Whatcha Got There?!🫣

Buckle up, tech fam! Every week, our dynamic duo “Spark” ✨ & “Trouble”😉 share some seriously cool learning resources we stumbled upon.

Product Labs🔬: Decoding M365 Declarative Agents

If you've ever wished you could create your own AI assistant without diving into complex code or infrastructure, Microsoft's Declarative Agents might just be your new best friend! Think of it as a LEGO set for building AI assistants – you just need to describe what you want, and voila, your assistant comes to life running on Microsoft 365 Copilot's infrastructure.

What's in it for you?

Declarative Agents take the "declare don't code" approach to a whole new level. Running on Microsoft 365 Copilot's infrastructure, these agents can be created using simple JSON files and natural language instructions.

The best part? You don't need any additional infrastructure beyond your existing M365 Copilot (commercial) license!

Declarative Agents is a great application of the "Network Effect" strategy.

• Initial Value: Each agent provides immediate utility to its direct users

• Compound Growth: Value multiplies as more agents are created and connected within the organization

• Ecosystem Expansion: Benefits increase with each new integration and knowledge source

With that let's break down the key features:

• No-Code Creation: Create agents using JSON files and natural language instructions. It's as simple as writing a recipe – describe what you want your agent to do!

• Knowledge Integration: Connect to SharePoint, OneDrive, Graph Connector knowledge, and even web content through Bing. Your agent becomes a knowledge powerhouse!

• Enterprise Ready: API plugins for external system integration, custom actions, and seamless integration with Microsoft 365 apps (Teams, Word, PowerPoint, Outlook).

Microsoft 365 makes building & using these agents super intuitive…

Your Agent, Your Way

You can create Declarative Agents via Teams by installing the Teams Toolkit on Visual Studio Code and, of course, via Copilot Studio.

Here is a quick video of Trouble, creating a Declarative Agent on Teams (to optimize & debug code snippets - how typical! 😂) faster than you can make Maggi!

Ready-to-Go Agents

Finding agents in Microsoft Teams or the Microsoft 365 app is simple and straightforward! Here's how:

• If you're in Copilot (within Teams or the Microsoft 365 app), just click Get Copilot agents on the right-side panel to explore the available agents in the app store.

• Alternatively, you can open the app store directly by clicking the Apps icon on the left sidebar of Teams or the Microsoft 365 app. From there, select Copilot agents on the left panel to browse through the list.

Once you've spotted an agent you want to use, hit Add either on the app tile or within the app's details page, and you're good to go!

There are also many agents that will be available for enterprise users to play around soon, wait to check it out on the app store.

Agents in Action

Now that you have added your agents, how do you use these agents in your business chat? Well, Microsoft offers 2 ways of achieving this…

What's the intrigue?

Here's where things get really interesting! According to Jeremy Thake (GPM, M365 Copilot) from the product team, Declarative Agents are just the beginning of what he calls a potential "AI mesh" within enterprises.

The future roadmap hints at some exciting possibilities:

• Agents talking to other agents (AI collaboration, anyone?)

• Direct document manipulation in Office apps

• Automated workflow orchestration

But perhaps the most intriguing aspect is how Microsoft is positioning this against competitors. While companies like Anthropic and OpenAI focus on building general-purpose AI assistants, Microsoft is taking a different approach – making it dead simple for enterprises to create their own specialized assistants.

This is also an example of the "Blue Ocean Strategy" framework:

Declarative Agents creates its blue ocean by addressing three key value innovations:

1. Democratizing AI Development: Creating a new market of non-technical AI assistant creators

2. Integrated Knowledge Management: Pioneering seamless enterprise knowledge connectivity

3. Low-Code Process Automation: Opening new possibilities for business process innovation

The future could see enterprises running dozens of specialized agents, each handling specific business domains while maintaining consistent enterprise context and compliance. It's like having a specialized AI workforce, each member an expert in their domain!

Who knows? Maybe someday we'll all be "programming" our AI assistants just by having a conversation with them!

You Asked 🙋‍♀️, We Answered ✔️

Question: AI systems like UroBot have surpassed human specialists in specific diagnostics. How can such breakthroughs ensure unbiased, equitable healthcare access, and what safeguards should accompany these developments?

Answer: AI in healthcare, like UroBot surpassing specialists in diagnostic accuracy, represents a paradigm shift. These systems excel in analyzing vast datasets quickly and consistently, reducing human error in areas like radiology and pathology. However, challenges arise in ensuring these models are unbiased and accessible.

Bias can stem from unrepresentative training data, potentially disadvantaging specific populations. Addressing this requires diverse datasets and transparent algorithms. Additionally, while AI offers cost-saving potential, infrastructure disparities could limit its reach in underfunded healthcare systems.

Safeguards include rigorous validation against global benchmarks, patient data protection, and establishing explainable AI protocols. Collaboration between technologists, medical professionals, and ethicists is crucial to ensure equity in AI-driven healthcare advancements.

Ultimately, AI should be an assistive partner to clinicians, not a replacement, amplifying human expertise while maintaining patient trust. Future adoption hinges on ethical, inclusive innovation.