Creating Agentic AI Magic

Exploring how CrewAI makes agent creation magical, who's leading the LLM rat race, and the dangers of "agent washing."

Author

AI Confidential Newsletter
May 20, 2025

Welcome to AI Confidential, your biweekly breakdown of the most interesting developments in confidential AI.

Today we’re exploring:

  • How CrewAI makes agentic AI feel magical

  • OpenAI’s leading spot in the enterprise AI adoption race

  • The latest in open source projects to check out

Also mentioned in this issue: João Moura, Gary Cohn, Mark Russinovich, Sachin Katti, Jason Clinton, Nelly Porter, Daniel Rohrer, James Kaplan, John Willis, Atin Sanyal, Jake Broekhuizen, Chester Chen, CrewAI, GitHub, Digital.ai, Checkpoint, Legit Security, IBM, OpenAI, Hugging Face, Anaconda, and Tether.

Let’s dive in!

This week on the AI Confidential Podcast, we’re diving into the fast-moving world of AI agents with João Moura, founder and CEO of CrewAI. 

CrewAI started as an agentic AI open source framework João worked on during a wedding anniversary trip in 2023. He launched the product just a few months later, which was deployed by employees at Oracle soon after. 

Since then, João’s continued success comes from one central vision:

Make agent creation feel borderline magical.

Inspired by the Ruby on Rails philosophy of “convention over configuration,” CrewAI is designed for speed, clarity, and ease of use.

But what truly sets CrewAI apart is its community-first approach to open source. 

From launching a job board to prioritizing self-hosted deployments, the team takes a developer-centric approach to agentic AI, tackling growth and innovation in three powerful directions:

  1. Builders — Agents that create tools, apps, and content

  2. Robotic Process Automation (RPA) — Agents that streamline/automate complex processes

  3. Synthetics — Agents that act like humans (we’re not fully there yet)

What we love about João’s approach to agentic AI is that he’s not just disrupting the market. He’s shifting the way enterprise leaders think about long-term workflows and processes, too.

As João says in the episode, the goal isn’t just to build agents.

It’s to use them to drive real business value.

In this episode, you’ll also hear João get into the nitty gritty about:

  • Why interoperability across your stack is essential for scaling AI

  • How managers are uniquely equipped to build effective agents

  • Why advanced AI enhances teams—not replaces them

  • The dangers of “agent washing” and how to see through the hype

Our conversation with João was an absolute blast. The podcast is out now. Listen here.

P.S. João is a featured speaker at the pre-summit workshop for the Confidential Computing Summit next month. Get your ticket now to hear him speak live!

Keeping It Confidential

In 2023, what percent of GitHub developers were using AI coding tools to improve their open source projects?

  1. 55%

  2. 78%

  3. 92%

  4. 100%

See answer at the bottom.

Code for Thought

Important AI news in <2 minutes

📈 Increased AI adoption is linked to an 18% jump in cyber attacks on mobile and web apps since 2024, new Digital.ai research finds.

🔐1 in every 80 prompts sent to LLMs exposes sensitive data to possible attackers, Checkpoint’s 2025 AI Security Report states.

🤖 61% of CEOs are going all-in on AI agent deployment across their businesses, an IBM study found. 

🏆 OpenAI’s large language models are winning the enterprise AI adoption race, beating the likes of Anthropic and Google. 

🏛️ U.S. senators introduced a law that would ban the federal government from using R1’s DeepSeek to prevent China from obtaining sensitive data.

Community Roundup

Updates involving OPAQUE and our partners

⏳ We’re just weeks away from OPAQUE’s 2025 Confidential Computing Summit, and the hype is real!

Happening June 16–18 in San Francisco, this three-day event brings together the pioneers shaping the future of secure, enterprise AI.

🎤 Attend keynote talks from industry leaders like:

Kick things off early on June 16 at our pre-Summit workshop, The New NORMAL: Normalizing AI Enterprise Architecture, featuring:

Get your ticket now before it’s too late. 

Use code AICNEWS25 for an exclusive discount. We can’t wait to see you there!

OPAQUE in the wild

✈️ Last week, OPAQUE’s own Giorgio Natili flew across the pond to speak at the Helix Summit in Barcelona, Spain. 

Speaking to leaders in continuing medical education and AI, Giorgio’s talk covered:

  • The benefits of confidential AI and retrieval-augmented generation (RAG) in healthcare training

  • Why AI is reshaping healthcare learning environments

  • How to harness AI to elevate learning and improve patient outcomes

Open source spotlight 

🌐 Hugging Face’s Open Computer Agent can interact with websites to perform tasks, like filling out forms and booking tickets, all on its own. 

🐍 Anaconda’s AI platform is designed for Python developers using open-source software to build AI tools in a secure environment. 

🪙 Tether’s AI tool lets users build AI applications with payment integrations for cryptocurrencies like USDT and Bitcoin, eliminating the need for a middleman.

Quotable

💡 “As AI adoption accelerates creating greater efficiency, and productivity gains, the ultimate pay-off will only come to CEOs with the courage to embrace risk as opportunity…. At this point, leaders who aren't leveraging AI and their own data to move forward are making a conscious business decision not to compete.”

— Gary Cohn, Vice Chairman, IBM in the foreword of IBM’s CEO study

Trivia answer: 92%

The use of AI coding tools on open source platforms has skyrocketed. Proving this, GitHub reported that 92% of developers were using AI tools for their projects in 2023, something they predict will drive a new wave of innovation on the platform.

Stay confidential!

- Your friends at OPAQUE

ICYMI: Links to past issues

How'd we do this week?

Vote below and let us know!

Login or Subscribe to participate in polls.

Reply

or to participate.