Everyone is asking “can AI do this?” But that’s the wrong question. The right questions are: does it have the right knowledge, the right tools, and clear enough instructions to do the job? And is your enterprise actually ready for what comes after?
In this episode, Jon Myer is joined by Christopher and Diego from EPI-USE — two experts in AI architecture, governance, and enterprise implementation — for a frank conversation about why so many agentic AI projects fail, what the successful ones have in common, and why the real bottleneck is almost never the AI itself.
Topics covered:
Why most companies are implementing AI for the wrong reasons and asking the wrong questions
The difference between generative AI and agentic AI — vending machine vs. sous chef
Pilot purgatory — why AI projects get stuck and never make it to production
Why the AI part is usually the easiest part and the foundation is the heavy lift
How governance becomes the enabler not the obstacle when done right
The difference between human in the loop and human on the loop — and why it matters for design
A real-world case study — a bank that ran 350 agentic AI use cases and what actually worked
Do we want true autonomy or accountability? The answer is both
⏱️ Timeline
0:00 — Introduction — Reframing agentic AI as an evolution that demands first principles thinking
0:50 — What companies are getting wrong at the very start — the fascination phase
2:13 — Is agentic AI truly novel or just a rebrand of what we already had
3:03 — The right questions to ask — knowledge context tooling and instruction
4:51 — Pilot purgatory — AI projects stuck in QA that never make it to production
5:46 — First principles have not changed — what problem are we solving and what outcome do we want
8:50 — Are we implementing true agentic AI or just rebranding existing workflows
9:59 — Generative AI as a vending machine vs agentic AI as a sous chef
11:51 — Model in the loop — when you are really just calling a more sophisticated API
13:33 — Where agentic AI adds the most value — exception handling in enterprise workflows
16:40 — Governance as an enabler — why dirty data is a governance failure not an AI failure
18:19 — What feels familiar and what feels genuinely new about AI adoption today
20:29 — The importance of foundation before implementation — and the rush that bypasses it
23:29 — 40% of AI projects will fail by 2027 — and the bank that ran 350 use cases
25:18 — What actually determines AI success vs quiet failure
26:28 — Human in the loop vs human on the loop — why design matters from day one
29:41 — Agentic AI is a business project not a technology project
31:05 — Do we want true autonomy or accountability? The answer is both
🔔 Don’t forget to Like, Subscribe, and hit the notification 🔔 ✔
Subscribe: https://www.youtube.com/jonmyer/?sub_confirmation=1
📱 Social Media Twitter: https://twitter.com/_JonMyer
Website: https://jonmyer.com
LinkedIN: https://www.linkedin.com/in/jon-myer/
Spotify: https://open.spotify.com/show/0wjJzdIwctq4o4pTXM2KSF
Like my sounds? Here’s my audio source: https://www.epidemicsound.com/referral/fj9o9k
#aws #awscloud #podcast #podcasting #costoptimization #finops #cloudcost #cloudoptimization