Helping Employees Succeed with Generative AI

Author: Paul Leonardi

If one universal law regarding the adoption of new technologies existed, it would be this: People will use digital tools in ways you can’t fully anticipate or control. The arrival of generative AI-based technologies using large language models (LLMs) such as ChatGPT and Bard raises a critical question for leaders in all types of organizations: How can you manage employees when the capabilities at their fingertips are constantly changing and the effects of those changes are unpredictable?

Two characteristics of generative AI make this problem more challenging than it was with previous digital solutions. First, ChatGPT is one of the most widely diffused and fastest-adopted products in history. Just two months after launch it had 100 million users. Instagram took two and a half years to acquire that many. Facebook took four and a half years. The faster technology spreads, the less time users have to learn from one another and mimic patterns of use. Second, unlike virtually any other digital technology we’re accustomed to, AI-enabled tools are designed to change by themselves—continuously. Each time you provide new data to an LLM to produce text or computer code, the technology learns and its capabilities grow. The things it can do for you next week won’t be the same as the things it could do for you this week. Thanks to the autonomous learning that characterizes the most-advanced AI-based tools, your employees aren’t learning to use a new technology once—they are learning to use it nearly every time they engage with it.

During a three-year research project with 10 knowledge-intensive companies at the leading edge of AI use, I devised a framework—STEP—that can help employees take advantage of new technologies. STEP consists of four interrelated activities to help leaders ensure that employee-facing AI plays a positive and productive role in their organizations: (1) segmenting tasks for either AI automation or AI augmentation; (2) transitioning tasks across work roles; (3) educating workers to take advantage of AI’s evolving capabilities and to acquire new skills that their changing jobs require; and (4) evaluating performance to reflect employees’ learning and the help they give others.

Several companies have already adopted the STEP framework. Their early experiences have demonstrated that it meets three critical needs: It empowers employees to actively participate in shaping their new responsibilities. It allows leaders to redistribute tasks and reimagine work roles in ways that add value to the company. And it provides a way to deal with the new reality that technological change is no longer episodic but, rather, a force that organizations and their employees must manage continuously.

This article describes how leaders can deploy STEP and how three companies (all of which have requested anonymity) have used it effectively. They include a marketing agency I’ll call MarkCo, a medical device manufacturer I’ll call HealthCo, and a metropolitan planning agency I’ll call UrbanGov. (Disclosure: I have served as a paid consultant for MarkCo and HealthCo.) Those organizations saw STEP as a new way to enable and encourage employees to capitalize on AI. Lessons from them will help leaders from other companies improve employees’ experience at work and create new value for their organizations.

Idea in Brief

• The Quandary

How can you manage employees when the capabilities at their fingertips are constantly changing, and the effects of those changes are unpredictable?

• The Context

Unlike virtually any other technology we’ve known, AI-enabled digital tools are designed to change by themselves—continuously. Given this rapid evolution, users have less and less time to learn from one another.

• The Solution

To help, the author has devised a framework for leaders—STEP—that consists of four interrelated activities:

1. Segmenting tasks for either AI automation or AI augmentation

2. Transitioning tasks across work roles

3. Educating workers to take advantage of AI’s evolving capabilities

4. Evaluating performance to reflect employees’ learning and the help they give others.

1. Segmentation

No single AI will do all the things that one person does in a work role. Informed leaders should ask, “How will AI affect the various tasks my employees engage in?” To determine the answer, have your employees create three categories: (1) tasks that AI can’t or shouldn’t do, (2) tasks for which AI can augment workers’ actions, and (3) tasks that can be automated by AI.

HealthCo adopted ChatGPT for its junior staff. Leaders encouraged staffers to first determine the tasks for which the AI would not be helpful. Determining how to comply with federal policy and how to safeguard the company’s IP when working with outside consultants quickly rose to the top of the list.

Next, employees decided which jobs the AI could help them with. One time-consuming task was ensuring that contracts accurately reflected the details of requests for proposals (RFPs). Here ChatGPT was very useful. After reading through an RFP and a standard contract template, it could generate a draft contract that reflected the terms of agreement. Paralegals could then review the draft for specific areas of concern that would need to be amended.

Finally, employees identified tasks that could be completely automated. One was the laborious job of emailing outside parties requesting changes to a contract. The AI could automatically generate those emails by reading through revised contract language.

Once the junior staffers had segmented their tasks into those three buckets, they began figuring out how the AI could augment or automate some of them. A preliminary analysis at HealthCo suggested that as a result of the deployment, the staffers each managed to free up five hours a week for additional tasks.

All the successful companies I studied encouraged employees to take the lead on segmentation and asked them to experiment with the tools. Their leaders convened meetings at which employees discussed the results of their experimentation. They allowed employees to help them reach consensus on best practices. Employees were proud that their leaders trusted them to deploy their expertise, and being part of the experimentation and planning gave them insight into how their companies would use AI, reassuring them that automating part of their jobs wouldn’t put them out of work.

2. Transition

Because AI either helps complete work tasks faster and more accurately (augmentation) or takes some of them over completely (automation), some employees will have less to do after AI is deployed. In some cases, companies might reduce head count. Yet among the 10 companies I studied, only one eliminated jobs in response to the efficiencies gained by augmenting and automating work. Two other, more common strategies were to transition work roles by deepening or upgrading them. Deepening roles allows employees to devote more time to certain tasks than they were previously able to. Upgrading roles frees them completely from some tasks and gives them more-critical ones instead.

MarkCo adopted a chat-based LLM to help junior associates create marketing collateral, such as PowerPoint presentations. That gave them time to spend on other tasks that added more value to the firm. Some employees began to do more competitor analysis. Others focused on testing and evaluating campaigns. Managers at MarkCo identified which employees had the aptitude and the interest to deepen their knowledge in those areas and drew on expertise within the company to create and provide their initial training.

Upgrading roles involves having employees perform tasks that had typically been conducted by someone more senior. For example, at UrbanGov junior planners often spent much of their time building land-use models for urban development. During the STEP segmentation process they augmented and automated certain tasks, giving them the bandwidth to assume scenario building and other higher-level tasks, some of which had been done by senior planners. That meant UrbanGov had to find new tasks for senior planners, so the lead planner turned his own responsibility for managing relationships with city planners over to the senior planners. “After careful evaluation, I decided to give them a big chunk of my job,” he explains. “That allowed them to feel comfortable giving up scenario building. Now I can focus my efforts in new directions too, since I’m freed from maintaining all those relationships.”

Deepening work roles was the most common strategy among the 10 companies in this study, representing nearly 70% of all transitions. Most leaders found it easier to help employees identify new value-adding activities within their roles than to take over tasks from more-senior employees.

3. Education

The first two stages of the STEP framework require workers to learn new skills, some of which are directly related to using data, algorithms, and AI. Employees need to know how AI tools work, how to train AI on documents or data proprietary to the company (often called “fine-tuning”), how to create effective commands or prompts (“prompt engineering”), and how to evaluate the validity of an AI’s predictions. Because AI tools are constantly evolving, employees can’t learn new skills once and be done. They need to revisit the segmentation process and continually refresh their learning about AI’s capabilities and the areas into which their work roles will transition. Over the three-year period, leaders and employees at the companies I worked with went through segmentation and transition an average of two and a half times. Employee education was thus a top priority.

HealthCo, MarkCo, and UrbanGov all embraced the need for continual employee reskilling, but in different ways. For AI and data skills, HealthCo created a “boot camp” for its employees. MarkCo contracted with a local university to create custom programs for teaching employees those skills. The university devised tests that employees had to pass to certify that they were “AI ready.” Every year the tests change according to evolutions in the technology. UrbanGov, which had a much smaller budget available, bought subscriptions for short courses on AI, simulation, and data management from companies that offer corporate learning, such as LinkedIn and Udacity. Employees on the planning team were encouraged to complete one course each month, whether they were directly using AI or not.

Companies that prioritized education had two things in common. First, they embedded the ethos of learning in their culture. Leaders and managers across the company framed AI as a learning opportunity. Employees were not expected to know how to use AI perfectly or how to segment their tasks around it. Instead they were expected to explore its capabilities and take time to determine how best to incorporate it into their work. Second, these companies provided time for employees to engage in the learning opportunities they provided. HealthCo, for example, expected employees to devote at least two days each quarter to attend the boot camp or refresh skills they’d learned in it. UrbanGov earmarked three hours a week for employees to take the online courses to which they subscribed.

This shift to providing ongoing education for employees can have other benefits, too. One company in the study found that employees given learning opportunities associated with the implementation of generative AI tools were roughly 30% less likely than those not given such opportunities to leave the organization.

4. Performance

The final stage of the STEP framework requires managers to rethink how they evaluate employee performance. Typically, employees are assessed on the speed, efficiency, creativity, or accuracy with which they complete certain tasks, and most discussions regarding the impact of AI focus on how their productivity will increase. But STEP gives employees responsibility for determining whether they can use AI to be faster or more accurate. As a result, performance evaluation shifted in several ways at all the companies I studied. Segmentation and transition quickly changed expectations regarding what tasks employees should do and how. And because roles altered multiple times in a year, often rendering objectives identified at the start of a performance period obsolete, annual performance evaluations no longer worked. Every company in the study shortened its performance evaluation period, most often to quarterly.

In addition, workers were constantly interacting with new people. Many of the updated performance evaluations I studied involved identifying the people with whom an employee interacted most often to determine whether that employee was a useful collaborator. Collaborators themselves usually provided the evaluation. The advantage of this was that they could better evaluate the help they received than managers could. And because the evaluations were done at short intervals, an employee could quickly put the feedback to use.

Of the companies I worked with, HealthCo had the most aggressive and technologically advanced performance-evaluation system. Its data scientists created a dashboard that drew information from email communications, Slack use, and calendars to show whom employees were most dependent on and who most depended on them. Every six weeks—the length of HealthCo’s new evaluation period—the dashboard automatically sent all employees a list of their most frequent collaborators and asked them to rate their interactions with those people. The data was compiled and shared with employees and their managers so that employees could gauge their collaborative performance and make changes. After two years using this new approach, employees reported a 72% increase in their satisfaction with the company’s evaluation system in comparison with the prior two years.

Lessons for Managers

As companies have applied STEP, leaders have needed to adjust their approach to managing their employees. Four principles have helped them navigate the challenges associated with AI.

Trust employees to experiment

Leaders who have followed the STEP framework have found that they do better when they let employees lead the process of discovering how best to use AI. They can set broad objectives, such as increasing accuracy or doing a particular task faster, but they should allow employees to determine how best to segment their work. That requires trusting them to make good decisions and acknowledging the fact that as the people closest to the work, they know best where AI can have the most impact.

Create conditions for learning and incentivize helping

Learning is the real imperative of successful AI use. Employees must learn and relearn how to use LLMs and other AI tools as they change, how to apply the new capabilities those tools provide to their work, and how to conduct new tasks that add value to the organization. Because employees need to learn with and from others, it becomes essential to adjust how they are evaluated and rewarded to ensure that they are motivated to help one another.

Rethink workforce planning

We typically think of work roles in terms of the tasks they involve. But as AI upends the distribution of tasks, work roles will become more amorphous. Good leaders must figure out how to forecast for head count and how to manage recruiting and promotion in a world where jobs are dynamic rather than static. That means they can’t just hire people who have deep expertise in one specific task. They need to ensure that their experts have a range of abilities and the potential to learn in adjacent areas.

Reimagine your own role

In the age of AI good leaders are those who create the conditions that enable their employees to adapt in the face of changing technologies. Managers need to develop a digital mindset for themselves, their departments, and their teams. And when employees use AI to start doing higher-value tasks, the best leaders will find ways to personally provide value both up and down the organizational hierarchy. The middle of the organization may well be the place where creativity is the most important.

…

The AI-powered organization is coming fast. Leaders should help their employees use AI to create value for themselves and their companies. STEP provides a useful framework for thinking through how AI will lead to changes in work. Most important, it can assist leaders in teaching employees to be successful with this new technology.