In the first few months of 2023, everyone was talking about generative AI (including me). Now that the hype and hysteria have died down, the industry has moved on to investigating how AI can become a useful fixture of the workplace. Currently, individual employees are far out in front of their employers in answering that question; at least four-fifths of knowledge workers have experimented with generative AI, and more than half have used it on the job. Meanwhile, only one company in four has developed a policy or set of guidelines for employees in the use of AI in the workplace. If we learned anything from the introduction of social media at work, we know how important it is to have guidelines and support for employees using new technology.

If you’re not already operating under a generative AI policy at your organization, matched with a rollout and training plan, now is a great time to get started sketching those things out. And you’re in luck, because two recent studies have lent important insight into what you should keep in mind while doing so. 

In September of this year, a collaborative study by BCG and business school scholars published the results of a study of generative AI in the workplace. Their findings were eye-opening: in creative and ideation tasks, the speed and quality of work improved across the board (as measured by a panel of judges who graded the output), but not equally so. The work of junior employees improved most radically—in many cases, the AI raised the quality of their work to the level of their more experienced peers. 

‍

Insight #1: Generative AI can bring immediate benefits to individuals and teams—and is particularly helpful for less experienced staff.

In the BCG study, the work did become more uniform—there was less individuality in AI-assisted work, which is a point worth considering. In general, though, this result makes a compelling business case for an official rollout of AI: your team’s work will almost immediately become better.

It's not all good news, however: study participants’ work assisted by AI got worse when it fell in an area that lay outside of data on which the AI was trained. Your team might run into this if, for instance, you’re performing research on a competitor or dealing with some other topic that has not been documented in widely-available sources (for example, a program internal to your company, or a product or service that has only recently been released). Generative AI doesn’t always signal when it lacks specific knowledge; when presented with a more obscure query, it may simply make something up, and the burden is on the user to notice when this is happening. 

‍

Insight #2: Leaving it up to individuals to figure out AI will have uneven results. 

You may be thinking: “Well, then, I’ll just train my team on the risks of generative AI!” Unfortunately, evidence suggests that this may not work. In the BCG study, the participants using AI were split into two groups—half received pre-training in prompt engineering and the limitations of the AI, while the other half did not. The participants who received the training actually performed worse on the out-of-dataset tasks than the untrained cohort. The reasons are uncertain, and this result calls for additional study, but it could be that participants who received the training were left feeling a little overconfident about their ability to navigate the AI’s strengths and weaknesses, and as a result were less likely to rework and improve lower-quality AI output.

‍

Insight #3: A simple L&D module on the use of AI will not be enough. 

The most successful participants in the BCG study pursued a “chimera” model: they understood where the AI was reliable and relied on their own judgment in areas where the AI fell short. In doing so, they drew on their own professional experience—proving that, even in the age of generative AI, experience and personal insight still counts for something. 

As you think about the use of generative AI on your employer brand team, one thing becomes clear: mentorship has an important role to play. AI can help anyone get up to speed quicker in areas where they lack experience, but junior employees still need the help and occasional oversight of more experienced team members to bring out the benefits of generative AI while avoiding the risks. 

Once you’re ready to encourage the use of AI within your team, a study by Charter sheds light on how to do so. Workers today often feel uncertain about generative AI, either because they’re afraid that the machines will take their jobs, or because they’re simply not comfortable with rapid (technological) change. Different members of your team are likely to experience the onset of generative AI differently. For instance, the Charter study found that Black employees are significantly more likely to fear job loss from AI than white employees (53% to 39%), while women are much less likely than men to agree with the statement, “I am excited about the prospect of using generative AI as part of my day-to-day work” (46% to 66%). 

‍

Insight #4: Change can be hard, especially when it involves technology, and not everyone will experience it the same way.

The Charter study argues that the best response to this situation is not to impose this change on your team. Instead, involve them. Allow them to identify pain points or opportunities where AI might be most valuable—after all, they know their own jobs better than anyone. When they raise concerns, listen to them and take them seriously. Make it clear that the intention behind the change is to help them achieve more, work faster, and clear the obstacles that are getting in their way. 

The evidence clearly states that adding generative AI to your team’s capabilities is a good idea. It makes the work better and puts you in position to connect with candidates and current employees more effectively. Bringing these capabilities to your team, though, calls for careful consideration into how you can make the rollout a collective experience and leverage the experience and insight of your entire team in doing so. 

We are, in the end, better together, whether we’re talking about humans partnering with AI or team members benefiting one another. The time when AI will do everything on its own still appears far away–and it may never arrive. In the meantime, it all gets back to our relationships, how well we work together, and the time we invest in making those connections stronger. 

*Note: The image at the top of this article was generated by DALL-E 3, using the contents of this article as a prompt.