Another Crazy Day in AI: The Two Problems Holding Back Gen AI

Hello, AI Enthusiasts.

Made it through the Monday blur? Same here. But the AI world’s already dissecting what’s holding GenAI back.

McKinsey’s team says too many AI programs crash because they aim high but forget how to scale. Their fix? Structure from day one.

Then there's Usama Fayyad, arguing the one-model-fits-all mindset is the wrong path—and that smaller, tailored AI works better in practice.

And speaking of bold moves, Apple is quietly stepping back into the AI spotlight with its new, ultra-efficient image generator that skips diffusion altogether.

Here's another crazy day in AI:

• Why your GenAI plans keep stalling

• AI expert says one-model approach is the wrong path

• Apple debuts image tech to compete with top image tools

• Some AI tools to try out

---

TODAY'S FEATURED ITEM: Lessons From GenAI Missteps

Image Credit: Wowza (created with Ideogram)

What happens when companies invest millions in genAI only to watch their projects collapse under the weight of compliance hurdles and scaling challenges?

In a new article published by McKinsey Digital, authors Curt Jacobsen, Erik Witte, Kaz Kazmier, and Oscar Villarreal—all McKinsey Partners—explore two major pitfalls that frequently derail generative AI programs: the failure to innovate and the failure to scale. Drawing on their work with over 150 companies, they unpack why promising projects often stall and share practical solutions for designing scalable, secure, and truly innovative genAI platforms.

The authors offer insights that reflect both the promise and the challenge of building enterprise-wide generative AI systems. Rather than zeroing in on model performance or toolkits, they draw attention to the broader organizational dynamics—how leadership, coordination, and infrastructure decisions play a critical role in whether genAI delivers real business value or fades into the background.

Some of the core ideas they bring forward include:

• Innovation often stalls because teams are limited to isolated use cases that don’t connect to broader goals

• Many platforms are built ad hoc, without a clear architecture that supports growth and reuse

• Security, governance, and compliance concerns tend to surface too late, disrupting momentum

• A coordinated platform vision, aligned across teams, leads to faster scaling and more consistent results

• The most effective organizations are those that view genAI not as a side project, but as a strategic discipline

These observations point to a deeper challenge: it’s not just about having the right tools or talent, but creating an environment where different teams can build on each other’s work without starting from scratch every time. When generative AI is rolled out in isolated pockets or through one-off pilots, the early promise rarely translates into lasting progress.

The takeaway here isn’t a set of quick fixes, but a prompt to reflect on how genAI is being positioned within the broader enterprise. Are teams solving for short-term wins, or are they investing in a foundation that allows experimentation to evolve into capability? What emerges from this piece is a sense that while the technology is moving quickly, the hard part often lies in the organizational work behind the scenes. And that’s where long-term value is either built—or lost.

Read the full article here.

---

OTHER INTERESTING AI HIGHLIGHTS:

AI Expert Says One-Model Approach Is the Wrong Path

/Cesareo Contreras, Reporter, on Northeastern Global News

Instead of rushing to implement massive, general-purpose AI models, businesses should focus on small, efficient models tailored to their specific needs. Usama Fayyad argues that the "one-model-fits-all" mindset—so popular in generative AI—misses the mark for most business applications. At the recent AI in Action Summit, he emphasized how smaller, private, and customizable models provide better performance, data control, and real-world value. Fayyad also called for greater AI literacy, data fluency, and thoughtful regulation to ensure ethical deployment.

Read more here.

---

Apple Debuts Image Tech to Compete with Top Image Tools

/Michael Nuñez, Editorial Director, on VentureBeat

Apple has introduced STARFlow, a breakthrough AI image generator that competes with leading models like DALL-E—without using diffusion. Developed in collaboration with university researchers, STARFlow uses normalizing flows and autoregressive transformers to generate high-res images efficiently. This marks Apple’s bold step toward regaining its AI edge, with innovations that support on-device intelligence and enterprise-level precision. Whether STARFlow evolves into a consumer-facing product remains to be seen, but it signals Apple’s commitment to unique, research-driven AI.

Read more here.

arXiv:2506.06276 [cs.CV]

---

SOME AI TOOLS TO TRY OUT:

• Betterfeedback – Turns surveys into natural chats that ask smarter follow-ups.

• Substrata – Reads real-time cues to help you sell smarter and close more deals.

• Eleven v3 – Converts text into expressive, multi-speaker speech in 70+ languages.

---

That’s a wrap on today’s Almost Daily craziness.

Catch us almost every day—almost! 😉

---

EXCITING NEWS:

The Another Crazy Day in AI newsletter is on LinkedIn!!!

Leveraging AI for Enhanced Content: As part of our commitment to exploring new technologies, we used AI to help curate and refine our newsletters. This enriches our content and keeps us at the forefront of digital innovation, ensuring you stay informed with the latest trends and developments.