Many enterprise IT leaders are downstream consumers of AMD AI chips — racking them in data centers, provisioning them for workloads and fine-tuning models against them.

As a “customer zero” for AMD, Hasmukh Ranjan engages with AMD technologies long before they reach enterprise data centers. As the company’s senior vice president and CIO, he and his team test products and anticipate customer needs before market release.

“We are one of the early adopters of new solutions,” Hasmukh says. “And as they are built, we continue to test with the research and development team.”

Hasmukh says this customer-zero approach , in which a company runs on its own products before they reach market, is a meaningful advantage for AMD. His team certifies new client products, tests AI PC use cases and returns first-hand feedback to engineering before anything ships.

The strategy shapes how Hasmukh thinks about a broader challenge facing today’s CIOs: how to move from early AI experimentation to enterprise systems that meaningfully drive the business.

For Hasmukh, success requires more than compute cycles; the winners will be companies that redesign core business workflows from the ground up. His conviction is reinforced by the customer-zero focus at AMD.

“The enterprises that will take the value of AI to the next level are the ones who have built autonomous systems to solve complex business workflows,” Hasmukh says. “AMD is right at the center of that.”

Forbes interviewed Hasmukh to learn how his team is deploying AI across AMD, from designing the data infrastructure to embedding AI into enterprise applications and scaling adoption across the workforce.

The Foundation: Building The Data Platform

According to Hasmukh, enterprises eyeing AI’s full potential must begin by building a foundation to support their business long term. While IT conversations often start with the LLMs (large language models), Hasmukh starts further down the stack: hardware first, then the enterprise data infrastructure and then the models themselves.

That order reflects how dramatically the data center has changed. In just a few years, server clusters have been reshaped for the GPU era. “The entire hardware data center has transformed completely,” Hasmukh says, as AI workloads demand higher levels of compute, storage and networking performance.

Then comes the data layer. At AMD, R&D activity has produced what Hasmukh calls an “explosion of data,” and like every enterprise, AMD currently works with just a fraction of it.

To build autonomous systems, more enterprise data must be analysis-ready, and the infrastructure supporting it needs the compute, memory and network performance to keep pace with AI workloads. To that end, Hasmukh and his team have built their own internal data platform, partnering with leading vendors and leveraging open and interoperable technologies that have matured in recent years.

“We need to have a very strong foundation around data platforms,” Hasmukh says. “We are very thoughtful in making sure our data architectures, computer architectures and network architectures are in harmony.”

Only when the hardware and data elements are in place can the enterprise turn its attention to LLMs.

Deployment: Embedding AI Into Applications

Once the foundational hardware and data layers are in place, companies need to decide where to apply AI and which LLMs best fit their needs. At AMD, deployment is structured around a clear principle: Simplify enterprise workflows to a single click.

Hasmukh calls this his “click series,” a set of initiatives designed to convert complex, expensive enterprise workflows into single-click actions for everything from provisioning engineering resources to closing the books, processing returns and managing claims.

At AMD, this approach is embedded in enterprise applications that bring AI directly into core business workflows. The most consequential of these initiatives is click-to-compute, where AMD uses AI to optimize how compute resources are allocated across chip design workloads.

Compute is the largest IT expense at AMD, Hasmukh explains, and it’s a resource chip designers consume in massive, unpredictable bursts as products move through their lifecycles.

AMD uses AI to better predict and distribute compute across teams, matching resources more closely to demand and cutting idle capacity. The result is a more efficient use of internal resources without expanding infrastructure. “With one click, any R&D person should have access to as much compute as they need for their job,” Hasmukh says.

The company’s early investment in click-to-compute is paying off. As high-memory servers grow more expensive, Hasmukh says the optimization work the company began years ago is offsetting costs many companies now face in real time.

“High-memory compute resources are extremely expensive in today’s market,” he says. “Acquiring equivalent capacity from third-party providers can be very costly.”

Scale: Enabling The Digital Workforce

One of the trickiest challenges in enterprise AI isn’t building the technology but driving adoption across the organization.

At AMD, Hasmukh describes AI adoption as a progression , from assistive tools that help employees work more efficiently to sophisticated systems that can take action, automate tasks and ultimately operate with more autonomy.

Today, many companies remain focused on assistive AI tools that support employees via chat interfaces and productivity apps layered onto existing work systems. In some parts of the business, where workflows are particularly repetitive or operationally intensive for instance, companies are starting to experiment with autonomous AI systems capable of handling defined tasks or orchestrating larger workflows.

The real productivity gains come when those systems run end-to-end at scale with minimal supervision, Hasmukh says. He closely tracks the “coverage ratio,” or the number of employees supported by each IT team member.

“At some point, for every 20 employees, there was one IT person,” Hasmukh explains. He expects this ratio to improve over time, as automation empowers IT teams to support a larger, more complex organization without scaling headcount at the same pace.

Hasmukh observes a familiar pattern in AI’s potential to reshape computing. Decades ago, the internet redefined IT. Cloud computing transformed it again. Now, AI is beginning to change what computers themselves are capable of.

“The IT job will shift more towards data, data management and management of agents,” Hasmukh says. “The business opportunity extends beyond simply doing the same work with fewer people. It’s about empowering employees to take on more interesting, fulfilling, higher-value work — the kind that can actually move the business and enhance employee experience. That’s how the promise of AI is playing out at AMD.”

Explore how AMD helps enterprises build, deploy and scale AI