Learning to Manage Uncertainty, With AI

Three Ways to Enhance Organizational Learning With AI

While it may be tempting to identify organizational learning as — or, more pointedly, reduce it to — knowledge management or learning and development, organizational learning involves far more than these important activities. It encompasses whether organizations view unsuccessful experiments as failures or as sources of learning; how organizations develop, not just manage, knowledge; and how organizations anticipate the unknown rather than merely capture what is known. It occasionally requires setting aside old ways of working to make learning new capabilities possible.3 What’s more, organizational learning encompasses synthesizing and analyzing information to glean what is and is not working in the enterprise. It also involves optimizing metrics, not merely maximizing performance on existing metrics. Finally, organizational learning addresses the communication, dissemination, and accessibility of knowledge.

Combining organizational learning with AI-specific learning yields more benefits than taking either approach alone. AI-specific learning can significantly enhance (at least) three areas of organizational learning: knowledge capture, knowledge synthesis, and knowledge dissemination. These are not incremental additions; Augmented Learners multiply their abilities in these areas.

Knowledge Capture

Adopting generative AI and embracing developments in traditional AI can expand an organization’s ability to capture knowledge. Organizations can use it to extract tacit knowledge resistant to legacy codification techniques, absorb vast quantities of external information, and even help crystallize knowledge that employees are still learning.

Using AI helps managers capture tacit knowledge, an often intractable challenge for traditional knowledge-capture techniques. Consider the example of NASA’s Mars 2020 mission. NASA wants to explore as large an area of Mars as possible, which means its Perseverance rover needs to be moving as much as possible. At the same time, the agency wants the rover to stop when it finds something “interesting” — a concept that’s difficult for operators to define. Plus, with 30 minutes of communication latency, human operators cannot immediately direct the rover to stop. Vandi Verma, a principal engineer at NASA’s Jet Propulsion Laboratory and chief engineer of Robotic Operations for Mars 2020, explains that AI has helped solve the problem. “We have AI capabilities on the rover where it’ll take a wide-angle image, look at a large swath of terrain, and then try to figure out what is the most interesting feature in there,” she says. Despite the difficulty that humans have articulating what is and is not interesting, the AI learns from past data to operate semiautonomously, without anyone explicitly defining criteria for “interesting.” Perseverance has had to overcome many obstacles while driving on Mars; using “interesting” as a guide for where to explore was an unexpected one. It had to re-create the tacit knowledge behind “interesting” to navigate the terrain successfully.4

Combining organizational learning with AI-specific learning yields more benefits than taking either approach alone.

AI technologies represent new capabilities for capturing existing tacit knowledge. In a more down-to-earth context, LG Nova’s Prasad observes that AI-based augmented reality (AR) glasses have the potential to capture the tacit knowledge of factory workers on the shop floor who have mastered a certain way of working with machines. “If they’re doing a technique on the shop floor that only they know, AR glasses can allow real-time content creation,” she says. While AR use is not common today, Prasad states this use case has the potential to become a more significant approach to capturing tacit knowledge as the technology/hardware matures.

Using AI to distill information at scale enables the capture of salient information that would otherwise be impossible for humans to discern. Since 2021, LG Nova’s mandate has been to work, develop, and collaborate with startups to build new business ventures — a typically daunting task, given the sheer number of potential targets worldwide. Prasad summarizes the question driving the subsidiary: “Can we use AI to find the right startups and create a deal flow that merits being put in front of our executive team?” The answer was decisive. AI has to work alongside humans and can narrow the search process and save executives’ time. While final decisions on which startups to recommend falls squarely on LG Nova’s human team, Prasad says that using AI can help generate a list of candidates for human evaluation and improves the company’s situational awareness while expanding the number of investment targets for its small team.

Employees can also use AI to clarify how their knowledge would work in contexts they haven’t yet experienced. Expedia Group is a case in point. The company, like any large online platform, faces constant security threats. “Travel, like most other industries, is a great target for bad actors,” notes Expedia Group’s chief architect, Rajesh Naidu. He says that the company is beginning to use generative AI to simulate attacks so it can prepare for them. Expedia Group learns from looking at “how an account-takeover scenario would work, or phishing, social engineering, things like that,” Naidu says. With the help of AI, the organization captures the knowledge it needs to prevent fraudulent activities before they occur.

Knowledge Synthesis

Making sense of vast data sets can overwhelm legacy analytics. AI, however, can more effectively systematize an organization’s data, pulling together internal and external data sets while making it all more digestible for managers, customers, and partners.

Jeff Cooper, formerly senior data science director at online personal styling service Stitch Fix, describes how effective generative AI can be in synthesizing and summarizing large volumes of content. “One of the spaces we’ve been working hard on and considering where it might be useful involves customers who have been with us for dozens and dozens of Fixes,” he says, using the company’s term for the delivery of stylist-selected clothing and accessories. “To have a stylist come in and look at all of the feedback they’ve given over years potentially can be really complicated. With our new generative tools, we have the possibility of creating summaries of those things and compressing some of that information a bit further. In this case, it’s almost like you have a stylist working alongside a partner that can help do some of the extra work.”5 Generative tools excel at summarizing, a valuable feature in a business environment, where desirable (and expected) response times are increasingly short.

Organizations need not build their own tools to synthesize data if a significant chunk of their data is in general business products with AI components. Our 2022 annual research report found that 55% of organizations were using third-party tools with these capabilities.6 (That number is likely higher now, with the widespread availability of tools based on large language models.)

Slack is an example of how work platforms are using AI to assist in synthesizing knowledge. More than 700 million messages are sent in Slack each day. In large organizations, especially, the volume of data produced across company channels can be challenging to keep up with.

In response, the company developed a native AI solution for its product that helps workers instantly tap into their knowledge base by answering questions, summarizing conversations, and providing daily recaps of channels. Rocca remarks on the value that a feature like this provides, saying, “We are creating GenAI solutions — a combination of generative AI and machine learning — that deliver a daily recap summarizing all the channels you want to get up to speed on without going through every single message. A sales leader I know uses recaps to stay in the loop on his top 10 accounts, and he has many more accounts than that across many more channels. He doesn’t want to know all the ins and outs that the team is doing to prepare for their next meeting.” The net effect is that users are able to quickly reduce ambiguity about what’s happening with the accounts that matter most.

Knowledge synthesis with AI is not only about synthesizing knowledge within a company, it also facilitates knowledge transfer from one context to another. Expedia Group has been using AI to synthesize data from over 3 million properties, 500 airlines, and 100 million loyalty members in the U.S. Collectively, the company manages more than 1.26 quadrillion combinations for hotel options alone, which includes variations like location and length of stay, all the way down to beachfront vistas and free parking; and assessments of how the presentation of images impacts the end-user experience. With AI, Expedia Group can make sense of all of that information and make detailed recommendations to its partners. Naidu says, “We have enough information to provide really good recommendations to our hotel partners on image selection, image quality, and what content needs to be there to help drive a booking. We have the ability to suggest winning formulas that are based on our insights from other properties.” Synthesizing knowledge, especially across organizational boundaries, can lead to value-creation opportunities with partners throughout a business ecosystem.

Knowledge Dissemination

Organizational learning depends on more than capturing and synthesizing information. A critical challenge in any learning organization is knowledge dissemination within the enterprise. Chief data officers’ mandate to get the right information to the right person at the right time reflects the importance of managing the dissemination challenge. Using AI to disseminate knowledge makes the process more inclusive and personal.

One executive in the cloud services industry observes, “With generative AI, we have an opportunity to ensure that everyone is getting a learning experience that is going to meet their specific needs. That could be someone who’s neurodiverse, or it could just be different learning styles or different languages. Often, we build technology and systems to meet the needs of the majority. AI can provide more rapid, lower-cost opportunities that prioritize some groups that might be underserved or whose needs may not otherwise be met. We can provide an experience more representative of your entire organization and consumer base.”

The opportunities to capture, synthesize, and disseminate knowledge with AI bring a fresh perspective to a remark commonly attributed to Lew Platt, the former CEO of Hewlett-Packard: “If only HP knew what HP knows, we would be three times more productive.” As more enterprises adopt AI and participate in business ecosystems, knowing what the company knows is only part of a larger learning challenge. Organizations also need to know what others, including suppliers, partners, customers, and competitors, know. Of course, knowing every detail would be overwhelming, particularly in environments where speed matters. Furthermore, it isn’t enough for the organization generally to know; individual employees need access to digestible information to make the knowledge useful.