The AI Riff

AI isn’t just going rogue anymore—people are getting really good at using it to go rogue themselves. In this episode, the crew digs into how AI is supercharging fraud, from fake documents to cloned voices and eerily convincing scams. What used to take effort now takes seconds—and the line between real and fake is getting dangerously thin.

From code words and call-backs to bigger questions about trust, privacy, and how far businesses should go to protect themselves, this conversation gets real fast. The takeaway? It’s not just about smarter tech—it’s about staying one step ahead of the humans using it.

Listen in before the next “trusted” voice you hear… isn’t.

What happens when AI doesn’t just get things wrong… but starts ignoring you altogether?

In this episode, the crew dives into a new (and unsettling) reality: AI that bends rules, bypasses instructions, and “thinks” it knows better. From rogue coding assistants to drifting models and conflicting AI opinions, this conversation explores what trust really means when the tools we rely on stop listening.

Is it just the same old hallucination problem—or something more? And how do you stay in control when the machine starts negotiating back?

Buckle up—this one gets philosophical fast.

The conversation delves into the impact of AI and its challenges, covering topics such as March Madness, Amazon's AI mishap, and the role of IT in AI management.Takeaways* AI and its Impact* Challenges of AI ImplementationChapters* 00:00 March Madness and AI* 20:45 The Role of IT in AI Management

The conversation covers the celebration of the 40th episode, the introduction of Steve Ahearn, Microsoft and Google's AI initiatives, practical implementation of AI tools, AI's impact on the mortgage industry, AI's role in contract writing, AI's application in home improvement, AI's influence on legal contracts, the future of the mortgage industry, and regulation and AI in the mortgage industry.

In Episode 29 of The AI Riff, the crew celebrates their near-30 milestone with a bold retrospective and a dive into one of the most audacious ideas in AI infrastructure yet: data centers in space.

What starts as a riff on “galactic brains” turns into a grounded but provocative exploration of energy beaming, compute offloading, and what it all means for the future economics of AI. Are these moonshot visions actually coming down to Earth—or are we just building targets in orbit?

Along the way, they revisit early predictions, throw some shade (and love) at Klarna’s pivot from AI-led CS to crypto wallets, and give a shoutout to Cursor’s game-changing debug mode.

Key Moments:

2:45 – Galactic brain: Why data centers in orbit might not be so sci-fi

7:01 – The energy-beaming angle: Limitless power or endless hype?

18:14 – Klarna’s arc: From AI poster child to crypto side quest

22:48 – Cursor’s new debug mode: “Brilliant” isn't an overstatementThis episode is all steak, some sizzle—and a few snarky grenades.

This episode breaks down Menlo Ventures’ “State of AI” report — and what the data really says about where enterprise AI is headed (and what everyone’s still getting wrong).

Key moments:
2:13 — The headline: enterprise AI is the fastest-scaling software category (and the “is this GenAI spend?” debate).
4:26 — The shift: money is moving from foundation models to the application layer (shovels vs. gold).
7:13 — Build vs. buy flips hard: enterprises are increasingly buying AI solutions from vendors instead of building in-house.
11:53 — The messy reality: employees are expensing AI on personal cards… while compliance teams try to keep the company out of the newspaper.

Big takeaway: we’re not in the “agent era” yet — most AI spend is still human-prompted — but the vendor ecosystem is accelerating fast, and governance is already playing catch-up.

Episode 28 is the “AI sitting next to your data” episode and why that’s both the dream and the danger.

Anthropic + Snowflake is the headline, but the real convo is: what happens when enterprise AI can finally query clean corporate data… and start suggesting decisions?

Key moments:

1:50 — Anthropic x Snowflake deal: Claude “next to the data” and why that model matters.

5:39 — The big unlock: clean data + AI = real analytics (and why “garbage in, garbage out” still rules).

9:27 — Mortgage use cases get real: condition pattern discovery + “negative” work you can eliminate.

10:43 — The warning label: scenario suggestions can become *steering* without explainability + fairness + monitoring.

Big takeaway: AI isn’t magic. But when it’s embedded where the data lives, it stops being a demo… and starts becoming operations.

If AI could safely sit on top of your company’s data tomorrow, what’s the first question you’d ask it?#EnterpriseAI #DataStrategy #Snowflake #Anthropic #AIinFinance

The crew reacts to a New York Times take on students flocking to AI majors—and asks the real question: does an “AI degree” translate to job-ready skills, or should AI be an overlay on domain expertise (law, accounting, engineering, mortgage, etc.)?

They dig into what companies actually need (applied reasoning, workflow orchestration, bias/GRC awareness), the risk of AI-native grads overestimating their readiness, and why “I’m just the AI guy” isn’t a role—yet.Key moments* 0:56 – NYT spark: college kids rushing into AI majors* 2:52 – What happens when AI-native grads enter compliance-heavy companies?* 6:54 – AI as a standalone major vs. AI as an overlay across every discipline* 11:17 – Why domain expertise + AI fluency beats “AI fluency alone”

This episode cuts through the noise on one of the most persistent misconceptions in the AI discourse: that implementation is easy.

Your hosts Kyle Meyers, Mario DiBenedetto, and Ruth Lee break down the harsh realities of AI adoption in real-world enterprise environments—especially lending. With humor, candor, and real stats, they explore what’s working, what’s not, and why many companies are stuck in the AI “evaluation” phase.

🧠 Top 3 Moments:

[2:00–3:43] AI Integration is Tougher Than Advertised

[9:01–12:10] Agentic AI? Not So Fast

[13:00–15:20] Why Mortgage AI Is an Uphill Battle

What happens when Microsoft builds a superintelligence team—but doesn’t chase AGI? Instead, they point their smartest minds at one thing: helping doctors diagnose disease. In this episode, we riff on the strategy behind that decision—and why narrow, high-stakes AI might be the smartest kind.

Kyle, Mario, and Rob unpack Microsoft’s approach to human-in-the-loop AI, the business model implications of medical breakthroughs, and what it means when AI becomes part of the healthcare supply chain. But things get personal, too: What happens when profit motives clash with cures? Can AI be a tool for humanity if it’s built by corporations?

We also explore:

* The slow pivot away from generalist AI toward specialist agents

* Open-source cellular modeling and the open-access vs. monetized AI debate

* The latest dev tools, including Cursor’s multi-step agent pipeline

And yeah, we check in on Copilot, take a quick detour into Ask Jeeves nostalgia, and pitch a new hit show: Diagnosis: AI.

A big episode on the future of medicine, money, and the role of humans in AI’s last mile.