Five essentials for enterprise agentic AI

A practical guide to smarter agentic AI for enterprises

Agentic AI—autonomous systems that perform complex tasks—are reshaping the future of business operations. In a recent episode of the Deloitte On Cloud Podcast, Gary Arora, managing director and chief architect for cloud & AI solutions at Deloitte, spoke with Brent Collins, vice president of enterprise AI strategy at Intel, about how enterprise organizations can responsibly deploy, govern and scale AI agents.

Their conversation revealed five actionable strategies for unlocking agentic AI value while managing risk.

1. Specialized AI agents vs. general AI: complement, don’t compete
The AI landscape is often described as a spectrum: at one end, the pursuit of artificial general intelligence (AGI); at the other, the rise of federated, task-specific agents. Collins emphasized these models are not mutually exclusive.

“AI for me is very similar to a human function,” Collins said. “You see generalists and you see specialists. Sometimes you just need something that’s very simple and very good at a specific function.”

Choose solutions—large general models or efficient specialists—based on business needs, complexity and cost. Don’t’ select a “sports car” when a reliable “sedan” will do.

2. Foundations: robust orchestration and data governance
Agentic AI thrives on strong orchestration and clear standards. Ensuring interoperability, privacy and security are non-negotiables for building trust and compliance.

“You need solid standards,” Collins said about connecting data. He also noted that compliance becomes more complex as enterprises federate data.

“If somebody tells you ‘We’ve got [enterprise data management] figured out,’ I would ask them to check their math...data atrophies over time.”

He recommended regularly validating and updating data throughout the agent workflow to maintain quality and regulatory standing.

3. Balance innovation with risk
Agentic AI brings new risks—prompt injection, data leakage and more. To drive innovation and progress, Collins urged leaders to iterate quickly, learn fast and adapt controls as solutions mature.

“Take advantage of the liberty right now to innovate,” he said. “Fail fast if you need to and move on. Over time, you’ve got to lock things down a lot more.”

Adopt pragmatic guardrails to support safe experimentation and scalable rollout.

4. Infrastructure: let workloads guide technology choices
Let specific workload needs, not hype, drive your decisions—whether that means CPUs, GPUs or AI-specific chips.

“I would rather be talking about the workload first, and the needs of the business, and then working your way into whatever makes the most sense.”

Factor in AI scalability, cost, and operational familiarity as you build an AI-ready infrastructure.

5. Rethink operations: AI agents demand new models
Orchestrating hundreds of autonomous agents transforms how you manage, monitor and govern digital work. 

“The way you manage agents is going to be similar to managing people,” Collins said. “Value stream mapping [reveals] how workflows should work versus how they actually work.”

Look for opportunities to optimize value streams, expand access to insights, and reshape how decisions are made across the enterprise. This may require a mindset shift for technology leaders.

“When you democratize intelligence...you’re going to be able to do so much more if you just open up your mind and really take advantage of that.”
 

Deloitte and Intel: leading the agentic AI journey

Deloitte and Intel combine deep experience in operational transformation and technology innovation to help clients unlock the full value of agentic AI.

Intel advances secure, scalable hardware and interoperability standards while Deloitte delivers operational readiness, governance and strategic alignment. Together, they guide organizations with tested methods for risk management, compliance and business-driven AI adoption.

As Collins concludes, bold leaders who embrace innovation, align technology to business outcomes and follow frameworks—supported by industry leaders like Deloitte and Intel—will lead in the era of agentic AI.