How organizations can achieve big value with smaller AI efforts

What you’ll learn:

Organizations are using generative artificial intelligence for “small t” transformations that help their businesses evolve. This means building capabilities and managing risks at each stage of a three-level risk slope:

• Level 1 focuses on low-risk individual tasks, such as email management and meeting summarization.
• Level 2 applies AI to specific roles, like coding and customer support.
• Level 3 integrates autonomous AI into products and operations.

Under pressure to leverage generative artificial intelligence, smart organizations are discovering that what works in pilot stages doesn’t always translate to large-scale implementations. 

Instead of initiating sweeping redesigns of major business functions, these organizations are working their way up a risk slope, pursuing a series of “small t” transformations that aim for incremental value, according to research from MIT Sloan senior lecturers Melissa Webster and George Westerman that was published by MIT Sloan Management Review.

These smaller transformations are also better suited for managing generative AI’s risks, including data security, AI ethics, and compliance challenges.

“Smart leaders are taking a much more measured and systematic approach to getting to the large things they want to implement with generative AI,” Westerman said during a recent webinar that detailed the research. “Every step of the way, they’re learning how to manage risk, they’re learning about the tools, and they’re building up capabilities to move forward to these bigger opportunities.”

Climbing the generative AI risk slope

Webster and Westerman defined three categories of AI transformation that represent different levels of risk. Here’s how they chart the journey.

Level 1: Individual productivity

This is where most companies are on the maturity curve. At this level, organizations make generative AI available to employees for low-risk, basic tasks related to their specific roles while keeping a human in the loop during interactions. One common use case is inbox management, such as using generative AI to summarize emails, draft replies, and flag priorities. Employees are also using generative AI to produce real-time transcriptions and meeting summaries, optimize their daily calendars and auto-schedule meetings, and prepare for briefings by getting quick summaries of markets, articles, and staffing levels. Many desktop tools now have integrated large language model capabilities to boost individual productivity.

In more advanced scenarios at this level, companies use generative AI to recast communications in a different voice or to adapt it based on cultural norms. Some leading companies, including McKinsey, are building company-specific LLMs that let employees access vast internal intellectual property resources, helping them perform tasks more efficiently with improved quality. 

Level 1 work sets the stage for doing more with generative AI. “These tasks get people comfortable and reduce some of the fear,” Webster said. “Then they can move to Level 2 tasks that start to transform the way the organization operates.”

Level 2: Specialized roles and tasks

At this stage, companies apply generative AI to specific tasks in job roles or business processes, such as coding and data science, human-in-the-loop customer support, and low-risk content generation and personalization. Software development is a particularly hot area, with generative AI giving programmers a leg up when writing and reviewing code, creating documentation, and conducting data analysis.

Generative AI is also reshaping some customer service and sales workflows. At CarMax, for example, LLMs are used to summarize reviews in hours versus having multiple workers labor for weeks on end, the researchers said. Other companies are generating personalized scripts for sales calls or using sophisticated chatbots to handle common customer queries while the more complex ones are routed to human agents.

“A general theme you see in Level 2 applications is humans and AI working together, finding the places where AI can support the humans, and for the humans to be overseeing the work of AI,” Webster said.

Level 3: Products and processes

This is when companies start adding more autonomous generative AI capabilities to products, customer-facing experiences, and internal operations. Webster and Westerman see companies at this stage using generative AI as part of a multifaceted toolkit that also includes different technologies and people.

With integrated capabilities, organizations such as Adobe, SAP, and Workday are starting to use generative AI to aid in rapid content creation, automate marketing campaigns, or deliver more sophisticated chatbots that make decisions and perform work independently.

Organizations that are already using AI tools can often start activating these Level 3 features, Westerman said. “These can require a whole lot of capability development, and these can require a whole lot of risk management,” he said. “And that’s why companies are taking a careful approach to get up to this stage.”

Leading the AI-Driven Organization

In person at MIT Sloan

Register Now

Achieving small transformations with generative AI

New technologies will improve what’s possible at each of the three levels. AI agents are an emerging area of focus, with autonomous task execution streamlining workflows and solving new business problems. This technology spans a continuum, from simple AI assistants to autonomous agents that act on their own based on guidelines and human input. Agentic AI, the multi-agent version of this technology, will require an “AI manager” to oversee the other specialized agents performing individual tasks, the researchers noted.

While there is plenty of enthusiasm for generative AI and AI agents, there remains a healthy amount of skepticism — another reason why a “small t” approach makes sense. Webster and Westerman shared the following recommendations for leaders implementing generative AI tools:

• Don’t treat all of your organization’s problems as nails that can be tamped down with a generative AI “hammer.” Zero in on problems where this technology might be most useful in solving them.
• Think about where your company is on the risk slope, and build management buy-in on the plan to move forward.
• Don’t force the technology on everyone. Find the people who are excited about it, and leverage their enthusiasm and successes to push for change.

The bottom line: Take your time. “Building the right strategy doesn’t go hand in hand with the gold rush mentality going on right now with generative AI,” Webster said. “Look closely at the work and the business need, and build up competencies before going for the larger swings.”

Watch the webinar: Scaling generative AI — Get big value from smaller efforts

This article is based on a webinar and research conducted by Melissa Webster and George Westerman that was published by MIT Sloan Management Review.

Melissa Webster is a senior lecturer in managerial communication at the MIT Sloan School of Management. She teaches oral, written, and interpersonal communication; persuading with data; teamwork; and leadership. She also investigates the adoption and implications of ChatGPT and other generative AI tools in both the professional and educational realms. Her research explores generative AI’s use by knowledge workers, and its integration in business education.

George Westerman is a senior lecturer at MIT Sloan. He helps executives understand the transformative potential of AI and other fast-moving digital technologies. His research-based insights show leaders the questions they should ask and the steps they can take to help their organizations thrive. His studies on digital-ready culture and workforce transformation provide important insights to move from transformation projects to transformation capability.