• Everyday AI
  • Posts
  • Copilot Studio: Autonomous AI agents for everyday people

Copilot Studio: Autonomous AI agents for everyday people

Inside scoop on Copilot Studio, new LLM beats o1 model, OpenAI deletes data in NYT lawsuit, Trump's AI policy reversal, and more!

Author

Everyday AI
November 21, 2024

Sponsored by

“We’re not being as creative as we can be when we think about AI,” says Microsoft Chief Scientist and Technical Fellow Jaime Teevan.

She joins the WorkLab podcast to discuss how leaders can unlock the full potential of AI for better brainstorming, ideation, and collaboration.

Learn more on WorkLab, available wherever you get your podcasts.

Outsmart The Future

Today in Everyday AI
7 minute read

🎙 Daily Podcast Episode: Autonomous agents are THE new thing in the AI space. So when Microsoft announced its autonomous AI agents inside Copilot, we knew we had to sit down with their VP of AI Agents to get the inside scoop. Give it a listen.

🕵️‍♂️ Fresh Finds: Huawei to launch new AI chip in 2025, Messenger video calls get AI backgrounds and NVIDIA to build AI school in Indonesia. Read on for Fresh Finds.

🗞 Byte Sized Daily AI News: New LLM beats OpenAI’s o1, OpenAI deletes data in NYT lawsuit, Trump's AI policy reversal sparks concern, NVIDIA speaks on new AI model concept. For that and more, read on for Byte Sized News.

🚀 AI In 5: We’re reviewing Meta’s new LLM Llama 3.2 and showing you how you can download it to your device. See it here

🧠 Learn & Leveraging AI: We’re breaking down everything you need to know about Copilot Studio and Microsoft’s new autonomous AI agents. Keep reading for that!

↩️ Don’t miss out: Did you miss our last newsletter? We talked about ChatGPT’s o1 model reclaiming top spot, ChatGPT voice model available on web, Atman’s Rain AI chip gains investments and Meta unveils new AI product team. Check it here!

Copilot Studio: Autonomous AI agents for everyday people 🕵

Ready for the buzzword(s) of 2025?

Autonomous Agents.

Ready for your cheatsheet for Autonomous Agents?

Brought to you by legit one of the worldwide leaders in the space -- Ray Smith.

Ray is the VP of AI Agents at Microsoft.

Ray took time out of his busy schedule to join us at the Microsoft Ignite conference.

This is one convo you legit cannot miss.

Join the conversation and ask Jordan and Ray questions on Microsoft AI here.

Also on the pod today:

• SharePoint Agents 📍
• Use Cases for Copilot Studio 🧑‍💻️
• Data Security and Governance 🔓

It’ll be worth your 32 minutes:

Listen on our site:

Click to listen

Subscribe and listen on your favorite podcast platform

Listen on:

Here’s our favorite AI finds from across the web:

New AI Tool Spotlight – Taurin is an AI-native email client for founders, Lovable is a full-stack AI engineer and notclass helps you search relevant content on YouTube.

AI Chips – Huawei aims to mass-produce its newest AI chip in early 2025.

Meta – Messenger video calls are getting better quality and AI backgrounds.

AI Models – Black Forest Labs has released FLUX.1 Tools, a suite of AI models for controllability.

NVIDIA – NVIDIA is building an AI school in Indonesia.

AI in Society - A child safety org is launching an AI model trained on detecting child sexual abuse images.

Trending in AI – AI pioneer Yoshua Bengio is calling for more research and “guardrails” to develop AI safely.

1. DeepSeek Launches High-Performance LLM, Beating OpenAI’s o1 🤯

DeepSeek has just unveiled the R1-Lite-Preview, an innovative large language model that emphasizes transparent reasoning, exclusively available through its chat platform. This new model claims to rival and even surpass OpenAI’s o1-preview, showcasing impressive performance in logical inference and problem-solving across various benchmarks.

With a commitment to open-source accessibility, DeepSeek plans to release its models and APIs in the near future, promising a significant impact on both the AI community and industries looking to leverage advanced reasoning capabilities.

2. Report: OpenAI Deletes Data in NYT Lawsuit Case ⚖️

In a recent twist in the ongoing legal battle between OpenAI and major publishers like The New York Times and Daily News, lawyers claim that OpenAI engineers accidentally deleted crucial data related to their copyrighted content. This misstep came after the publishers spent over 150 hours searching OpenAI’s training data for potential infringements, only to find that the recovered data was unusable for tracing specific article usage.

While the plaintiffs do not suspect intentional wrongdoing, this incident highlights the challenges of ensuring fair use in AI model training.

3. Nvidia's $19 Billion Quarter Sparks AI Debate 🤔

Nvidia’s recent earnings report revealed a staggering $19 billion in net income, yet concerns linger among investors about the sustainability of its growth amidst emerging AI methodologies. CEO Jensen Huang emphasized the potential of “test-time scaling,” a concept gaining traction in AI model development, suggesting it could reshape Nvidia's future.

While the chipmaker currently dominates AI model training, the rise of nimble startups in AI inference presents a competitive challenge..

4. Trump's AI Policy Reversal Stirs Global Concern 😨

President-elect Donald Trump has pledged to repeal President Biden’s AI policies, stirring concerns among global leaders gathered in San Francisco for an AI safety conference.

This meeting, which includes representatives from several allied nations, aims to tackle pressing issues like deepfakes and the broader implications of AI on society. U.S. Commerce Secretary Gina Raimondo emphasized that safety in AI development is essential for fostering trust and innovation, urging collaboration across nations.

5. Google's AI Tool Uncovers Critical Vulnerabilities in OpenSSL 🔐

Google revealed that its AI-powered fuzzing tool, OSS-Fuzz, has identified 26 vulnerabilities in open-source code, including a medium-severity flaw in OpenSSL (CVE-2024-9143). This particular issue, which could lead to application crashes or remote code execution, highlights the effectiveness of AI in automating security checks that humans might miss.

With the integration of large language models into OSS-Fuzz, Google has significantly enhanced code coverage across numerous projects, adding over 370,000 lines of new code.

6. U.S. Reigns Supreme in AI Innovation 🇺🇸

A new report from Stanford University reveals that the United States continues to lead the global charge in artificial intelligence development, significantly outpacing China in both investment and research. With private AI funding in the U.S. soaring to $67.2 billion last year, compared to China's $7.8 billion, the gap is reportedly widening, according to computer scientist Ray Perrault.

The rankings highlight the importance of responsible AI practices, showcasing the U.S. as a hub for major players like Google and OpenAI, while also underlining the rapid advancements of countries like the UK and India.

Llama 3.2 Quick Review - Meta releases new multimodal and on-device models

Meta recently released Llama 3.2, its multimodal model.

We’re going over what’s new with Llama 3.2 vs 3.1 and how you can download Llama 3.2 to your device and test it out.

🦾How You Can Leverage:

Ready for the buzzword(s) of 2025? 👇 

↳ Autonomous AI Agents. 

Ready for your cheatsheet for Autonomous Agents? 

↳ We gave it to you today, courtesy of one of the worldwide leaders in the space -- Ray Smith. 

Ray is the VP of AI Agents at Microsoft. 

(You know.... the current company leading the world's development in .... Autonomous Agents.)  

Ray took time out of his busy schedule to join us at the Microsoft Ignite conference.

This is one convo you legit cannot miss. 

Don't say we didn't warn ya shorties. 

Think about it…. those repetitive tasks you hate? They're about to vanish. If you follow this guide. 

Need to process 10,000 invoices by morning?

Tired of copy-pasting data between systems?

Ready to automate your entire department?

Miss this, and you'll spend 2025 watching competitors build AI workforces while you're still copying and pasting from ChatGPT.

Let’s get autonomous y’all. 🦾

1 – The Three-Tier Agent Architecture That Changes Everything 📚

Microsoft just democratized agent creation across three distinct layers, each targeting different technical abilities.

Ray broke down the three tier approach to agentic implementation. 

The first hits Microsoft 365 users, enabling instant agent creation from any SharePoint site. Zero code required.

The second unleashes Copilot Studio, where business units can build autonomous agents that trigger on emails, monitor SAP, and execute complex workflows.

The third opens Azure AI Foundry to developers, offering raw API access for custom agent development.

Try This: 

Start with your most-accessed SharePoint folder. Create an agent today – it takes two clicks. (Alright, maybe 3.5)

Point it at your company policies. Task it with cross-referencing three months of team communications against those policies. Measure response accuracy against your current manual process.

 Target: 85% accuracy in the first week.

2 – Teams Just Became Your Universal AI Command Center ☎️

Picture this: Your German team member speaks. Your Japanese colleague hears it in real-time in their native language.

That's live now.

(Yes, actual mind blown emjoi.)

The new Teams facilitator agent doesn't just track action items. It balances speaking time, flags off-topic discussions, and ensures meeting objectives stay on target.

This isn't about replacing human interaction. It's about injecting AI into every collaborative moment to multiply its impact.

Try This: 

Deploy the Teams translator in tomorrow's international meeting. 

Measure participation rates from non-native English speakers before and after. Track comprehension through follow-up task completion rates. Set a baseline by recording current international meeting participation metrics tonight.

Who knew… using AI is actually a way to create more empathy between humans. 

3 – The SharePoint Security Model That Makes Legal Teams Sleep Better 🔐

Every new SharePoint agent inherits your existing security framework.

No new permissions needed. No additional security reviews required.

It just…. Works. 

(Side note, we talked to the SharePoint team at Ignite and there’s some REALLY good things coming.)

These Sharepoint agents read only what humans can access. They respect every label, every policy, every restriction already in place.

When they pass information between departments? They maintain those permissions dynamically. In real-time. At enterprise scale.

Try This: 

Select your most sensitive SharePoint site with cross-department access restrictions. Create an agent. Test it with users from different permission levels. Have them request information outside their clearance. Document every response. Verify against your current permission matrix. Complete this security audit by end of week.

Your existing Microsoft security stack just became your AI governance framework. Without adding a single new policy.

⌚

Numbers to watch

$686 Million

Crusoe, a rumored OpenAI data center supplier, has secured $686M in new funds.

Now This …

Let us know your thoughts!

Vote to see live results

What type of technical coding and development episodes, if any, might you wanna see?

Click to see live results

Login or Subscribe to participate in polls.

Reply

or to participate.