Why Agentic AI Projects Fail | Key Insights from EPI-USE | Jon Myer Podcast

About this episode

Most agentic AI projects don't fail because the technology doesn't work. They fail because teams start with the technology instead of the business problem. In this episode, Jon Myer sits down with Christopher and Diego from EPI-USE to unpack why so many organizations get stuck in "pilot purgatory" and what it actually takes to move an AI agent from demo to production. They walk through the three fundamentals every agent needs before deployment: context, tooling, and instruction, using the analogy of onboarding a new human employee. The conversation also draws a sharp line between generative AI (the "vending machine," reactive and prompt-driven) and agentic AI (the "sous-chef," proactive and empowered to act), and makes the case that governance isn't a blocker to AI adoption but the thing that actually enables it. Christopher and Diego share a case study of a bank that scaled to 350 agentic AI use cases, not through moonshot projects, but by targeting "boring" processes like invoice reconciliation, IT support, and HR onboarding. They close with a candid look at the autonomy-versus-accountability tradeoff, and why human-in-the-loop and human-on-the-loop design need to scale with risk.

Key Takeaways:

Start with the business problem, not the technology. Projects that lead with "let's use AI" instead of "let's solve this problem" are the ones most likely to stall out or fail.
Before deploying an agent, it needs the same three things a new hire needs: context about the business, the right tools to act, and clear instructions on what success looks like.
"Pilot purgatory" happens when organizations run endless proofs of concept without a real path to production, momentum dies without executive sponsorship and a scaling plan.
Generative AI and agentic AI solve different problems: generative AI responds when prompted (the vending machine), while agentic AI can act proactively on your behalf (the sous-chef).
Governance shouldn't be treated as a brake on AI adoption, done well it's what gives leaders the confidence to actually deploy agents at scale.
The organizations seeing real ROI aren't chasing flashy use cases, they're applying agentic AI to unglamorous, repeatable processes like reconciliation, IT support, and onboarding.
Human-in-the-loop and human-on-the-loop aren't one-size-fits-all, the right level of human oversight should scale with the risk of the task the agent is performing.
Analysts project that up to 40% of agentic AI projects will be abandoned by 2027, underscoring how much of this is still an execution and governance challenge, not just a technology one.

Presented by:

Jon Myer

Chief Content Creator, Myer Media | Host, The Jon Myer Podcast

Jon Myer is the Chief Content Creator for Myer Media and host of The Jon Myer Podcast. He has spent the last 20 years working in technology, including the last 8 years focused on cloud technology.

Early in his career, Jon had the opportunity to teach others about tech, from hardware to software, never expecting it would lead him to create technical content for a wider audience. He posted his first tutorial on YouTube more than three years ago.

In mid-2021, Jon launched his podcast, interviewing guests and giving them space for natural conversation to tell their own stories and he hasn't stopped since.

Christopher Belford

Senior Director, EPI-USE Services for AWS

Christopher Belford is a Senior Director at EPI-USE Services for AWS with over 20 years of expertise in high-volume systems for Government, Financial Services, Retail, and Life Sciences. He has held roles including CTO, CIO, and Group Architect, specializing in system design, ERP integration, and cloud migrations.

Currently leading public sector solution strategy at EPI-USE services for AWS, Christopher drives global initiatives in Smart City, FinTech, and SAP cloud migrations. His accomplishments include designing e-filing platforms for 6 million users, transactional banking APIs, and tax reconciliation systems.

Recognized as an Adobe Ambassador, Christopher's leadership has transformed enterprise systems, improving workflows and customer experiences globally.

Carlos Diego Cavalcanti

CEO & CTO, Valcann, an EPI-USE Company

Carlos Diego Cavalcanti is the CEO/CTO of Valcann, EPI-USE's Latin America cloud services division, where he leads strategy and technology direction for one of the region's most trusted SAP and cloud services partners.

Carlos bootstrapped Valcann from the ground up, building the business on a simple philosophy: you can't scale a relationship. Client trust has to be earned individually, even as the company grows. That approach included launching a university internship program in 2019 to develop local talent and build an engineering pipeline grounded in Valcann's own culture rather than hiring purely for headcount.

When the pandemic hit, Carlos and his team navigated the uncertainty month by month instead of committing to long-term bets, a discipline that carried Valcann through to its 2023 integration into Group Elephant and EPI-USE. He views resilience less as a fixed trait and more as a discipline built through consistent decision-making under pressure, and is now focused on how AI is reshaping the professional services model for firms like Valcann.