Generative AI in the Workplace: Maximizing Its Value

There’s been a lot of hype around generative AI in the workplace in recent years.

Much of the excitement around the advances in generative AI and machine learning is warranted—these technologies have the potential to improve productivity and supplement creativity.

However, it’s easy to get swept away by the hype without a clear vision of how to actually implement these tools effectively.

“There’s so much rhetoric—falsely placed I think—that we’re gonna have these massive transformations on day one,” says Paul Leonardi, Duca Family Professor of Technology Management at UC Santa Barbara. “We are in discovery mode still.”

In this article, learn how leaders can move beyond experimentation and leverage AI effectively to create value in their organizations.

Supporting Employee Experimentation

The first step to moving beyond experimentation is experimenting.

If you haven’t explored Large Language Models (LLMs) yet, now is a good time to begin.

“Go to chatgpt.com, or download the app and ask a question. It is so easy,” advises Meagen Eisenberg, Chief Marketing Officer of Lacework. “I’m in awe of how many people haven’t even just gone to the URL and asked it a question.”

Employees want to use AI in their roles. According to research from Workday, 71% say AI can help obtain information faster and more easily, and 73% want their company to explore more ways to incorporate AI.

Unfortunately, Gallup reports that 53% of employees don’t feel sufficiently prepared to work with AI, and 26% aren’t prepared at all.

“Our job as a leader is really to make sure we’re creating the right environment in which employees using our new tools and their capabilities can thrive,” says Leonardi. “So that means in many cases, that we’re encouraging them to experiment and try out the new capabilities that are enabled by our technologies.”

Common Obstacles to AI Adoption

1. Hallucinations

When AI confidently states incorrect information as fact, this is known as a “hallucination.”

The New York Times estimates that chatbots hallucinate anywhere from 3% to 27% of the time.

This means that overdependence on AI can lead to significant issues—as one lawyer found who relied on ChatGPT in a court case.

An article from MIT highlighted three reasons for these hallucinations:

1. Training data sources: The data that trains these models contains both accurate and inaccurate information, as well as human biases.
2. Limitations: Generative AI is predictive, meaning that its responses to prompts are predictions of the most likely next word or sequence. It can’t discern the truth.
3. Design of generative models: Due to the generative nature of this technology, even if it was trained on accurate data, it could still generate inaccurate responses.

Understanding the nature of generative AI is important for ensuring that you’re able to use it effectively. It’s important to keep in mind that because AI can’t inherently discern fact from fiction, every answer it gives is essentially “made up.”

“One of the points I always make sure to try to convey to my clients and my students is that all AI answers are hallucinations,” says Leonardi. “Everything is a prediction that’s happening in the moment, and some hallucinations correspond with our understanding of reality, and others don’t.”

Because AI is often more accurate than humans, it’s a good idea to weigh the risks vs. rewards of how you use AI in your organization.

If you’re using it to write, research, or perform other similar tasks, always verify the information. If you’re using an API key to create a custom chatbot or similar resource for your company, ensure that safeguards are in place to mitigate the risk of hallucinations.

2. Employee Resistance

Another common obstacle to AI adoption is employee resistance.

According to Workday, just over half of employees are concerned about AI putting them out of a job.

Although this is a common concern, there are other factors that play into employee resistance.

For example, the same report shows that 69% worry that their personal data will be misused, and 61% say their employers aren’t transparent about AI use and its impact to employees.

The good news is that 64% of employees say their concerns about AI would be alleviated with clear guidelines and transparency around the use of AI.

3. Security

Another concern surrounding AI adoption is security. Many of the current AI models—especially free ones—rely on user prompts and feedback to help train their models.

This can be problematic for organizations with sensitive data.

“When GenAI was hype, we didn’t want all employees going out and take getting free licenses and starting to put sensitive corporate data into those different LLMs that take data for training,” explains Deepika Rayala, Chief Digital & Information Officer at Cornerstone OnDemand.

This means if you’re going to use an existing AI model, it’s important to familiarize yourself with the privacy and security standards in place.
“If you choose the wrong model and the wrong architecture, you are sharing data externally, and you want to think about how to overcome that,” says Arvind KC, Chief People & Systems Officer at Roblox.

Progressing Beyond Experimentation: 7 Tips

So how can you move beyond the initial experimentation phase of AI?

“So many companies I find are struggling with thinking about ‘where do we even begin to implement GenAI?’” says Leonardi. “What does success with GenAI look like, or AI more broadly?”

Here are seven tips for moving beyond experimentation and maximizing AI’s value at your organization.

1. Give Employees Safe Ways to Use AI

Because of security concerns and uncertainty, employees might hesitate to use AI to improve productivity.

“What you need to create for your companies is that environment of innovation, and access to these tools, and learning and training them,” says Eisenberg.

This requires openness and transparency surrounding what employees are permitted to use AI for, and what they should stay away from.

Create a document that outlines guidelines for using AI. This should include:

• What tools employees can use
• Clear guidelines surrounding the data that can be fed into AI models
• The potential for bias in AI and the importance of verifying any information it provides
• Use cases for AI at your company
• Best practices for AI use

It’s also important to encourage managers to have conversations with employees about their individual roles and how AI will impact their jobs to alleviate concerns about job security.

2. Determine Where AI Can Create Value

In order to maximize the potential of generative AI in the workplace, it’s important to consider where AI can add the most value to employees and the organization alike.

A good place to start is considering monotonous or boring tasks that could be automated.

“What are the things that we want to be doing that really add value?” says Leonardi. “What are the things we have to do to today that are important but don’t get anybody excited? And for which we have lots of data upon which we could train a model and have the model do a lot of that work.”

Lenardi has developed a framework for determining where AI can create value called the STEP framework:

1. Segmentation: Identify tasks that AI shouldn’t or can’t perform, those that AI can augment or enhance, and those that can be fully automated.
2. Transition: Determine how employees’ jobs can be enhanced by leveraging the time freed up through AI implementation.
3. Education: Empower employees with the knowledge and skills to effectively leverage AI.
4. Performance: Consider how performance evaluations should change to account for the use of generative AI.

In his Teamraderie experience, Adopting GenAI, Paul Leonardi can join your team live to determine ways to implement this framework in your organization.

3. Begin and Continue

While it’s a mistake to get caught up in the “hype cycle” of AI, you should also avoid underemphasizing its value.

“People overestimated [this] trend in the short term but underestimated in the long run,” says KC. “The only thing that you can do if a trend is true in the long run is to begin and continue.”

• Begin: Don’t be the last to start using AI. Consider how it can be used now at your organization to improve productivity.
• Continue: Explore AI’s use cases at your company and revise as needed.

Here are three tips KC provides for beginning and continuing:

1. Consider where you have a high probability of success
2. Uncover the real problem you’re facing and how AI can help mitigate it
3. Don’t start too big—take an incremental approach to AI adoption

It’s important to start somewhere, but in doing so don’t put too much pressure on yourself—or your team—to make massive changes right away.

4. Inch Employees Into AI

One surefire way to make employees nervous about AI is to do too much with it all at once.

“One thing that I often see, you know, when I’m working with companies doing these kinds of implementations is we often start too big…” says Leonardi. “We try to do too much at once, rather than… inching people into it.”

Instead of trying to find ways AI can make an enormous difference in your company immediately, consider incorporating it into your processes incrementally.

Leonardi provides three suggestions for easing employees into AI:

1. Give employees confidence in using AI and understanding its value
2. Uncover potential use cases for your organization
3. Make strategic bets in areas where you think you can create the most value

The world of AI is still unfamiliar—Leonardi compares it to the “Wild West.”

If you want to move beyond experimentation, you need to determine where AI is most valuable and pursue opportunities in those areas.

5. Enable Mid-Level Managers to Work With AI

While C-suite executives can push for AI adoption, true innovation thrives when mid-level managers—those closest to the day-to-day operations—are empowered to experiment and integrate AI into their work.

But how do you equip mid-level managers with the tools and confidence to leverage AI without compromising security?

“Our goal has been to not make the ability to do a thing be a function of the position you are in,” explains KC.

For example, if a project manager has a brilliant idea for using AI to streamline a workflow, they might be hesitant to pursue it due to security concerns or uncertainty regarding their authority to pursue those opportunities.

Instead of requiring complex technical knowledge or waiting for approval from higher-ups, mid-level managers should have access to:

• Data classification levels: Clear guidelines regarding the sensitivity of different corporate data.
• Instructions for data classification levels: Clear guidelines on which models can be used for each sensitivity level of data.

“We need to make sure that we are creating the right kinds of policies and procedures in place that aren’t constraining people from doing what they want, but enabling them to take on these new initiatives,” says Leonardi.

6. Don’t Expect AI To Be 100% Right

There are often different expectations for AI and humans when it comes to accuracy.

It’s important to remember that employees are often wrong more frequently than AI models, especially when these models are trained with your data.

“I think about how many reps are out there just giving the wrong answer,” says Eisenberg.

“We certainly all live in this illusion that our machines should be a hundred percent accurate,” explains Leonardi. “That’s an unrealistic expectation for these tools, especially because they’re constantly learning and evolving based on the data that we feed them.”

While it’s important to ensure that your AI products are accurate, it’s also important to be realistic about AI’s capabilities. Don’t expect AI to be completely right all the time—instead, verify any information it provides.

7. Talk to University Students

Finally, it’s not a bad idea to visit your local university and simply discuss AI with the students.

“Find some university students you can go and hang out with because you just have to have an osmosis of ideas that you not often don’t get in the company,” says KC.

Having these conversations can open your eyes to potential use cases and ideas that you otherwise wouldn’t have considered.

Maximize GenAI’s Value With Teamraderie

If you’re interested in working with Paul Leonardi to determine how the STEP framework can be leveraged to integrate generative AI into your organization, consider our team experience, Adopting GenAI.

In this live, virtual, interactive workshop, Paul will help your team become more adept at integrating AI into your processes, creating value for employees and your organization as a whole.