Agentic AI isn’t just a new technology, it’s the future—and it’s already here. Systems based on agentic AI can plan, act, and learn on their own, thereby accelerating efficiency, innovation, and growth. But to seize this potential, companies must navigate uncharted ground, managing a technology that combines capabilities typical of tools with others characteristic of people. The task involves reimagining workflows, roles, governance, and learning, and devising investment strategies to provide the flexibility necessary to create value with agentic AI.
This insight comes from the ninth annual global research study on artificial intelligence and business strategy conducted by MIT Sloan Management Review and BCG. A total of 2,102 respondents spanning 21 industries and 116 countries participated in the survey, and the results reveal that agentic AI is rapidly emerging in enterprises: 35% of organizations have begun to use it; and another 44% plan to join their ranks soon. On the other hand, 47% indicate that they don’t have a strategy for what they are going to do with AI.
The dual nature of agentic AI further complicates the situation. Blending qualities of tool and human, agentic AI doesn’t fit neatly into traditional management frameworks. Managing it purely as a tool or purely as a worker creates four strategic tensions: scalability versus adaptability, investment versus employment, supervision versus autonomy, and process retrofitting versus process reimagining. Yet few organizations are restructuring to manage the competing demands.
Although substantial effort is necessary, companies that embrace agentic AI’s dual nature put themselves in a position to unlock the technology’s full potential. The path forward starts by clarifying the underlying value thesis. Agentic AI systems have the ability to appreciate with use, getting smarter and more capable as time passes. Companies that treat agentic AI solely as a cost-saving tool risk missing its greater value as an engine for scalable learning, adaptability, and innovation. So organizations should first ask, “What are we trying to optimize for?” Then they should take five practical steps:
- Redesign work around agentic-first workflows. Rather than automating isolated tasks, reimagine processes to integrate agentic AI’s tool-like scalability and human-like adaptability. Almost two-thirds of agentic AI leaders expect to see changes to their operating model. (See the exhibit.) A best practice: build workflows that can shift smoothly between efficiency optimization and innovative problem solving—rather than focusing on one or the other.
- Upgrade governance and redesign decision rights. Since agentic AI systems fall somewhere between tools (owned and predictable) and people (autonomous and requiring supervision), governance must be adaptive. Our survey results indicate that 58% of leading organizations anticipate changes to governance and decision-making rights. Companies should create governance hubs and equip them with enterprise-wide guardrails. But they should also treat decision rights as dynamic—tailored by workflow—and build system-level oversight mechanisms to monitor and manage varying degrees of AI autonomy.
- Redefine roles for human-agentic interaction. When AI agents coordinate workflows, traditional spans of control widen. This calls for a flatter organization (45% of agentic AI leaders expect to see a reduction in middle management layers) in which managerial roles evolve to orchestrate hybrid human-AI teams. With this in mind, companies should consider creating dual career paths for generalist orchestrators and AI-augmented specialists. Both are critical.
- Prioritize continuous learning for humans and for AI agents. Organizations that fail to care for their AI agents risk seeing their systems become outdated, inaccurate, or unpredictable. Employee training programs should cover not only how to use AI systems, but also how to supervise, critique, and direct them. Like their human counterparts, agentic AI benefits from continuous learning and support. This can take different forms, including fine-tuning agents through retraining as more data becomes available and conducting AI performance reviews to evaluate the system’s accuracy, adaptability, and bias—and when necessary, to trigger intervention.
- Anchor investment decisions in value. Agentic AI can scale exponentially, evolve quickly, and create value in multiple ways simultaneously. As such, investment strategies should combine scale, adaptability, and focus. Agentic AI leaders ensure that investments remain anchored to the value thesis, whether it be efficiency, innovation, revenue growth, or some combination of these. They also build adaptable investment plans that possess the flexibility to support ongoing training over the AI agent’s life cycle, as well as to respond to shifts in technology and evolving infrastructure needs.
A mix of tool and collaborator, agentic AI is unlike any other technology. Businesses that recognize its dual nature and align their processes, governance, talent, and investments accordingly can capture its full value. Although agentic AI will inevitably challenge established management paradigms, with the right strategy, imagination, and execution, it won’t just offer promise—it will deliver impact.
