Why CIOs need a two-tier approach to gen AI

CIOs failing to gain organizational traction with generative AI might want to rethink how they are introducing the technology — and how they are honing their AI strategies to suit.

When IT leaders consider generative AI, they should create separate strategies when rolling out productivity-enhancing AI tools than when deploying business-case-driven AI solutions, according to new research from the Massachusetts Institute of Technology.

So-called AI tools and AI solutions offer two distinct opportunities for organizations interested in gen AI and may require separate approaches to deployment to succeed, according to a recent MIT Center for Information Systems Research (CISR) briefing, “Managing the Two Faces of Generative AI.”

The key to getting value from both AI tools and AI solutions is to deploy them in the right way with the right expectations, the report’s authors say. The research brief was driven in part by IT leaders’ questions about why they aren’t getting the same value from gen AI as they have from data and analytics technologies in the past decades, says Barbara Wixom, principal research scientist at MIT CISR.

One reason may be because they have taken a one-size-fits-all approach to AI. Another may be not recognizing that rolling out certain AI tools can breed familiarity, laying the groundwork for bigger advances with more ambitious initiatives to come.

AI tools require training

Breaking down gen AI use in the enterprise today, the MIT CISR report differentiates AI tools such as ChatGPT, Adobe Acrobat AI Assistant, and Microsoft Copilot, which enable productivity enhancements, from more complex and strategic AI solutions aimed at generating financial returns by changing processes and systems at scale.

More than just productivity plays, AI tools help employees get comfortable with using AI, Wixom says. As such, IT leaders need to see tools such as AI assistants and copilots “as really important mechanisms for building their data democracy,” she adds.

“The more their employees and workforce are using these generic tools and becoming more comfortable and creative in using AI, that’s actually going to uplift your employee skills and also lead to more innovation happening on the solution side,” she says.

Guidelines, however, are important especially with this tier of AI offerings because some may repurpose the company data that’s entered. Moreover, training is also essential in order to ensure AI assistants and copilots can be effective gateway technologies to more sophisticated AI solutions down the line.

“There’s value in that kind of tinkering and experimentation on the employee level, but you want to do it safely,” says Nick van der Meulen, a research scientist at MIT CISR. “A lot of these generative AI tools are available in the public domain, so you want to make sure that you have some good guidelines, maybe even that you have a few select tools from chosen providers that you trust.”

Many CIOs in the early days of AI copilots found themselves not entirely sold on the promise of the tools. This was in part because their full effects may be hard to calculate, given that employees don’t often track the time savings. One executive the researchers interviewed for the report suggested AI tools are productivity “shaves,” because they save users a few minutes on each task by summarizing documents or by helping to draft an email, for example.

AI solutions need strategic plans

In contrast to AI tools, AI solutions address strategic business needs, the researchers note. A large language model (LLM) used by a contact center to process the content and tone of conversations and provide real-time coaching to agents is a prime example. Organizations deploying AI solutions should create formal and transparent AI innovation processes and build guidelines for AI development that prioritize competitive differentiation through customization, the research note recommends.

“To avoid falling into what one executive called a ‘gen AI laundry list mentality,’ organizations need clear governance structures, early and consistent stakeholder engagement, and a focus on scalable solutions,” the research report says. “The executive’s organization created a senior-level working group to guide its gen AI initiatives, tapping diverse sources like hackathons and external consultants to surface stakeholder ideas for GenAI solutions.”

In some cases, the value of AI solutions can become evident sooner than the value of AI tools, Wixom says.

“If you are working in a content-heavy, expertise-heavy, document-heavy industry, with all that unstructured data, then there are incredible opportunities that exist,” she says. “They’re going to have their golden use cases that they address best.”

Customized responses

Dhaval Gajjar, CTO of SaaS text marketing platform Textdrip, agrees that these two types of gen AI implementation require different strategies.

For example, successful use of AI tools, which tend to be easier to deploy, hinges on user training, says Gajjar, also CEO of Pranshtech, a website and mobile app development firm. “Standardization by vendor should go along with guidelines and best practices for their effective use,” he says, echoing recommendations from MIT CISR.

Solutions like AI-driven fraud detection or predictive analytics systems are more complex, he adds. “For my part, any AI solution would require a structured and formal approach to the launch,” Gajjar adds. “It will therefore take cross-functional collaboration to deliver this value in scale, with rigorous testing and clear governance.”

The division between AI tools and solutions is more of a practical framework for deployment than a technical division, adds Moe Asgharnia, CIO of accounting and consulting firm BPM.

“It is important to remember that the underlying technology is the same,” he says. “Where it differs is how they are applied and used. Where AI tools like chatbots or writing assistants enhance productivity with specific tasks, AI solutions such as an AI-powered customer support platform are designed to transform an entire workflow using multiple AI capabilities.”

Asgharnia agrees that AI tools such as copilots can lead to employee acceptance of more complex solutions. “AI tools can deliver quick wins in terms of productivity and efficiency, and they can act as a steppingstone for more complex AI implementations in the future,” he says. “By building familiarity and comfort with these tools, companies can create a stronger foundation for deploying larger AI solutions down the line.”

Both types of gen AI technology can be useful if deployed correctly, he adds.

“The difference in the scope and complexity of each, directly impacts how they should be rolled out or implemented,” Asgharnia says. “In both cases, organizations should measure the success by both immediate impact and how well these tools and solutions align with the long-term business goals.”