As enterprises rush to implement AI across operations, Dell Technologies Systems Engineering Director Eric Leung says most hit the same two walls: soaring costs and unreliable data. Drawing from Dell's own experience, he urges businesses to first ask two essential questions: What assets do we already have? What specific problem are we trying to solve?
"Since we adopted AI, my boss has gotten laser-focused with his requests—I can't dodge anything anymore," Leung quipped during a keynote at AI EXPO Taiwan 2025, hosted by DIGITIMES in Taipei.
Start with cost and data, then define your use case
As a core Nvidia partner and supplier of GB-series servers and AI Factory infrastructure, Dell has seen how AI hype often eclipses real-world hurdles. Leung said that despite overwhelming interest, most companies stall at the same two friction points.
"Look at GPUs," he said. "Nvidia's 5090 is priced at US$1,500, but in Taiwan, the price surged to nearly NT$300,000 (US$9,060). That shows how cutthroat the demand for computing has become."
But pouring money into AI won't pay off if the data isn't up to par. Many companies dump massive datasets into models, only to get back outputs that Leung describes as "barely usable." The cleanup effort alone can kill momentum.
Inventory your assets before investing
To avoid this pitfall, companies should first inventory their internal assets—proprietary data, workflows, and domain expertise. Then comes the critical step: identifying the exact problems AI is meant to solve. Only after that should they evaluate platforms or tools.
Leung highlighted Dell's direct-to-customer model as a strategic advantage over competitors relying on channel sales. That clarity around customer needs gave Dell an edge during the pandemic, enabling agile supply chain responses while rivals scrambled.
AI now powers Dell's end-to-end supply chain, tracking everything from backend component flow to front-end sales trends. "You need both ends of the equation," said Leung, emphasizing that proprietary data has become a crucial edge in the era of public LLMs.
Raising expectations across the organization
AI is also reshaping workflows inside the company. "Everyone thinks AI makes my job easier—it doesn't," Leung joked. "Now our boss expects weekly AI use from every employee. If you're not using it, you get flagged. The pressure's real."
Article edited by Jerry Chen
