5

AI

JD Dillon

SOME SAY ARTIFICIAL INTELLIGENCE (AI) is already starting to transform workplace learning. But others say AI is overhyped and years away from any meaningful learning applications. Are you confused? Conflicted? Not sure who or what to believe? That pretty much sums up the conversation within L&D around AI throughout 2019.

The talent development field is rife with new technologies and buzzworthy trends. While we are quick to explore new tools, we are often quite slow to adopt them. Take mobile devices as an example: The first iPhone was released in 2007, but many organizations still haven’t figured out how to leverage smartphones that their employees are carrying around in their pockets to help them get better at their jobs. Given the everyday ubiquity of mobile devices, this would seem to be a pretty straightforward challenge. However, applying AI, a technically complicated and relatively unfamiliar concept, within workplace learning is likely to be much more challenging right out of the gate.

As the topic approaches the “peak of inflated expectations” on the LearnGeek Hype Cycle (Dillon 2019), let’s explore the AI conversation and determine what L&D professionals need to know to effectively leverage this high-potential workplace technology.

AI Defined

Many L&D concepts are left open for interpretation. For example, what is the industry-standard definition of microlearning? Trick question: It doesn’t exist. A big chunk of L&D concepts (such as social learning, mobile learning, and e-learning) are just established learning principles stuffed into a consumer-friendly, buzzworthy package. As a result, the industry often struggles for years to determine the potential for these “new” ideas.

However, the concepts involved in AI are the result of decades of work by very smart people—people who have nothing to do with corporate learning. That means that this time, L&D doesn’t have to make up definitions; they already exist. Here’s how McKinsey & Company defines AI and key related terms in “An Executive’s Guide to AI” (Chui, Kamalnath, and McCarthy 2018):

• Artificial intelligence: The ability of a machine to perform the cognitive functions we associate with human minds, such as perceiving, reasoning, learning, interacting with the environment, problem solving, and even exercising creativity.

• Machine learning: Algorithms that detect patterns and learn how to make predictions and recommendations by processing data and experiences, rather than by receiving explicit programming instructions.

• Deep learning: A type of machine learning that can process a wider range of data resources, requires less data preprocessing by humans, and can often produce more accurate results than traditional machine-learning approaches.

Figure 5-1 is commonly used to explain the relationship between these three important terms.

Figure 5-1. Cousins of AI—Toward Data Science

Source: Singh (2018)

By no means do L&D professionals need to become AI experts. However, as should be the case with any important workplace technology, we should have enough of a baseline understanding to be dangerous. We need to be able to have informed conversations and make educated decisions regarding how our practices should evolve through AI. Sure, you could hire new team members with AI skills. But, if you don’t have that luxury, you must build your own knowledge using the information available and in partnership with those within your organization who already work with AI.

The Hype

Thirty years ago, the term “artificial intelligence” conjured images of The Terminator and a battle between humans and machines for survival. Just five years ago, people proclaimed that AI and automation were coming to steal our jobs. Today, most of the fear seems to have subsided, giving way to a more practical and nuanced conversation about the impact of AI on the workplace. The World Economic Forum (2018) now estimates that, while machines will displace 75 million jobs by 2022, an additional 133 million new roles will be created. That’s a net gain of 58 million jobs worldwide. And with these new positions will come new skill requirements and support needs.

Before L&D gives into the hype and jumps to apply AI to every solution, we must understand how AI is changing the work itself. According to McKinsey, while 50 percent of the tasks people do today can be automated, only 5 percent of the roles can be entirely automated (Manyika et al. 2017). This means more and more employees will be leveraging AI-driven tools within their workflow, thereby changing their development and support needs. Machine learning is also already being applied across an array of industries, from retail to manufacturing to healthcare. Therefore, if you are not already having conversations about the impact of AI on the workplace, you are already behind the curve.

As machines take on more predictable and repeatable tasks, “people skills” are expected to become a key focus within organizations. According to ManpowerGroup (2019), “by 2030, demand for human skills—social and emotional soft skills—will grow across all industries by 26 percent in the U.S. and by 22 percent in Europe.” But “soft” is a clear mislabeling of these skills, as they are often much more difficult to find and develop. L&D must not only be ready to apply AI tools, but we must also be prepared to help people build skills like analytical thinking and communication, which are topics that have not been a strength of our industry to date.

L&D, We Have a Problem

L&D isn’t good with data. That’s not just my personal observation. It’s an industry-wide, readily acknowledged fact. For example, according to Donald Taylor (2019), chairman of the Learning and Performance Institute (LPI), the skill set in which L&D pros ranked themselves the lowest as part of the LPI Capability Map was data analytics. No wonder “learning analytics” jumped from unranked to #3 on his L&D Global Sentiment Survey of Hot Workplace L&D Topics this year. Traditional industry measurement methods, such as the Kirkpatrick Model, have failed to establish a foundation for data-driven decision making within our field. Now, as we face a technology revolution built on piles and piles of data, we have a giant hole in our metaphorical boat. There’s no way we can put it into the water before figuring out a way to patch it—fast.

But there’s good news. The modern workplace is overflowing with data. Organizations are tracking everything from sales results to safety incidents to call metrics. L&D just hasn’t tapped into this data flow in meaningful, scalable ways. And that leads me to even more good news: L&D pros don’t have to become data scientists. Again, you can bring data skills into your team if you have the opportunity to hire more people. But the smartest data person in your company probably doesn’t work in L&D. They are in your operations group or your business intelligence team. Just as L&D should have been developing strong relationships with IT for the past decade, now we must make friends with our data partners. They not only have the resources we need, but they can also help us understand how we can better apply data to inform our solutions.

To maximize the potential of AI, L&D must evolve in the same ways marketing has transformed their practices during the past 15 years. Back in the day, marketing pros positioned billboards along highways because they knew how many cars would pass by over a given period. But they couldn’t directly attribute that billboard to customer buying decisions. Today, a marketing pro can determine exactly how an online ad influenced changes in sales revenue. The difference? Data. While digital tools are now commonplace at work, most organizations haven’t adopted the same data-first mindset when building the employee experience. This is especially true for L&D, where course completions, test scores, and survey results make up most of our data strategy.

Going from our traditional measurement approach to a “big data” model may feel like big jump. But it’s a required leap if we hope to take advantage of data-driven technology. And, just like marketing, L&D must move forward with purpose when expanding its definition of “learning data.” Thankfully, we can borrow some of marketing’s tools to help us along the way. For example, the 5 Vs (Figure 5-2; Marr 2014) are a popular marketing concept for applying data within a business context.

Figure 5-2. The 5 Vs of Marketing

Every organization’s data strategy should be custom-built based on the needs of their business. Regardless, L&D can no longer think about “learning data” in isolation. Rather, we need to recognize that what we have historically defined as learning data is just one small part of the much larger performance data. Thinking about data—and the rest of our work—in a holistic way is the first step in establishing the mindset needed to successfully implement AI within L&D.

So Many Possibilities

If you can solve the data problem, there’s still a really big question to answer: Does L&D even need AI? My answer is an emphatic yes, but not just because it’s the latest and greatest thing. L&D should move quickly to adopt AI because we desperately need what it can provide: agility.

L&D is the proverbial cartoon coyote in today’s companies. We are always chasing “the business” to find out what they want and need. Then, when we think we have it all figured out, we use our shiny new tools to develop a robust, crowd-pleasing solution. But, right before we implement our solution, the business shifts and a new problem becomes the big priority. And where do we end up? At the bottom of a ravine with an anvil on our heads.

As businesses become more agile to fend off near-constant disruption, L&D must also become more agile. After all, we play a big role in making sure employees have the knowledge and skills necessary to thrive in the current business environment. If we fall behind, employees are potentially left to fend for themselves. Therefore, we must do whatever we can to move faster, ask better questions, and become more proactive when providing learning and support resources. And that’s where AI comes in.

Think back to the definition of machine learning and what the technology is built to do: Detect patterns and learn how to make predictions and recommendations by processing data and experiences. When you apply this concept to our newly expanded mindset on L&D data, an array of potential use cases immediately jumps to mind. What they all have in common is agility—helping L&D get faster and more accurate. But another key consideration is scale: L&D has historically failed to balance the needs of individual employees with those of the entire business. With our limited resources, scale often wins, and we are forced to provide generic solutions to everyone. However, when you introduce AI and improved data practices, balancing these two competing priorities suddenly doesn’t seem so impossible.

So how can AI be specifically applied within workplace learning? Well, I could write an entire book on potential AI use cases. But I’ve come up with a shortlist of applications that are already being implemented with meaningful results for both end users and administrators. Figure 5-3 shows each based on their current level of adoption and perceived level of organizational value.

Figure 5-3. Current AI Applications in Workplace Learning

Personalization

AI can leverage a range of data to provide the right support to the right person at the right time—at the scale of a global business. And this doesn’t just mean pushing online courses in front of employees. It can also include the full range of potential support tactics, including training, coaching, and performance support.

Gap Identification

L&D often has to wait until something bad happens before instituting a new training program. Job requirements are also a moving target. The skills needed to be successful in a role today are often considerably different than even just five years ago. AI can be leveraged to proactively identify performance gaps and skill requirements and then direct employees to the right development resources so they can prepare for future roles.

Impact Analysis

L&D is under pressure to justify its value and prove that it has an impact on business results. Similar to the way marketing connects ads to buying decisions, machine learning can be applied to show exactly how L&D solutions are (or are not) affecting targeted business goals.

Smart Coaching

Managers aren’t always watching employees do their work, which means that many coaching conversations are generic or misinformed. AI can support more accurate, robust coaching. It can also be applied to fill in the gaps when a real-world manager is not available and provide timely, targeted feedback directly to employees.

Expertise Matching

Let’s say that an employee in Orlando, Florida, has a problem. Unfortunately, they aren’t aware that another employee in Phoenix, Arizona, has the knowledge needed to help solve that problem. AI can measure knowledge and performance at scale and connect people for a variety of reasons, including mentorship, projects, and basic problem solving.

Translation

Translating training content is a time-consuming and expensive process. Therefore, employees are often limited to a select set of options, which may not include their preferred language. AI is getting very close to being able to accurately translate content in real time into any available language, without the need for extra work by the L&D team.

Search

Natural language processing (NLP) is a machine learning technique used within a range of AI applications, including chat bots and content authoring. In this case, NLP is used to improve search results by understanding not just the provided keywords but also the added context behind the inquiry.

Content Authoring

A considerable amount of time, money, and capacity is spent building training content. In many cases, L&D pros are acting as middlemen between subject matter experts and employees. Now, machines can write content faster and at a quality level that is pretty darn close to human authors. In fact, you’ve likely read many online articles written by AI, but you just didn’t know it.

Content Recommendation

Often referred to as “Netflix for learning,” AI can connect employees with the most relevant training and support content—without the need for extended searching. This is especially important when trying to navigate the huge catalogs of aggregated content that are now available within many organizations. But AI can also take the idea of recommendation to another level by suggesting resources that have proven effective in helping people with similar backgrounds overcome similar challenges.

Chat Bots

This is the most common use case for AI in L&D today. Rather than searching for reference content with keywords, employees engage in human-like conversations to find specific answers to their questions more quickly.

Tagging

Tagging online content with metadata is an extremely time-consuming process. However, it is necessary to make content searchable within a large catalog. AI can now take care of this tagging process on its own by scraping the resource to determine its topics of focus.

System Administration

Every traditional learning technology has at least one system administrator sitting behind it, assigning training content to employees by team, role, location, or some other combination of attributes. AI can lighten the administrative workload by dynamically identifying training needs and assigning the necessary components as those needs evolve.

Again, that’s just a quick sample of the ways AI is being applied within L&D today. This list will only grow as the technology improves and L&D teams become increasingly savvy in their ideation and execution.

A People-First Approach

So far, even this exploration of AI in workplace learning has focused mostly on technology, data, and content. But ultimately this is a conversation about people—how L&D can provide better support to the employees working on the frontlines of our organizations. We cannot afford to get distracted by how cool or promising the technology is to the detriment of the people we are trying to help. Rather, like everything else L&D does, this must be a people-first conversation.

The introduction of AI within workplace learning will require a foundational shift in the way people think about the work L&D does. Think about how similar technologies have changed the way you look at online shopping or streaming entertainment. We are fundamentally talking about using data to provide a more agile, form-fitting learning and support experience. Employees will quickly see the difference. And they need to be ready—and willing—to take advantage of this new approach. Otherwise, our effort will be wasted.

Similar to our data practices, this is something L&D pros can start working on now, well before they get into the nitty-gritty of AI tech. Preparation all comes back to trust. Do the people you support trust you to do what’s best for them and provide them with the resources they need to do their best work? If you can’t answer yes with confidence, you have work to do! Trust also plays a big role when it comes to improving our use of data. Privacy is a sticky subject in everyday life. These same concerns cannot be allowed to permeate workplace conversations and limit the potential of new technology. Instead, L&D must work with partners to ensure transparency is at the forefront of any data strategy. Some employees will not care how their workplace data is being collected and applied. Others will very much care. Therefore, this detail should be made readily available for those who are interested, and all related practices, including the use of AI to enable the learning and support experience, should be openly documented.

AI is already improving the value L&D can provide to stakeholders and employees. But that value will be severely limited if we fail to include the people we support—frontline employees and management—in our evolution. As I said, not everyone is interested in how we develop learning and support resources. But we should allow them to make that decision for themselves and not assume that they will just buy into whatever we implement because we believe it’s an awesome idea.

Where Do We Go From Here?

Deloitte summarizes the current mindset around workplace AI perfectly in their 2019 Global Human Capital Trends report: “The value of automation and AI lies not in the ability to replace human labor with machines but in augmenting the workforce and enabling human work to be reframed in terms of problem-solving and the ability to create new knowledge.”

This represents a real test for the talent development industry. Not only is this technology changing the way work is done around us, but it will also fundamentally alter how we do our jobs and enable performance. How? Well, the details are still to be determined. But this evolution falls in line with what many industry influencers have been saying for the past several years when they talk about modern workplace learning. The potential applications I shared give you a sense of how the focus of L&D work will likely change—moving from building content and administering programs to addressing business problems, solving problems with a data-enabled mindset, and fostering connections between people and resources.

AI is the fastest-growing technology in the history of the workplace. L&D cannot afford to take five or 10 years to figure this one out. But, in a marketing-heavy profession, will we be able to distinguish what’s real and what’s not when it comes to AI tech? Can we see through the hype and dig down into the fundamental principles that will help us maximize these new tools? Can we develop our own subject matter knowledge while establishing the critical partnerships that help bring these new ideas to life?

It’s safe to say that L&D will not be the department that first introduces AI to most companies. But we still don’t have to wait for someone to come to us to explain how the whole thing works. Start the conversation with your team. Find people in your organization who are already working with AI. Explore online resources and connect with subject matter experts. Improve your knowledge and awareness so that you can start building the foundation of your AI strategy right now. We’re going to have a lot more to talk about when we explore this topic again in subsequent years.