5 ways to make agentic AI a competitive advantage

When it comes to creating an AI-driven organization, enterprise leaders may feel like they’re steering a cruise ship while startups are navigating around them in kayaks, MIT Sloan School of Management senior lecturer Paul Cheek said.

“Emerging firms can start and scale faster than we’ve ever seen before,” he said. “I encourage the big ships to deploy a dinghy to explore a cove, figure out if they should shift in that direction, and then upgrade the dinghy to a speedboat that can operate at the same speed and scale of the AI native startup that’s starting from scratch.”

Cheek and Paul McDonagh-Smith, a visiting senior lecturer at MIT Sloan, led a recent webinar highlighting how enterprises can turn agentic AI into a competitive advantage. The pair will teach an upcoming MIT Sloan Executive Education course on the same topic. 

They shared five key considerations for enterprises looking to make the most of agentic AI.

Embrace agentic AI, even if it means redesigning work. Cheek said that the workforce is experiencing a rapid shift in which AI agents are becoming dominant actors. Those agents will apply system instructions, their own knowledge, and a range of third-party systems to inputs to produce outputs — acting similarly to humans, but with different constraints. 

The biggest risks organizations face won’t stem from replacing humans with AI, Cheek said, but in failing to embed AI in functional areas. The established firms doing this “are getting compounding gains compared to you and have the potential to steal market share every single day.”

To realize these gains, McDonagh-Smith said, organizations must consider redesigning workflows, the workforce, and the workplace. “It’s through rolling our sleeves up that we actually build AI literacy and fluency,” he said. 

Set realistic goals, and understand limitations. Of course, rolling up our sleeves means getting started. Too many organizations falter by “setting their North Star as some version of perfection,” McDonagh-Smith said. Instead, the target should be a marked improvement on existing benchmarks, which obviously will vary by organization and use case.

Additionally, as organizations build AI fluency, they need to understand where tools fall short. “That’s a signpost to the human capabilities the organization needs to invest in and develop,” McDonagh-Smith said. Successful leaders won’t simply adopt AI; they’ll adapt their organizations around it and orchestrate human-machine relationships in a way that will help them move forward and deliver results, he said. 

Agentic AI: Business Implications and Applications

In person at MIT Sloan

Register now

To get out of the sandbox, rethink the role of governance. Many AI pilots fail to move forward due to a half-finished data strategy, McDonagh-Smith said. Coordinated development efforts require data governance, which should be viewed not as an obstacle to progress but as a well-timed tap on the brakes.

“If you’re driving around [an obstacle], if you want to be super effective, you have to hit the brakes at the right time and use the guardrails as an opportunity to accelerate better outcomes to get better results,” McDonagh-Smith said. With a governance framework in place, organizations should see AI projects move from the testing sandbox into production.

Rearrange the org chart to reflect agentic AI’s role too. As solutions emerge from the sandbox, Cheek said, leaders will want to look at how AI agents slot into org charts. It’s imperative to formally document who will oversee updates, training, token use, embedding, bias control, and regulatory compliance.

“There must be a line to governance, oversight, and accountability,” Cheek said. “At the governance layer, there are humans who are responsible for taxes, legal or regulatory considerations, and things that go wrong. They have the potential to reap the reward — but they also wind up in the position where they’re on the hook.”

Know that AI adoption won’t mean business as usual. A new org chart is important, but it should follow a broader audit of the organization that considers how AI and other technologies can “supercharge” human users, Cheek said. 

Organizations should evaluate and prioritize AI use cases based on risk, feasibility, business impact, and “human vs. AI,” which considers legal and ethical considerations as well as the organization’s core values. “Is that something that we want to be delivered by AI — or do we want to maintain that human relationship?” Cheek said. 

All of this requires new ways of thinking. “We have to be careful not to just try to reengineer the same process using AI agents or an agentic system that we would do already with humans,” Cheek said. “We want to make sure it’s productive and additive to the organization.”

Agentic AI: Business Implications and Applications is offered through MIT Sloan Executive Education. The course will be taught in person on the MIT campus July 27–29 and online Nov. 3–6.

Paul Cheek is a senior lecturer at the MIT Sloan School of Management and a global expert in innovation-driven entrepreneurship. He is a serial tech entrepreneur as well as an educator, a software engineer, and the bestselling author of “Disciplined Entrepreneurship: Startup Tactics.” He was MIT’s first “hacker in residence” and has since taught, mentored, and advised thousands of entrepreneurs around the world. His upcoming book, “No One Works Here: How AI-Driven Enterprises Are Dramatically Redefining Business, Leadership, and Competition,” will be released Aug. 24.  

Paul McDonagh-Smith is a visiting senior lecturer at MIT Sloan. In his research and teaching, he creates key intersection points between technology and business. He specializes in translating computer and data science into measurable business value that evolves organizational capability, transformation, and strategy. He is an advisor to NASA Goddard Space Flight Center and provides digital transformation, business model, and strategy guidance to a range of organizations and international government departments.