How to manage two types of generative artificial intelligence

As organizations continue to experiment with and realize business value from generative artificial intelligence, leaders are implementing the technology in two distinct ways. 

According to a new research briefing by researchers Nick van der Meulen and Barbara H. Wixom at the MIT Center for Information Systems Research, organizations are distinguishing between two types of generative AI implementations. The first, broadly applicable generative AI tools, are used to boost personal productivity. The second, tailored generative AI solutions, are designed for use by specific groups of organizational stakeholders.

The research, which is based on roundtable discussions with members of the MIT CISR Data Research Advisory Board and interviews with executives, outlines the two approaches and highlights unique challenges and management principles for both.

Broadly applicable generative AI tools

Generative AI tools like conversational AI systems and digital assistants embedded in productivity software are broadly applicable by design. They are versatile, and their use is typically defined and refined by its users, the researchers write. 

“This is AI for everyone,” said J.D. Williams, a vice president and chief data and analytics officer at global animal health company Zoetis, which is a member of the MIT CISR data board. “It’s where you’re bringing in external products and privatizing them within the company so your data is protected.”

Generative AI tools pose four key challenges to organizations, according to the researchers:

1. Because generative AI tools are based on large language models trained to predict the most likely sequence of words in a given context, they often produce output that is common. As a result, the quality and relevance of the output depends on the specificity of the prompts a user enters.
2. Generative AI tools can lack context, contain bias, present false or misleading information as fact, and fail at doing math. Consequently, users must continually critically evaluate a tool’s output to avoid accepting bias or inaccurate assertions.
3. Unvetted, publicly available generative AI tools can present significant risks, particularly when employees use them for work. These risks include data loss, loss of intellectual property, copyright violations, and security breaches. 
4. Generative AI tools are expensive. Providing users with licenses to tools from multiple vendors can quickly become costly, once free trials and early-adoption incentives expire.

To counter these concerns, companies should provide employees with sanctioned access to a select number of generative AI tools to create a safe space for experimentation. To enable safe and successful use of generative AI, the researchers suggest that leaders do the following:

• Develop clear usage guidelines. These guidelines should be developed by cross-functional leadership teams with representatives from technology, legal, privacy, and governance interests. Guidelines should specify which tools are permissible and under what conditions, and articulate associated risks and potential consequences. Williams said that mitigating risk, protecting data, and ensuring regulatory compliance are critical for any AI governance framework. “You want to be innovative and speedy, but you also want to be risk aware, data secure and compliant,” he said.
• Invest in training. Organizations should establish AI direction and evaluation practices, including teaching employees to effectively instruct and interrogate generative AI tools, understand the underlying models, and use the tools responsibly.
• Standardize with a select set of vendors. Form a cross-functional team of likely generative AI tool users to help determine which tools hold the most potential for your organization. 

Generative AI as a tailored solution

Generative AI solutions are business-case-driven development initiatives that address strategic business objectives and create value for specific groups of organizational stakeholders, ideally at scale, the researchers write. Organizations fund these solutions after they meet innovation criteria related to end-user desirability, technical feasibility, and business viability.

“[Businesses] deploy these solutions in specific functions that perform specific tasks,” Williams said. “In manufacturing, for example, it might be around monitoring processes and products to make sure they’re trending in the right direction as they’re manufactured. There are a lot of great applications here.”

Although generative AI solutions share some similarities with other AI initiatives, they present three unique challenges, according to the researchers:

1. As more employees begin to realize the potential of generative AI, organizations risk the development of “shadow generative AI,” in which groups of stakeholders independently pursue unsanctioned solutions with the help of eager vendors.
2. A few vendors own and control most of the foundation models that support generative AI solutions. This complicates organizations’ understanding of the models and their own ability to assess biases and predict model behavior, which can introduce various risks, including data leaks and inaccurate outputs. Uncertainty around future usage, model performance, and pricing also makes it difficult for businesses to estimate long-term operating costs of generative AI solutions.
3. The value that organizations realize from generative AI solutions depends on whether the organization buys a solution, enhances a vendor model, or builds its own solution. Depending on the approach, there are trade-offs in transparency, context awareness, and cost. 

Organizations can realize value from generative AI solutions by making them cross-functional efforts. To succeed with targeted generative AI solutions, organizations can also do the following, the researchers write:

• Establish a formal, transparent generative AI innovation process. Organizations need clear governance structures, early and consistent stakeholder engagement, and a focus on scalable solutions. 
• Formulate guidelines for generative AI development decisions. Leaders must differentiate between the various generative AI development approaches to help teams make informed decisions, given that there are different benefits and detriments when buying, building, or enhancing generative AI models. 
• Create a generative AI vendor partnership strategy. Effective partnerships with generative AI vendors rely on mutual understanding and long-term collaboration. This fosters adaptability and continuous improvement, which benefits both parties.
   

AI Executive Academy

In person at MIT Sloan

Register Now