Turbocharging Organizational Learning With GenAI

Generative AI can radically improve an organization’s ability to learn. With OpenAI’s introduction of ChatGPT in November 2022, for the first time in the 200-plus-year history of advanced automation, the machines talked back. Instead of having to “speak” Java or Python, people could use everyday dialogue — which is why the tool garnered more than 100 million users in its first two months of public availability.1

More profoundly, by facilitating interaction in human language and deftly handling unstructured words, images, numbers, and sounds, GenAI opened up an entirely new way of creating, capturing, and transferring organizational knowledge. In this article, we’ll argue that leaders need to embrace generative AI as a new organizational capability, and not just because it automates a variety of tasks economically. Combined with traditional AI, generative AI expands the scope of potential improvement in many processes and decisions and the ease with which this new knowledge can be applied. This, in turn, creates the potential for a positive compounding effect on organizational learning, with human and machine agents working in concert to create new competitive advantages.

Creating New Capability, Not Just Adopting New Technology

Paradoxically, the very generality and broad applicability of generative AI can make it challenging to adopt. In our research, we found that the most advanced organizations view GenAI not as a stand-alone technology but rather as an organizational capability. Two examples illustrate this perspective:

• Blue Cross Blue Shield of Michigan (BCBSM): This $35 billion health insurer implemented a cross-functional GenAI/AI leadership team to educate employees about how to use the technologies, follow responsible AI practices, measurably enhance efficiency in a series of projects, and innovate on key processes such as contract review and benefits administration. When BCBSM applied a generative AI tool that enabled better analysis and standardization of terms and pricing across the company’s services, it recouped more than $10 million in savings after applying its GenAI tool to its existing IT contracts, which enabled better analysis and standardization for terms and pricing across the company’s services.
• Wolters Kluwer: This 4.2 billion euro business and academic information company built in organizational slack for learning. Leaders scheduled an hour a week during which everyone attended peer-led learning sessions, with the majority focused on practical GenAI/AI applications. This approach boosted employee skills, innovation, and even retention. Ultimately, employees felt more connected to the company because the sessions not only enabled the creation of a dynamic community of learners but also provided a vehicle for them to communicate their innovations to a broader audience.2

Both companies succeeded by viewing GenAI/AI as an enterprisewide capability. They committed to both top-down leadership and bottom-up enthusiasm, fostering learning environments that enabled rapid experimentation and value creation.

If we look at rigorous productivity studies of generative AI’s impact, we see a broad pattern of improvement. One study looked at the productivity of call center agents when they were given access to a GenAI-based conversational assistant and found an improvement in productivity of at least 14%, as well as higher service quality and faster onboarding of new agents.3 Another research initiative involved an extensive study of more than 758 Boston Consulting Group consultants and found that when the generative AI tool matched the task, productivity increased by 12% and speed of task completion by 25%.4 As in the call center study, the lower-performing employees got a bigger boost from the tool than did more experienced workers. In the time since these studies were conducted in 2023, the capabilities of the large language models (LLMs) and other supporting software have been steadily increasing.

GenAI represents a fundamental shift in the economics of WINS work — work involving words, images, numbers, and sounds. We introduced this new category of work in a September 2023 Harvard Business Review article because we believe the idea of cognitive work is too broad.5 A carpenter is a cognitive worker, as is an attorney, but the impact of GenAI will be greater and faster for an attorney than for a carpenter.

For that reason, we created this new subcategory of WINS work. Examples of WINS work include management consulting (every major consulting firm has adopted LLMs), customer service (LLMs have been deployed at scale in many enterprises), and even movie production (Tyler Perry halted a massive studio expansion project after seeing the capabilities of OpenAI’s text-to-video product, Sora).6

The bottom line: Every organization has WINS work in it — and it can be radically improved.

Establishing a New Management Model

While a professor at the MIT Sloan School of Management in the 1960s, Douglas McGregor suggested that many management systems began with a negative view of the worker and created incentives and punishments to control workers’ behavior. He called this Theory X and suggested that there was room for more worker engagement and participation — what he called Theory Y. This would be an organization with more respect for and empowerment of its people. Shortly after this, psychologist Abraham Maslow suggested Theory Z, in which the company’s culture, meaning, and purpose would lead to an even more powerful staff commitment to the organization’s goals and processes.

Along these lines, perhaps we need a Theory A, in which the core unit of analysis is agents — human and machine — working together in dialogue. Those agents may assist, augment, or automate, depending on how well they can progressively structure the task at hand. Over time, this creates a vastly different learning curve in WINS work and thereby throughout the entire organization because, in most cases, WINS work is the frontier for organizational learning and productivity. Throughout history, breakthroughs in management — such as scientific management, the quality movement, and lean practices — have driven competitive advantage. Each approach introduced new learning mechanisms, allowing organizations to improve productivity, quality, and value. GenAI represents the next step in this evolution, with the new technology enabling three fundamental new capabilities:

1. Dialogue-enabled processes and products. Generative AI is the first technology to interact using human language, allowing processes and products to explain themselves. Soon, the idea of machines not being able to articulate their workings will be as rare as using a physical map for a road trip.

2. Citizen programmers. GenAI enables employees to create their own AI agents and tools. This is akin to machinists crafting custom jigs, but for cognitive work. The resulting explosion of WINS tools will make organizations vastly more productive.

3. Handling unstructured data. Generative AI is able to process the 80% to 90% of information that is unstructured. For example, MIT created a chatbot using CustomGPT.ai to guide students in course selection by simply uploading course descriptions and schedules.

Together, these three pillars allow GenAI to accelerate an organization’s learning curve through a combination of progressive structuring of WINS work and a reallocation of decision responsibilities that enables the product or process to explain itself in any format the learner prefers. Moreover, it can work hand in glove with traditional AI. This approach, like scientific management before it, offers a new way to capture and transfer knowledge and establish a platform for continuous improvement.

Dialogue-enabled processes and products that can tutor employees; citizen programmers who can create great new cognitive “power tools”; and the ability to efficiently process the vast amount of unstructured information that’s typically largely outside the reach of traditional automation represent a triad that will vastly change organizational learning curves. Furthermore, a combination of the human agenda and the GenAI agents arranged in a learning system is becoming the new unit of analysis when work is being designed.

Driving Productivity With GenAI- and AI-Enabled Learning

In the many executive briefings and training sessions we’ve led, we’ve observed that most people don’t have good intuition regarding how GenAI and AI create economic value, and they have an all-or-nothing mindset about automation. To frame this issue practically, we turn to the Keen-Scott Morton model, which offers insight into how GenAI elevates productivity. Building on previous research, Peter G.W. Keen and Michael S. Scott Morton mapped decisions along two dimensions: management role in operational, managerial, and strategic decisions; and the level of “structuredness” of the decision or task. (See “GenAI or Traditional AI?”)7 Operational control systems deal with day-to-day transactions (for example, did the product ship?); managerial decisions often involve improvement or problem-solving (for instance, are our margins going up or down?); and strategic decisions concern where and how to compete (for example, should we even be in this market?). Concerning the other dimensions of structuredness, fully structured problems have a clear language of description, established methods for finding answers, and objective measures on the quality of the answer. Think about FICO scores: They structure the hard problem of credit evaluation and have fully structured a potentially difficult and fluid assessment.

The following is an example of how one company used generative AI to structure an otherwise unstructured task. Jerry Insurance, a startup that helps people choose auto insurance and refinance auto loans, has more than 4 million customers. Before it adopted GenAI, handling customer requests via chat was semistructured, requiring human intervention. (See “A Chatbot Changed the Economics.”)

After implementing GenAI in April 2023, the company automated 89% of those interactions, allowing for the escalation of only the 11% requiring human input. Moreover, the group of chatbots the company created included one that could consider the tone and emotion of the caller and escalate the call immediately, if needed. Ultimately, humans dealt with the tasks that needed the human touch.

The GenAI system at Jerry Insurance progressively structured these previously ambiguous texts and chats with customers, a process maintained by the customer service staff. The company’s approach illustrates how generative AI enhances productivity by progressively structuring the task at hand. The core AI team built the GenAI models, then the customer service personnel were trained on how to use and update the models so that they could continually improve.

This type of progressive structuring changes the nature of the work and its capacity. For example, Jerry Insurance’s leaders believe that they can handle three or more times the volume of customers using the same capacity of 32 customer service agents. This is not a one-and-done automation: Now that they have begun this type of improvement, they are turning to other processes, starting with voice calls, in addition to text and chat.

Perhaps most importantly, this organizational learning is not limited to occurring inside of companies. In fact, GenAI is the new user interface. LLMs can choose the language, emotional tone, level of expertise, output format, output modality (such as images, text, music, and so on), point of view (for example, supportive or adversarial), and many other dimensions. We expect that, in time, every product, process, and interaction of substance will have an intelligent agent to explain how to use it and diagnose any customer need.

Organizing for Success: GenAI as a New Quality Movement

To take advantage of such a broad and general capability and a new concept of human and machine agents working in tandem, it is essential that leaders view GenAI as more than just another technology. Drawing inspiration from the quality movement, we recommend that they take a structured approach to building generative AI capabilities, akin to the belt ranking system used in martial arts, with clear paths for developing expertise:

• White belt: Introduce core GenAI concepts through hands-on experience. Understand the power of dialogue in improving work. Establish a foundation for widespread adoption.
• Yellow belt: Develop deeper understanding, including prompt engineering and the creation of chat agents. Foster knowledge-sharing.
• Green belt: Build and implement multiple agents. Lead teams and projects. Understand security and ROI, and scale at least one pilot. Teach others how to create value with GenAI. Demonstrate skills, GenAI’s real-world application, and the ability to spread knowledge.
• Black belt: Deepen understanding of GenAI/AI tools and techniques. Gain knowledge of MLOps (machine learning operations) and AIOps (AI for IT operations). Develop the ability to teach and to design a teaching curriculum. Demonstrate the ability to manage GenAI talent and oversee a portfolio of AI projects.

Using an approach like that of the quality movement enables senior executives to have some measure of organizational capability in their companies. For example, if you are in a WINS-intensive organization, such as a law firm, you ought to create a broader and deeper pyramid of capabilities that are easy to measure and report on. In addition, the approach is immensely practical because you don’t have to hire as much expensive AI talent, and you won’t waste time translating the work context to a consultant or technologist. The local experts can build it themselves, with help from others familiar with the context.

In a sophisticated machine shop, the machinist who can create new jigs to help set up new jobs, or develop equipment to fix or enhance existing machines, makes the entire shop more productive. Likewise, it is easy for people to create their own agents and robots with the same outcome. We are seeing an explosion of new GenAI-powered tools for WINS work that anyone in an organization can use — tools that will continue to grow the learning ability of organizations and anneal hard-won process or product understanding into freestanding agents or simple robots. They will be just like apps in the app store, but ever so much easier to create.

Theory A asks people to think about how they can go from assistance to augmentation to full automation and about what new tasks humans can undertake with these new models. Using a Theory A approach could invite educated people into the progressive structuring of valuable and as-yet-unautomated data and processes. It is not a fully formed new management system, but we invite thoughtful executives into the exploration of new learning and competitive territory.

Rather than chasing a killer app, organizations should focus on building comprehensive GenAI capabilities, allowing themselves to adapt to the rapid evolution and use these tools for continuous improvement across all operations. By adopting an approach that draws from the quality movement and imagines organizations not as people and dumb machines but as a dialogue among co-intelligences, leaders will be able to navigate this new landscape and unlock the vast potential of combining GenAI and AI for enduring competitive advantage.

“The MIT Sloan Management Review is a research-based magazine and digital platform for business executives published at the MIT Sloan School of Management.”

Please visit the firm link to site