The Enterprise’s New Hire Is an AI Agent

Case in point: Harvard Business Review argues that many companies are thinking about agentic AI the wrong way. The real challenge is not just adopting a new technology. It is redesigning work so people and AI agents can operate together in a clear and practical way. The article says many firms are still far from that goal, even as interest in autonomous AI grows. Its core point is simple: Companies will get more value from AI when they start managing agents more like co-workers, with defined roles, limits and accountability.

The article lays out a straightforward framework for doing that. HBR says every AI agent should have a job description that spells out what it is responsible for, where its authority stops and when it must ask for human input. It also says agents should be designed to take on tedious, repetitive work that frustrates employees, not just work that looks easy to automate on paper. From there, companies should review each agent on a regular cycle, using clear measures that look beyond accuracy to include reliability, timeliness and real process outcomes.

The article also makes the case for stronger human oversight during the rollout phase. It recommends giving every AI agent a human supervisor and treating new agents more like interns than full-time hires until they prove they can perform well in a real business setting. It also suggests giving each agent a clear name so employees can talk about its role in concrete terms and understand who or what is shaping a decision. The broader takeaway from Harvard Business Review is that enterprises will not succeed with agentic AI by plugging in tools alone. They need to build structure, oversight and trust around how the work gets done.

After Onboarding

Once companies learn how to onboard AI agents and define their roles, the next opportunity is bigger: rethinking how work gets done across the enterprise. EY says agentic AI gives companies a chance to move beyond narrow automation and redesign full processes around systems that can interpret context, make decisions and adapt as conditions change.

The article argues that this is where earlier automation efforts often fell short. Traditional tools could handle repetitive steps, but they did not change the workflow itself. Agentic AI, by contrast, can work across systems, data sources and teams, which gives enterprises a way to rethink operations more deeply and pursue far larger efficiency gains.

EY also says companies are already seeing practical uses emerge in areas such as compliance and customer service. AI agents can help pull together information from multiple systems, prepare reports, spot data issues and handle routine customer requests so employees can focus on harder problems.

But the article is clear that these gains come with real risk. Because agentic AI systems can behave in less predictable ways than traditional software, companies need stronger testing, closer monitoring and better governance before putting them into production. EY’s main point is that the biggest payoff will not come from scattering AI across small tasks, but from going deep in a few areas where businesses are ready to redesign work from the ground up.

The Enterprise Data Angle

Once companies start redesigning work around AI agents, the next challenge is making sure those agents can actually use enterprise data safely and in real time. Oracle’s article argues that this is becoming a central issue for the enterprise. The company says agentic AI will be far more useful when it is built close to the data itself, rather than relying on complex data movement between systems.

Oracle positions its latest database tools as a way to help companies build and run AI agents that can draw on live business information, work across different data types and scale for production use. The broader point is that enterprise AI will depend less on isolated models and more on how well companies connect AI to the systems where their core data already lives.

Oracle also puts heavy emphasis on security and control. The article says enterprises need stronger protections as AI agents begin accessing sensitive data and acting on behalf of users. Its new features are designed to limit who and what an agent can see, keep data inside private environments when needed and reduce the risk of exposing information to outside models.

Oracle also argues that open standards and flexible deployment matter, so companies can run these systems across cloud, hybrid and on-premises environments without getting boxed in. The takeaway is that enterprise agentic AI will be shaped not just by smarter agents, but by better data architecture, tighter guardrails and systems that let businesses use their own information with more confidence.

For all PYMNTS AI coverage, subscribe to the daily AI Newsletter.