Practical AI implementation: Success stories from MIT Sloan Management Review

Generative artificial can be intimidating and risky. But many firms are making it work in practical and successful ways. New insights from MIT Sloan Management Review show how companies are focusing on small and medium-sized wins while ensuring that powerful AI tools — such as agents capable of presenting choices and making decisions — are used appropriately. 

Achieving small-scale transformation with generative AI

The early stages of generative AI use are best described as “extensive experimentation,” according to MIT Sloan lecturer Melissa Webster and senior lecturer George Westerman. Enterprises are pursuing small-scale transformation with generative AI, taking a targeted approach that creates value while minimizing risk and laying a foundation for large-scale efforts. 

Webster and Westerman describe three categories of small-scale transformation: 

• Tasks common to employees in many roles. Large language models are popular for tasks such as synthesizing information and documenting meetings. While most enterprises prefer to license private instances of publicly available LLMs and integrate them with productivity tools already in place, some have OK’d the use of external tools when confidential information isn’t involved.
• Specialized uses for specific roles and tasks. Enterprises with a bit more risk tolerance are willing to use generative AI for business processes. Popular use cases include coding, supporting customer service, guiding the creative process, and creating content at scale. For example, CarMax uses generative AI to summarize customer reviews, with the summaries posted to research pages for customers to use. Highly regulated industries such as financial services, meanwhile, have rolled out AI to generate reports and review contracts.
• Products and consumer-facing applications. E-commerce companies are debuting chatbots and otherwise supporting more personalized shopping experiences. Companies such as Adobe and Canva, both of which make graphic design software, are embedding generative AI tools into their products. 

These use cases are the riskiest small-scale transformations, as they largely remove the human in the loop.

Leaders looking to make the most of generative AI should consider their risk tolerance, the potential to scale from the pilot stage, and the need for “foundational investments,” such as data cleansing and model training, to get AI projects off the ground, Webster and Westerman write.

Read: Generate value from Generative AI with ‘Small t’ transformations

Leading the AI-Driven Organization

In person at MIT Sloan

Register Now

Encouraging innovation at Colgate-Palmolive 

Consumer products companies have a history of market research and analysis that long predates the advent of generative AI. Thomas H. Davenport, a fellow at the MIT Initiative on the Digital Economy, and co-author Randy Bean explore how Colgate-Palmolive applies the technology to tried-and-true business practices and generates measurable business value.

One example is fast access to research data. Colgate-Palmolive is applying retrieval-augmented generation to LLMs that process a treasure trove of proprietary consumer research, third-party data, and Google search trends. Instead of poring over a pile of market research reports, employees can use generative AI to query the entire dataset.

Another source of value is the development and testing of new product concepts. Generative AI systems can help employees produce copy and imagery for a new concept within minutes. From there, concepts can be tested on digital consumer twins that play a role similar to that of in-person focus groups but don’t experience the fatigue of, say, testing two dozen toothpaste flavors in one sitting.

Colgate-Palmolive hosts its AI tools in an internal hub. To access the company’s AI Hub, employees must undergo training that covers both responsible and practical use of AI. This emphasis on skill-building has paid off: According to the company, thousands of employees have reported an increase in the quality and creativity of their work when using AI. 

Read: The Generative AI focus shifts to innovation at Colgate-Palmolive

Reframing how AI assists with decision-making

AI is now able to generate choice sets — as opposed to a singular “best” decision — and has the capacity to explain trade-offs, identify new opportunities, and learn from past outcomes. Enterprises using AI in this way are implementing intelligent choice architectures, defined by researchers Michael Schrage and David Kiron as “dynamic systems that combine generative and predictive AI capabilities to create, refine, prioritize, and present choices for and with decision makers.”

Liberty Mutual uses an AI-informed intelligent choice architecture to help claims adjusters triage incoming calls and resolve inquiries. And at pharmaceutical company Sanofi, AI systems guide managers in optimizing investments and overcoming sunk-cost bias, a common barrier to abandoning unsuccessful projects.

These are powerful tools. As they both inform and make decisions, they present a governance challenge. Leaders need to strike a balance between responsible use and oversight while expanding the capabilities of intelligent choice architectures, write Schrage, a research fellow at the MIT Initiative on the Digital Economy, and Kiron, editorial director, research, at MIT Sloan Management Review. 

Ultimately, that means the hierarchy of decision rights within the enterprise will shift, with the power to shape the decision environment having greater impact than the power to make decisions outright.

Read: How intelligent choice architectures rewrite decision rights