Show Notes
Is your AI delivering business value, or are you just paying to generate expensive garbage? In this episode of Let’s Solve IT!, Matt Brown talks with Cecile Kellam, NetApp Senior Solutions Architect, about why successful AI initiatives depend on more than good prompts. They discuss the importance of clean, relevant, secure data; why data ownership and custodianship must involve the business; and how organizations can avoid wasting tokens, budget, and trust as AI projects move from pilot to production. You’ll hear:
Why many organizations are wasting AI tokens on poor-quality data Why great prompts alone cannot overcome bad or outdated data How to identify the right data sources for different AI use cases Who should own data quality, governance, and AI success Why business leaders—not just IT—must be involved in AI initiatives How data integrity directly impacts AI accuracy, trust, and adoption How to build AI initiatives that deliver measurable business value instead of unnecessary costs
If your AI can’t trust your data, why should your employees trust your AI? You are not alone. Let’s Solve IT.
Learn More: IT case studies | NetApp
Connect with us! https://www.linkedin.com/in/cmattbrown Cecile Kellam | LinkedIn
Cecile Kellam on AI tokenomics: why clean data beats clever prompts Episode overview In this episode of "Let's Solve IT!," host Matt Brown of NetApp talks with Cecile Kellam, Senior Solutions Architect at NetApp, about a problem she sees across enterprise AI projects: teams "spending precious tokens on trash." Cecile Kellam defines tokens as the cost of every AI query, and Cecile Kellam explains that this cost covers more than money. Tokens also consume energy and people-management resources, so every prompt an organization sends counts as a real investment. The central argument from Cecile Kellam is direct: clean, secure, well-governed data matters more than a perfectly written prompt. Cecile Kellam points out that a strong prompt still fails when the underlying system pulls from outdated engineering code, incomplete records, or poorly governed HR data. To make the point concrete, Cecile Kellam contrasts AI needs across engineering, finance, and HR departments, since each department queries very different tools and data. Cecile Kellam also warns about the cost of scaling AI from a small pilot to full production. Cecile Kellam describes customers who successfully built digital twins in the cloud, then watched costs turn "astronomical" when they tried to run those digital twins for more than a couple of hours a day. Cecile Kellam ties this risk back to a familiar pattern: companies that went "all in" on the cloud without clear goals and later faced surprise bills. Key takeaways The following takeaways summarize the practical guidance Cecile Kellam shared for IT leaders planning or scaling enterprise AI.
Tokens represent the full cost of AI queries, including financial spend, energy use, and people-management costs, so wasted prompts waste real budget.
A well-written prompt fails to deliver good results when the AI system lacks access to clean, current, and relevant data.
Data accuracy is an ongoing discipline, not a one-time task, because information that was correct years ago can produce wrong answers today.
Data ownership and governance often cause friction at the start of AI projects, so organizations should settle who owns, manages, and uses the data early.
Engineering, finance, and HR departments each require different tools and data access, so AI systems must be built around each department's specific use case.
Line-of-business owners hold the strongest stake in AI success, since they see the return when their teams save time and automate processes.
End-user buy-in improves when leaders position AI as a tool that removes tedious work rather than a replacement for jobs.
Scaling a successful AI pilot to full production can create "disastrous" bills, so
More from Let’s Solve IT! Podcast
Can AI help us find answers faster without replacing the human experience, context, and trust needed to solve technical problems? Recorded at NetApp INSIGHT 2026, host Matt Brown joins NetApp’s Drew Claybrook and Lex Rutledge to explore why people and communities remain essential in an AI-driven world. They discuss how experts turn information into practical guidance, from understanding a customer’s environment to checking whether an AI-generated answer is relevant. They also share what makes technical communities valuable: trusted relationships, shared experience, and the freedom to ask questions. In this episode, you’ll learn:
AI agents can follow instructions, but how do they learn your organization's unwritten rules? Matt Brown welcomes Robert Vaughn, IT AI domain owner at Intel, to explore the “AI intern”: a metaphor for systems that use institutional knowledge and ongoing feedback to understand better what matters. They discuss the difference between AI reduction and synthesis, how risk compounds across agentic workflows, and why human oversight remains essential. Discover why building more useful, trustworthy enterprise AI takes more than connecting bots. In this episode, you’ll learn:
Organizations are collecting more telemetry than ever, but is all that data actually delivering value? Matt Brown sits down with Marty Mayer, Senior Director of Product Management at NetApp, to explore how NetApp turns massive amounts of telemetry into actionable insights. Marty shares how Active IQ uses data from more than 300,000 systems across 24,000 customers to improve support, identify risks, optimize infrastructure, and help customers get more from their NetApp investments. They also discuss the importance of continually evaluating what data is worth collecting, how Value Insights can help quantify business outcomes, and how AI is moving telemetry beyond recommendations toward automated action. In this episode, you’ll learn:
A strong data strategy isn’t built when the project begins; it’s built in the preparation that comes before it. In this episode of Let’s Solve IT!, Matt Brown is joined by Spencer Hamons, Executive Architect at NetApp, who explains why organizations must measure, classify, and understand their data before investing in infrastructure. The discussion covers how to identify hot, warm, and cold data, why storage performance matters more than ever as SSD prices rise, and how telemetry and observability help organizations make smarter infrastructure decisions. In this episode you’ll learn:
The hypervisor landscape is changing. Licensing costs are rising. AI is reshaping enterprise IT. For many organizations, it feels like two major waves colliding at once. In this episode of Let’s Solve IT!, Matt Brown sits down with Spencer Sells, Vice President of Global Technology Alliances at NetApp, to discuss why today’s virtualization disruption shouldn’t simply be viewed as a cost crisis—it should be viewed as an opportunity. Drawing inspiration from The Odyssey, they explore how IT leaders can navigate uncertainty, make more intentional technology decisions, invest in their teams, and build platforms that will support the next decade of innovation. The conversation also examines why successful IT transformations depend as much on people as technology and why organizations that neglect training, experimentation, and culture risk falling behind. In this episode you’ll learn: