by the Boston Consulting Group found that while 86 percent of workers believed they would need training in AI, only 14 percent of front-line employees reported receiving any upskilling training to date. Similarly, another study conducted last year by Randstad’s Workmonitor Pulse found that only 1 in 10 workers had been offered any AI-specific training, despite a 2,000 percent growth in job postings requiring AI skills.

The GenAI skills gap is causing restless nights for CEOs, CHROs and chief learning officers. In a 2023 survey of HR leaders by TalentLMS, a learning management system provider in San Francisco, 64 percent said the rise of generative AI has changed the key in-demand skills needed in their organizations. More than half say AI literacy is a new must-have skill for employees in all roles.

Training in GenAI has also emerged as a key perk in attracting job candidates and retaining top employees. “If you’re a company that has innovation as part of its talent brand, and you’re trying to attract job candidates who want to continuously build their digital skills, promising them training in GenAI should be a no-brainer,” says Allison Horn, executive director of talent and organization strategy for consulting firm Accenture and a former HR executive with that company. “The same goes for making GenAI training available to existing employees as a retention tool.”

Creating Learning Strategies

Organizations that are meeting the challenge of training their workforces in the use of GenAI often rely on a handful of best practices. Most understand the importance of using companywide literacy training as a first step to demystify the technology and reduce the “fear factor” of potential job loss to GenAI tools. Learning experts say these companies also are careful to calibrate the right mix of external and internal learning content so they can scale training while at the same time maintaining some agility to keep pace with the rapidly changing technology.

Best practices also include dedicating considerable time to teaching the responsible and ethical use of GenAI. Problems such as the technology “hallucinating” nonsensical content or simply providing inaccurate responses will diminish as the capabilities of large language models (LLMs) such as ChatGPT improve with new product releases and as more organizations create their own custom versions of LLMs. Still, the technology has flaws and there remain risks, including employees inadvertently entering confidential company data into GenAI tools.

Given that the arrival of GenAI will leave few jobs untouched, learning experts say training initiatives should start with companywide “fluency” education that includes both conversations and formal courses that cover what the technology is, how it works, and where its strengths and limitations lie.

“Ideally, it starts with senior leaders and change agents in the organization, those with influence and knowledge, talking either in person or through use of short videos about why GenAI is important and what it can do both for jobs and the organization,” Horn says.

This awareness training should address the trepidation many employees still have about GenAI to get them in the right mindset to learn.

“It’s important to help people get over the very natural ‘threat response’ they have when introduced to new technologies,” Horn says. “This is especially true with GenAI, because it’s had a more immediate impact on people’s work and personal lives than other new technologies introduced in recent years.”

Horn says this training can be more effective if real-world examples are used to show workers how GenAI can amplify what they like best about their jobs and reduce the time they have to spend on more mundane or repetitive tasks.

Evelyn McMullen, a research manager specializing in talent management and employee experience with Nucleus Research, believes a lack of clarity and understanding is a significant barrier to building GenAI capabilities in an organization. “HR teams must improve employee training, not just on how to use GenAI, but to address the risks and apprehension many user bases face,” she says.

A Tiered Learning Approach

Before launching additional stages of learning, experts say it’s critical for leaders and employees to first identify and segment job tasks where GenAI can have the biggest impact and where it may not be as helpful. That includes identifying tasks that might be fully automated without human assistance. Company leaders can then build learning plans around those tasks for which GenAI will augment human skills.

A new report from SHRM and The Burning Glass Institute, titled Generative Artificial Intelligence and the Workforce, found that it will become increasingly important for leaders to evaluate the composition of their workforces and determine how exposed specific jobs are to disruptions from GenAI. For those roles most impacted, the study found that reskilling and upskilling will be essential as GenAI grows more central to business operations.

“In this dynamic landscape, blending in-house training with strategic external hiring will be pivotal to harness the full potential of GenAI innovations,” the study authors wrote. They also found that as AI disrupts more traditional job roles, the importance of such uniquely human attributes as critical thinking, empathy and adaptability will become even more pronounced in parallel.

A tiered learning approach customizes training to the specific needs of different job types. Technical positions such as software developers, data scientists and cybersecurity experts require one type of GenAI training related to building applications or integrating AI into back-end systems, for example. HR, recruiting and learning professionals require different approaches tailored to performing specific tasks with GenAI, such as writing job descriptions, creating engagement surveys or building training modules.

Experts say a common thread running through training for all job types should be teaching effective “prompt engineering,” which refers to crafting good instructions to guide GenAI to the most useful and relevant responses. This is the modern-day version of “garbage in, garbage out”—the quality of prompts determines the quality of answers.

“It’s teaching the art of asking good questions,” Horn says. “Employees need to know how to tune their prompts to receive the best answers back from GenAI tools. Training should show examples of bad prompts, average prompts and excellent prompts to show the different results all three will yield.”

Some organizations also have found value in creating “prompt libraries,” a collection of best-practice prompts that have been used with success for specific tasks or projects. Employees can customize these prompt templates for their own unique needs.

The Right Mix of Content

For company leaders, a key decision is whether to use externally or internally developed resources for GenAI training. Experts say e-learning courses from third-party providers can play an important role, but they also have limitations, such as becoming outdated as GenAI tools change.

“If the GenAI learning content you need already exists externally, there’s little reason to re-create it internally,” Horn says. “Companies should leverage external content as an accelerator to get people up to speed and then build their custom learning layers on top of that to address company- or industry-specific needs.”

Some organizations create GenAI boot camps where they contract with local universities to create custom programs for their workers. Others tap into the vast amount of GenAI learning resources now available from third-party providers, many of whom have seen a surge of interest in their offerings.

Udemy is one of those providers. Stephanie Stapleton Sudbury, president of Udemy Business, says courses on the platform related to ChatGPT grew by 4,400 percent in the past year, as measured by learning hours. That makes it among the most in-demand content on the platform. Udemy now offers more than 1,600 courses tied to ChatGPT and prompt engineering in multiple languages, with almost 3 million enrollments, she says.

“GenAI is becoming the new benchmark for tech skills in the workplace, including for nontechnical roles,” Sudbury says. “The demand we’re seeing on our platform reflects that.”

The course-publishing model Udemy uses, which relies on instructors who also are technology practitioners to create and publish courses to the platform, keeps content timely, she says. And demand for GenAI learning courses isn’t coming just from people in technical jobs. Some popular Udemy GenAI courses include “Discover, validate and launch new business ideas with ChatGPT,” “ChatGPT marketing: Creating complete campaigns with ChatGPT” and “Generative AI: From big picture to idea to implementation.”

Other third-party content providers are experiencing a similar explosion of interest in GenAI-related courses. Learning vendor edX now offers more than 500 AI­related courses, including content tailored to specific industries such as education, health care and technology.

“These are clear indicators of the growing appetite for AI knowledge in organizations,” says Andy Morgan, executive vice president and head of edX Enterprise.

Keeping Pace With a Rapidly Changing Technology

Efforts to equip workers with the skills to use GenAI are complicated by the fact that the technology is a moving target and learning content often has a limited shelf life. Creators of GenAI tools regularly release new features, and the technology is evolving quickly.

For example, when ChatGPT was first introduced in November 2022, it was still a text-only tool that routinely produced hallucinations and inaccuracies. It could only answer user questions based on data up to September 2021. Today, most GenAI applications are multimodal, with the ability, for example, to conduct searches using photos and respond to voice commands. They can also understand context and nuance better, and tools such as ChatGPT 4 are now trained on information up to fall 2023 or later.

Keeping pace with such a fast-moving technology requires agile approaches to instruction that can be easily updated and modified.

Bryan Ackermann, head of AI strategy and transformation for management consulting firm Korn Ferry, says that using strategies such as creating “communities of learning” can help organizations stay more responsive to changes than relying on tools such as e-learning courses.

“E-learning can quickly grow outdated because of the lead time usually needed for content and course production,” Ackermann says. “By the time these courses get to end users, the GenAI tools have changed again. So, you end up seeing a lot of generic content that’s trying to be relevant despite the evolution of the technology.”

Ackermann says the most effective training methods he has seen involve creating learning communities that encourage employees to experiment with the latest versions of GenAI tools and learn best practices from internal subject matter experts or experienced peers.

“These communities are given different LLM tools to use, along with the appropriate guidelines and policies about how to use the technology safely and responsibly in an enterprise context,” he says. “The focus is on helping people learn via trial and error, including using supporting resources like libraries of best-practice prompts employees can learn from or customize.”

Organizations including LinkedIn, Harvard University, Meta and Walmart have created generative AI “playgrounds” to help employees learn about and master the technology. Playgrounds contain software, domain-specific data and policies that encourage technical as well as nontechnical workers to experiment with GenAI in a safe space. This arrangement avoids some of the risks involved when using external computer servers, such as leaking proprietary company data or violating copyright law.

Teaching Responsible, Safe Use

Among the biggest fears of top executives when GenAI tools first became available was that workers would get their companies in legal hot water with copyright violations or would inadvertently leak confidential data when using the applications.

While these concerns remain, more senior leaders have calculated that the benefits of GenAI outweigh its risks. As a result, they are creating more rigorous GenAI governance structures and encouraging their chief learning officers to double down on using employee training to help protect against such potential problems.

In a 2024 workplace forecast, analysts from Gartner said organizations must continue to actively manage the risks of GenAI. That includes creating more rigorous access and file classification policies, as well as improving quality control and judgment when using the outputs of GenAI tools.

Many rank-and-file employees also recognize the risks of using the technology irresponsibly. A 2023 study by Salesforce found that 62 percent of desk workers said they don’t have the skills to “safely and effectively” use GenAI, and 70 percent of business leaders said they didn’t believe their teams had the skills to “safely” use the technology. The study surveyed more than 4,000 workers in the U.S., U.K. and Australia.

“Ensuring safe and responsible use is a big topic in many of the learning communities we’ve seen organizations set up around generative AI,” Ackermann says.

As the technology continues to improve, experts say it will remain important to teach employees to stay vigilant in evaluating GenAI’s outputs. A recent study detailed in The Wall Street Journal found that employees who work on tasks alongside machines tend to pay less attention to detail and miss more errors than if they worked alone.

“Answers you receive from GenAI have a confident, authoritative tone, and that will only increase with time,” Horn says. “Human experts will still need to decide where content produced by GenAI is reliable and accurate and where it might be more shaky. It’s important not to take GenAI’s output as gospel, even as the technology innovates and gets better and better.”

Dave Zielinski is a freelance business journalist in Minneapolis.