How To Jump-Start Learning At Work

How To Jump-Start Learning At Work

There is a paradox at the heart of the new AI technologies.

One the one hand, they make almost everything more efficient. Andy Jassy, CEO of Amazon, reports that Amazon has already saved $260 million and 4500 developer-years of work just in software upgrades. Klarna, a customer service company, claims that its AI assistant handles two-thirds of customer service chats and is doing the work of 700 full time agents. A carefully controlled study of BCG consultants showed increases in both the volume and speed of work on a range of consulting tasks. Artists, photographers, authors and film makers are also becoming both more productive and more creative.

Given the economics of these technologies, companies will have no choice but to adopt them. And professionals will have no choice but to learn how to use them.

There is a downside, however. The use of these technologies can also inhibit the development of important job-related skills. While the technologies, in theory, hold the promise of freeing up time for iteration and learning, in practice people often use them to do a task quickly and at a level of quality that is “good enough.” The ease of use of the technologies, in fact, encourages this.

Flipping the Script in Schools

The dilemma may be most obvious in schools. Students use ChatGPT to write papers, teachers use other generative AI tools to catch the cheating, and a cat-and-mouse game ensues. The students avoid both work and learning. The teachers face an existential question: what’s the point?

What if, instead of using the new tools in an old context, we could flip the script and use them to dramatically increase learning? There is evidence that this may be possible.

In 1984, a set of studies at the University of Chicago under the guidance of Benjamin Bloom demonstrated that the average student taught through a process of mastery learning performed one standard deviation above the average student in a typical classroom. Mastery learning is not rocket science; it simply seeks to assure each student understands basic skills before moving on to skills that build on them. But it takes time.

The studies also found that students taught through tutoring performed two standard deviations above the average student in a classroom setting. This means that the average student who was tutored performed better than 97% of the students in a classroom setting. That’s an astounding effect.

These studies caused Bloom to begin a search for “methods of group instruction as effective as one-to-one tutoring.” AI tutors may be the answer. Many such tutors are available already, covering subjects from philosophy to programming in Python. I have used several and was amazed at how good the basic models are. Khan Academy has already begun efforts to integrate tutors into its curricula. The promise is tutoring at scale, something that only the rich can afford today.

Flipping the Script in Business

What is the equivalent of this phenomenon in business? In 2023, according to the 2023 Training Industry Report, companies spent over $100 billion on training (including the time of their employees dedicated to this training). Much industry training mirrors classroom training (or worse, is simply online slideware). How might new approaches using AI tutorials improve both the efficiency and effectiveness of this training?

The design would certainly depend on the topic of interest. Tutors focused on conveying practical knowledge could be customized to the skill level of the student, based on the principle of mastery learning. Other topics – those related to interpersonal relationships or cultural awareness, for example – require deeper understanding or even behavioral changes. These might benefit even more from the affordances of these systems. They could enable employees to challenge, explore and question the material through interaction with the tutor. Although most corporations today are focused on using generative AI to make employees more efficient, the big payoff may be in making them more effective.

The challenge goes beyond the learning of individual employees. AI technologies and robotics can automate work to a degree that inhibits learning for the next generation of professionals. In his book, The Skill Code, Matt Beane gives the example of robotic surgery systems. The use of these robots by senior surgeons obviates the need for the surgeon to rely on residents to assist with surgeries, which prevents them from getting necessary, hands-on experience. One manufacturer of such a system has developed a mentoring mode, which permits a student to do the surgery while a surgical mentor observes. The surgeon maintains the ability to jump in if necessary. The mentoring capability, however, is rarely used. Beane explores the implications of this for learning in many contexts. What is needed is a rethinking of work practices to sustain skill development, one that incorporates not only the short-term need for efficiency but also the longer-term need for highly skilled people.

Managerial Take-Aways

So, what should executives do in this AI-enabled world to jump-start learning in their organizations?

◦ First, seek to use AI to make internal training courses more effective; shift from online slideware to AI tutors – many of which can be developed in-house.

◦ Second, make learning a meaningful objective for each employee; too often, development is viewed as a secondary objective, one that employees can get to when time permits; this mindset causes learning to be under-valued, which is dangerous in a time of rapid change.

◦ Third, don’t fall into the trap of maximizing the extraction of productivity from AI at the expense of learning; leave employees the slack time necessary for continued growth.

◦ Finally, encourage those in engineering and IT to think through the consequences of their design decisions on the learning of those affected; it may be difficult to fully account for the implications of automation – many consequences are surprises that appear only during use – but if the issues are raised and discussed during design there is a better chance of avoiding a blunder.

These steps take discipline because they require reinvesting some of the savings of AI in the development of employees. But an increased emphasis on learning is likely to be a requirement of whatever new world emerges as AI continues to permeate every aspect of our jobs. It is best to begin the shift in mindset now.