Three Reasons Your Chief AI Officer May Soon Look More Like Your Head Of HR

Kazuhiro Gomi, NTT Research CEO. Leads research in physics & informatics, cryptography & information security, medical & health informatics.

gettyIn the past, CIOs adopted artificial intelligence (AI) in the same ways they approached any technology: Evaluate the platform, secure the data, integrate the systems and deploy the software.​

Agentic AI will drastically change this. ​

CIOs aren’t just deploying a single AI agent; they’re deploying specialized teams of AI agents designed to reason, collaborate and complete fairly complex business practices simultaneously. This is transforming enterprise AI from a technology initiative to an organizational one. ​

These new AI workforces will leave CIOs and chief AI officers (CAIOs) asking questions similar to the ones hiring managers answer. How many AI agents should perform a task? How should responsibilities be divided? Which agents should collaborate? What training should be provided, and how do we determine the agent is ready? When should humans intervene? Who remains accountable when something goes wrong?​

A recent study from my organization, NTT Research Physics of Artificial Intelligence (PAI) Lab, and Harvard University’s Center for Brain Science illustrates why these questions matter. After studying what researchers called a “Society of AI Agents” via a simple Flag Game, they found that adding more AI agents alone didn’t improve task performance. Instead, the performance of multi-agent AI systems depended on the same principles that influence human-based teams, such as organizational structure, communication, diversity and human guidance.​​

As enterprise AI matures, the CIO or the CAIO may take on a role resembling that of chief human resources officers.​​

Today, the AI conversation focuses on how to select the best LLM. Tomorrow, that conversation will be how to build the best AI organization. ​

While one agent analyzes contracts, another agent may monitor cybersecurity threats. Yet another may prepare executive recommendations. Rather than choosing the right LLM, the challenges will evolve to understanding how agents can work together. ​

The research cited above demonstrates why this distinction matters. To accomplish a goal by multiple AI agents working collaboratively, adding more AI agents initially improved performance.

However, after reaching the optimal point, problems emerged. Instead of increasing effectiveness, AI agents had communication overhead and competing viewpoints. Another interesting research finding is that diverse teams of AI agents combining different skill sets, and placing the right model in the right role, had a significant effect on performance.​​

These results imply team formation matters, even in the AI agent world.​

Agentic AI will require an assignment process similar to what organizations use when hiring, training and evaluating employees. ​

Whether they’re handling financial analysis, software development or customer support, organizations should establish standards for readiness, performance measurement and oversight structure before deploying AI agents.

Organizations must consider how much training AI agents should receive, how much autonomy they should have and where humans fit in this collaboration.​

These governance concerns resemble organization design and workforce management. As AI becomes more embedded within the enterprise, CIOs or CAIOs will increasingly shape organizational processes, not just technical architectures.​

Although AI agents are increasing their capabilities, the responsibility for business outcomes remains with humans.

Human employees operate within legal, ethical and organizational accountability frameworks. When and if their performance is poor or critical mistakes are made, the particular human employee takes responsibility.​

AI agents have no such constraints. Organizations may be able to retrain, or even replace, an AI system, but cannot make it accountable for its mistakes.​

As a result, human oversight remains essential. Successful enterprises will understand this and redefine where human judgment is most valuable, rather than removing humans from decision-making entirely.​

Deploying a single AI agent for an isolated task may be relatively simple. However, the next generation of enterprise AI will be made up of hybrid teams of agents and humans working together.

Some tasks may be entirely automated, while others will require more human involvement. Overall, most tasks are expected to require collaboration between humans and AI agents.​​

The responsibilities of CIOs or CAIOs will evolve into architects of hybrid organizations of AI agents and humans, which may be outside the scope of traditional definitions of CIOs and CAIOs. This new role will involve determining how work flows between humans and AI agents, when AI’s decisions require human intervention and strategies for sharing institutional knowledge across both.​

A competitive advantage has always been given to organizations that have upgraded their operating models, not simply those that adopted the technology first. This has been true during every major technology transition, and agentic AI is no different.​

Some organizations will aggressively experiment with AI workforces. Others will be more cautious, opting to learn from the early adopters before scaling.

Neither strategy is inherently right or wrong, but it would be risky to assume transformation can wait indefinitely. Competitors learning how to integrate AI into their organizations will move forward, while those who wait too long will stand still.​​

The CIOs or CAIOs who succeed won’t simply ask, “Which AI model should we deploy?”​ Instead, they’ll ask:​

• How many agents should be engaged?

• How do we divide tasks, choose AI agent types, create training standards and measure performance?

• Where should humans remain involved?

These are all organizational questions, not purely technology questions. In the meantime, of course, CIOs and CAIOs need to ensure the financial advantage of the new organizational structure while addressing security concerns.​

As CIOs or CAIOs prepare for the next era of enterprise AI, they will probably discover that their leadership strength doesn’t come from technology but from human resources.​

Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?