Gen AI looks easy. That’s what makes it so hard



GettyImages 1271693370 e1715701526578

This commentary is from McKinsey & Company, a Fortune Global Forum Knowledge Partner. Rodney Zemmel is a senior partner in McKinsey & Company’s New York office and global leader of McKinsey Digital. He is a coauthor of Rewired: The McKinsey Guide to Outcompeting in the Age of Digital and AI.

The natural-language capabilities of generative AI are so user-friendly that even CEOs, who typically are not early adopters, experiment with it. Less than a year after gen-AI-based tools burst onto the scene in late 2022, one quarter of C-suite executives were already using it. 

The widespread interest in gen AI has created a massive wave of use cases and experiments—and there’s the rub. Such efforts are relatively easy to launch but can chew up resources without creating much value. 

To escape from this pilot purgatory, the priority must be to connect gen AI to business outcomes. Here are four ways CEOs can make that happen.

Focus on something important. When gen AI is diffused across a range of pilots, it can look like a technology in search of a problem. However, meaningful change happens when gen AI is directed at a domain that is big enough to make a difference, such as a customer journey or a functional area. For example, McKinsey worked with financial services giant ING, which created a gen-AI-powered solution whose language and data capabilities enabled it to respond to customers with precise solutions. That improved service, while releasing agents to deal with more complex issues. 

Develop a business-led technology roadmap. Gen AI comes with so many unknowns that it requires a central team, composed of all relevant competencies, including risk, legal, compliance, finance, human resources, and strategy, to develop protocols and standards. That effort has to begin with the CEO and C-suite agreeing on what needs to be done. The CEO then needs to work closely with the chief information or chief technology officer (CIO or CTO) to translate that commitment into a specific roadmap that will direct how the company proceeds. Of course, transforming a domain isn’t just about gen AI applications; process digitization and other forms of AI will also be involved. If the applications are built around reusable modules, they can apply to many kinds of future problems too.

Build a talent bench. Building up a talent bench is a non-negotiable. Partnering with external providers, such as senior engineers who have already built gen AI products, can be an important part of a gen AI strategy. But just as much or more focus needs to be on in-house talent—and not just among tech teams. Those on the business side also need to have a sense of what gen AI can, and cannot, do. 

Companies can upskill their data engineers, for example, to learn multimodal processing and vector database management, while data scientists can develop prompt engineering and bias detection skills. And it is critical to retain these experts. A recent McKinsey survey of almost 13,000 employees found that 51% of gen AI creators and heavy users plan to leave their roles in the next three to six months. Compensation will always be important, but talented people are more inclined to stay if they can grow their skills, work on meaningful initiatives, and have opportunities for advancement. 

For example, McKinsey worked with Singapore’s DBS bank, completed a successful digital transformation, and found the winning ratio was 80% of talent insourced, and 20% outsourced. This combination allowed the organization to move more quickly and make decisions faster. The principle is clear: Greatness cannot be outsourced. 

Focus on what matters. Businesses are using up a lot of oxygen deciding which large language models (LLMs) to use. But all the new-generation LLMs can do amazing things. It’s more important to put the right effort in the right places, such as context engineering, security, governance, and ensuring that technology upgrades support gen AI at scale. This may sound obvious, but many pilots have been set up in protected environments that don’t reflect the realities on the ground. 

Improving the data needed for specific solutions can have an enormous impact on the quality of output. So, too, will investing in an orchestration engine: Gen AI requires many interactions and integrations between models and applications. An application programming interface (API) gateway is an important element of this orchestration capability because it mediates access and enforces compliance. A good API will not only help to reduce risk but also give teams confidence. 

The gaps in performance between leaders and laggards in digital and AI technologies are widening, with the leaders seeing much better financial performance. If that trend spills over into gen AI, the laggards could fall even further behind.

It is certainly possible to capture real value from gen AI, but is more difficult than meets the eye—in part because it seems so easy. It just isn’t.

The opinions expressed in Fortune.com commentary pieces are solely the views of their authors and do not necessarily reflect the opinions and beliefs of Fortune.



Source link

About The Author

Scroll to Top