Interest in generative AI (GenAI) is so high that many companies are trying to implement it in their business.

How much value businesses will get from those investments, however, remains unclear. In a survey of approximately 100 chief data officers in late 2023, 80% said they were experimenting with GenAI but only 5% had put it into production at scale. The survey, conducted annually since 2012 by data analytics expert Randy Bean and published by consulting firm Wavestone, also found that only half had the talent needed to implement the technology well.

That talent—those employees and how a company organizes them—can determine how much business value an organization gets from GenAI. What type of talent do companies need in order to develop and deploy this technology at scale? What makes the ideal GenAI dream team? When we asked several AI experts those questions, their answers were wide-ranging, each coming from a different perspective:

• Randy Bean, author of Fail Fast, Learn Fast: Lessons in Data-Driven Leadership, stresses that it’s important to have executive leadership. But exactly which executive should have responsibility for AI is still in flux.
• Daniel Kolodziej, principal at ChangeLogic, a management consultancy, suggests two different GenAI teams: one for the startup phase, the other for scaling up.
• Clas Neumann, senior vice president and global head of SAP Labs, notes the importance of including a wide range of cultural viewpoints, and having team members who demonstrate empathy for customers’ experiences.
• John J. Sviokla, co-founder of GAI Insights, a consultancy specializing in AI, believes that education about and practical experience with generative AI is a prerequisite that must be met well before any substantial implementation.
• Andreas Welsch, founder and chief AI strategist at consulting firm Intelligence Briefing and author of AI Leadership Handbook: A Practical Guide to Turning Technology Hype into Business Outcomes, recommends establishing a center of excellence on GenAI to build organizational capability.

While these experts all stressed that it’s early days for GenAI implementations, each offered advice about how organizations can begin assessing opportunities, mitigating risks, and laying the groundwork for achieving substantial value from this technology. Here is what they had to say, in several categories.

Wanted: Leaders with data experience—and exposure to GenAI

As with any new endeavor, strong executive leadership is a must. “The more connected the initiative is to sponsorship from the C-suite, the greater the probability of success,” says Bean.

Neumann points out the value of having a leader who balances a strategic approach that considers the possibilities of the technology with a principled view that keeps in mind how these systems influence customers’ and workers’ lives.

But which executive should serve as point? The continued evolution in data analytics and AI leadership roles could make it a challenge to identify which executive would be best.

For example, the chief data officer role has been a high-turnover position for years, notes Bean. Meanwhile, a new role—chief AI officer—has emerged.

Some companies have both. Capital One, for example, recently appointed a chief AI officer in addition to its chief data officer, he says. Cleveland Clinic and Mayo Clinic also have both, according to Bean.

Leadership also requires practical experience with and understanding of GenAI. Such education, starting at the top, is a key enabler of value-producing GenAI, says Sviokla.

Sviokla says that few executives really understand GenAI, how it works, and the potential effect it can have on their business. He recommends requiring executives to have at least five hours of hands-on experience with GenAI. And by this, he means literally hands-on.

“What I’ve seen happen again and again is a senior executive will go on to the free version of ChatGPT 3.5, ask it a trivial question and get a trivial answer, so they’re not impressed,” Sviokla says. “What they need to do is to work for five to ten hours on a real problem that’s of interest to them, to really push the tool and understand its power.”

Spreading the knowledge: The importance of familiarizing employees with GenAI

Education is not only important for leaders—it’s needed for changing cultures so that organizations can implement GenAI in productive, profitable ways.

Once an executive leader has gained practical, hands-on knowledge and understanding of GenAI, that leader can sponsor the education of the entire workforce, including encouraging people to play and experiment with GenAI tools, Sviokla says. This builds not only acceptance, but an organizational capability, very much like the Six Sigma quality movement, where there were various levels of training and knowledge from white belt to yellow belt to green belt to black belt. Organizations need this capability in order to identify and develop opportunities for generative AI throughout the business.

Bean emphasizes that making people at all levels of the organization more familiar with generative AI is vital. He co-wrote a 2024 Harvard Business Review article with Thomas H. Davenport, in which the pair argued that “since the technology is broadly applicable across businesses, almost every company should make such education available at all levels.”

Education at all levels and throughout the company would also chip away at a long-standing cultural barrier. Over more than a decade, Bean’s survey has found that only a small percentage of companies believe they have established a strong data and analytics culture.

While that portion doubled in the last year—from 20% in 2023 to nearly 43% in 2024—it’s clear that companies need to do better, Bean says. In the survey, half of organizations still report they do not manage data as an asset and have not created a data-driven organization, while three-quarters mention cultural hurdles as impeding their progress using data, analytics, and AI.

Educating leaders, middle managers, and others about the use of generative AI is valuable—if only to help them catch up to leading-edge employees who are already exploring the technology, notes Welsch. ChatGPT is so easy to get that many workers are already using it. Encouraging workers to come out of the shadows and explain how they are using GenAI in their jobs could serve to foster a positive culture and reinforce the technology’s productive use. Organizations might even reward employees for the best applications.

Another avenue to encourage learning and experimentation is to look for GenAI features in existing apps and software and start using them, even customizing them. For example, AI capabilities embedded in human capital management software can help hiring managers write job descriptions or suggest questions they could ask a candidate during an interview, Welsch explains.

Once companies decide to implement GenAI on a large scale and develop custom projects, Welsch recommends developing a center of excellence.

The need for contributors whose talents extend well beyond AI skills

Neumann’s prescription calls for including technical skills in data science and machine learning models, and people “with a very high awareness and knowledge about security topics,” he says. Critical thinkers willing to question the outputs of GenAI models will help the team scrutinize results and ensure relevance and quality.

Neumann argues it’s crucial to include a range of individuals to reduce the risk of bias in the use of data—and to increase the team’s awareness of cultural issues and relevant policies for using AI. “The number one criteria for choosing a great GenAI dream team is cultural and regional diversity,” he says. “A high diversity within the community developing models will be key. What is permitted or socially accepted for an AI to do in Country A may be a compliance issue or unacceptable in Country B.”

In addition, the team should build bridges with other players. “Beyond technical skills, it is important to know that an AI team will never work in isolation, but always in close collaboration with application teams, technology platform teams, and of course customers,” Neumann says.

Sviokla's concrete suggestions for a team highlight data analytics skills, creativity, and adaptability. His ideal team would include a business leader who can talk in terms of both technology and business, as well as handle politics, and an operating executive who understands how to deal with a complex, dynamic environment.

The team should also include “open-minded data people,” who can pick out the signal from the noise. The point is not to have perfect data, but to have good enough data to improve business decisions and create competitive advantage. “It’s like the old story of the bear and the two people in the forest. I don’t have to outrun the bear. I just have to outrun you,” says Sviokla.

Finally, the team should include a user-interface designer who is innovative enough to take a new approach that Sviokla calls design for dialogue. “AI is the new user interface,” Sviokla says. “This is the first technology in the history of humanity where we talk to it and it talks back.”

An important point to remember when selecting members of a GenAI dream team is that both the technology and the jobs required in a potential team are constantly changing. Experimentation is a logical starting approach.

“Things change month to month, even week to week,” says Bean. “I think organizations are still experimenting to find out what combinations of people, skill sets, and disciplines, work.”

Kolodziej says that, as with any project, organizations should consider two different teams for GenAI development work, depending on the needs of the business as well as the stage of the project. Each team would have different skillsets.

The “0 to 1 team” would have a goal of discovering, defining, and validating a use case. It should be staffed by people from various disciplines who are comfortable with ambiguity and divergent thinking. Its job is to start with a business problem, identify what’s needed and what’s feasible, and gauge what value it will create. It would build an early proof of concept.

The second team is for scaling the project and consists of people skilled in process optimization and scaling technology.

“It’s a difference in mindset: figuring out the right thing to build (0–1 team) versus building the thing right,” says Kolodziej. The second team requires heavy-duty technical skills (IT, data, security) as well as expertise in legal, procurement, and other business areas. He compares it to digital transformation projects, “where if it’s going to touch many customers and many different departments, then everyone needs a seat at the table.”

GenAI teams will evolve with experience

Kolodziej emphasizes how early we are in the development and application of a potentially revolutionary technology that some compare to the Internet. Today, companies may use GenAI applications to increase productivity and reduce costs, which he calls incremental innovation.

GenAI “is one of the world’s most impressive innovations in recent times, but it’s still predominantly being used for incremental means,” Kolodziej says. “Incremental innovation saves you money or optimizes processes, but radical innovation can generate revenue.”

This radical innovation will eventually come, Kolodziej says, but for now “everyone’s just getting their head around [GenAI], understanding what it means.” After all, “the creators of LLMs [large language models] can’t fully explain how those models work, so how can we understand what’s possible when we don’t really have a clear mental model of how things fit together?”

Read more