AI-Powered Business Applications Bootcamp Recap - Day 3¶

AI-Driven Digital Concierge¶

Speaker¶

Andros Lee – Lead Solutions Architect, NCS Pte. Ltd.

A Microsoft Customers' AI Transformation Journey¶

Needs¶

“I need to reduce the cost of my operations while remaining competitive.”

Concerns: “Can we still be innovative while being efficient?”
Hopes: “I want to boost productivity for my employees.”
Microsoft’s NCS Values:
- Identify key personas.
- Establish areas where Copilot can provide benefits.

Considerations¶

“I think AI and automation can help my organization. How do I get started, and what should I consider?”

Concerns: “How do I prepare my company’s data, security, and systems to be AI-ready?”
Hopes: “I want to enable my company for an AI future and avoid pitfalls.”
Microsoft’s NCS Values:
- Review the current security posture.
- Develop an implementation and change management strategy.

Adoption¶

“I’m ready to turn on Copilot now. How do I maximize its utility?”

Concerns: “How do I optimize licenses while ensuring employees fully utilize Copilot’s potential?”
Hopes: “I want my organization to gain maximum benefits from AI.”
Microsoft’s NCS Values:
- Organize workshop sessions.
- Release FAQ guides showcasing diverse Copilot use cases.

Extension¶

“We’ve implemented Copilot successfully, but now I want to target individual needs.”

Concerns: “What role-specific Copilots can I deploy, and what systems are required?”
Hopes: “I want Copilot to empower all Lines of Business (LoBs).”
Microsoft’s NCS Values:
- Integrate role-specific Copilots into workflows.
- Engage subject matter experts for knowledge transfer.

Transformation¶

“My current systems are outdated, and my data is scattered.”

Concerns: “How can I modernize my enterprise systems and consolidate data?”
Hopes: “I want to future-proof my business.”
Microsoft’s NCS Values:
- Modernize legacy applications and databases.
- Embed automation and GenAI into business applications.

Breakdown of AI and Automation in the Microsoft Ecosystem¶

Microsoft 365¶

M365 Copilot: A GenAI and Orchestration Assistant that enhances productivity across Word, Excel, PowerPoint, Outlook, and Teams.

Copilot for Industry Solutions¶

Copilot for Sales: Generate meeting briefs, email summaries, CRM updates, and sales insights.
Copilot for Service: Optimize customer service with AI-driven responses, case resolution, and knowledge base integration.

Dynamics 365 Copilots¶

Sales
Customer Service
Finance
Supply Chain Management
Marketing
Field Service
Project Operations

Unified Channels of Engagement¶

Power Apps: Develop low-code business apps.
Power Pages: Create low/pro-code web platforms.
Copilot Studio: Build chatbots with GenAI and orchestration.
Power Automate: Handle RPA and business logic automation.
Power BI: Leverage analytics and reporting.
Azure AI Studio: Build custom AI models.
Azure OpenAI: Utilize pre-trained models (e.g., GPT 3.5, 4, etc.).

Copilot for GitHub¶

Bespoke Applications: Create custom solutions with GenAI and automation tools.

Highlighting the Capabilities of a Digital Concierge¶

I do my best to capture key notes, but these upcoming slides were packed;

Slides Presented

Highlighting the above "Breakdown of AI and Automation in the Microsoft Ecosystem"

aibootcamp2025-day3-aiecosystem

Highlighting the components from Breakdown of AI and Automation" for the use of the Digital Concierge

aibootcamp2025-day3-aidigitalconcierge

Highlighting the Application Architecture of a Digital Concierge

aibootcamp2025-day3-apparchitecture

Highlighting capabilities associated with a Digital Concierge function.

aibootcamp2025-day3-digitalconcierge

The Art of Crafting Effective Prompts for Your Power Platform AI Tools¶

Speaker¶

Elaiza Benitez – Senior Cloud Advocate, Microsoft

"AI Won’t Replace You, But People Using AI Will."

What Makes Generative AI Different?¶

Traditional Rule-Based AI¶

Guided by predefined rules.
Rigid and inflexible.
Unable to create new things.

Generative AI¶

Utilizes large, diverse datasets to guide outputs.
Flexible and trainable with provided data.
Capable of creating and producing new content.

The Shift to Generative AI¶

Current Apps → Intelligent Apps:

Constrained Interactions → Natural Language Interactions
Fixed Data Sets → Data-Driven, Personalized Experiences
Costly Changes → Rapid Feature Delivery

Copilot in Power Apps¶

Formula Drafting: Describe desired functionality, and Copilot creates the formula for you.
Error Identification: Use Copilot to locate and fix errors, improving app quality.
Formula Explanation: Gain insights into formulas you didn’t write, with Copilot’s explanations.

Enhancing App End User Experience with Copilot¶

Built-In Copilot Studio: Customize Copilot for app-specific needs.
Copilot Controls: Enable data-focused chat functionalities.
Rich Text Editing Integration: Enhance app usability with integrated capabilities.

Unlocking Possibilities¶

Create Automations with Prompts¶

"Go from idea to automation using natural language."

Build Functional Websites¶

"Effortlessly develop internal portals and customer-facing websites."

Prompt Building with AI Builder¶

"Design, test, and embed GenAI prompts across products using a single interface."

Prompt Formulas¶

Basic Formula¶

Task → Context → Expectations → Output

Example: "Create a product feedback summary based on recent customer reviews. Identify key themes, sentiment, and recurring praises or concerns. Format the summary to be concise and suitable for a marketing team meeting."

Expanded Formula¶

Persona → Context → Task → Exemplar → Format → Tone

Example: "You are a senior product marketing manager tasked with evaluating feedback on the Contoso Widget. Create a summary based on customer reviews, identifying key themes, overall sentiment, and recurring praises or issues. Use clear and concise language suitable for a marketing meeting. Write in a confident yet friendly tone."

Prompt Crafting Best Practices¶

Order Matters: Large Language Models (LLMs) process prompts sequentially, so the order impacts the response. Test variations for optimal outcomes.
Be Mindful of Tokens:
- AI breaks text into chunks (tokens), encompassing words or parts of words.
- Token limits apply to both input and output, so concise prompts are critical.
Beware of Hallucinations:
- AI may fabricate information when prompts are unclear or overly broad.
- Use specific, fact-based prompts and verify outputs.

Essential Guidelines¶

Know the Limits: Copilot is in early stages; understand its constraints.
Trial and Error: Expect prompt revisions for the best results.
Copilot ≠ Autopilot: Craft thoughtful prompts for effective outcomes.
Fact-Check Outputs: Verify AI-generated content for accuracy and relevance.

Power Automate Copilot Formula¶

When → Task → Format

Example: "At the end of each week, send me a summary of pending tasks in Planner. Format it as bullet points with task name, status, and due date in a Microsoft Teams message."

Copilot Studio Topic Formula¶

Task → Context → Expectations

Example: "Allow a user to check a flight’s status by entering the flight number and date. Include 2 message variations and a speech option for each question."

Key Takeaways¶

Understand the Models: Each model serves specific purposes, with varying costs.
Balance Outcomes: Balance Generative AI’s creativity and grounded data insights.
Educate Prompts: Link data sources to improve accuracy and relevance.
Privacy Assured: Copilot and agents do not share your data or use it for model training.
Access Control: Copilot only interacts with data you are authorized to access.

Prompt Library: aka.ms/power-prompts

Deliver Hyper-Automation Enterprise-Wide with Power Platform¶

Speakers¶

Taiki Yoshida – Principal Program Manager
Scott Berry – Senior Technical Specialist

The Transformative Power of AI¶

Universal Interface: Seamlessly interact with systems.
Reasoning & Planning: Anticipate needs and optimize workflows.
Memory & Context: Leverage historical and real-time data to adapt intelligently.

"Copilot is the UI for AI Across Platforms"

Copilot
Copilot Devices
Copilot & AI Stack

Copilot and Agents Source Data From:

External Data Connectors
Microsoft Graph (Microsoft 365)
Microsoft Dataverse (Business Data)
Microsoft Fabric (Analytics)

Hyperautomation: A Holistic, End-to-End Approach¶

Origin: Identify processes ripe for modernization.
Building: Offer no-cliffs tools, enabling users to choose the right automation technology for their needs.
Scale: Empower users and organizations to automate at higher volumes.
Optimize: Drive continuous iteration and improvement.

How Copilot Transforms:

What you build.
How you build it.

Microsoft's Leadership in Hyperautomation

Robotic Process Automation (RPA) – New
Process Mining – New
Intelligent Document Processing

What Can Process Mining Do?¶

Uncover Inefficiencies: Gain x-ray visibility into processes, identifying bottlenecks and comparing them to effective workflows.
Identify Optimization Opportunities: Receive actionable insights for automating and deploying solutions.
Continually Monitor Performance: Ensure processes are optimized and track ongoing improvements.

Latest Announcements¶

1. AI Recorder (Public Preview)¶

A new multi-modal Copilot to create desktop automations.
Self-healing capabilities for UI automation using AI.
Automate from description with “Describe with Copilot.”

2. Generative Actions (Announced at Ignite)¶

Add AI-powered steps directly to cloud flows for smarter, dynamic automations.

3. GPT-Infused Advanced Document Processing (Public Preview)¶

Intelligently extract and process unstructured data using natural language.

4. Hosted Machines (General Availability)¶

Power Automate offers hosted machines for automation without infrastructure management.
Scale up or down within automations seamlessly.

5. Automation Center (Public Preview)¶

A Copilot-enhanced single pane of glass for managing all automations within an environment.
Promotes shift-left governance where makers contribute alongside Centers of Excellence (CoEs) and admins.

Governance + Security of Copilot Agents and Power Platform¶

Speakers¶

Ryan Jones – Partner Director of Product, Power Apps Scale
Shawn Nandi – Group Product Manager, Power Platform
Lianne Zelsman – Senior Product Manager

Agenda¶

Continuum of agents and low-code development.
Security measures.
Governance strategies.
Operational excellence.
Customer Spotlight: Microsoft Digital.

Spectrum of Agents¶

Simple → Advanced Capabilities¶

Retrieval (GA): Retrieve information, reason, summarize.
Task (GA): Execute actions, automate workflows.
Autonomous (Preview): Independently plan, orchestrate other agents, and learn dynamically.

Simple → Advanced Tools¶

Agent Builder: Designed for end users.
Copilot Studio: Tailored for makers.
Copilot Studio + Azure AI: For developers building advanced solutions.

Common Questions on Managed Environments¶

Chief Information Security Officer (CISO)¶

How can I protect data from external threats?
How do I prevent data exfiltration?
How can I stay compliant with regulations?
How can I assess risks effectively?

Chief Information Officer (CIO)¶

How can I ensure solutions adhere to guidelines?
How do I gain visibility into solution usage?
How can experts review agents before wide deployment?
How do I drive cost efficiency and maximize ROI?

Operations Director¶

How can complex solutions be deployed at enterprise scale?
How do I get timely alerts on critical solution failures?

Announcements¶

Managed Security (Secure)¶

Power Platform Admin Center:

Security Hub offers advanced protection for AI-driven ecosystems.

Capabilities:

Security Posture Management: Guidance and assessments.
Threat Protection: Enforceable solution checker; Microsoft Sentinel integration.
Data Protection & Privacy:
- Customer-managed encryption keys.
- Data masking capabilities.
- Microsoft Purview Data Map integration.
- Advanced connector policies for environment groups.
- Network isolation using v-net.
Identity & Access Management:
- Conditional access for apps.
- Privileged access management.
- IP firewall and cookie binding.
Compliance:
- Lockbox service.
- Dataverse audit tools.

Managed Governance (Govern)¶

Power Platform Admin Center:

Agent Administration: A central hub for managing readiness, governance, and GenAI deployment at scale.
Environment Groups Management: Route makers to development environments and apply policies efficiently.
Licensing Management: Simplified licensing, capacity oversight, and compliance tracking.

Managed Operations (Operate)¶

Power Platform Admin Center:

Deployment Pipelines: Scale operations with approval-based production changes and source control integration.
Monitoring Tools: Real-time operational health metrics such as app open rates, time-to-interact, and full load times.
Observability: Integration with Azure App Insights for alerts, logs, and custom metrics.
Data Resilience: Extended backup retention, long-term recovery, and self-serve disaster recovery options.

Microsoft Digital¶

Who Are We?¶

Microsoft Digital (MS IT): Microsoft's first customer managing its vast tenant ecosystem.

Responsibilities:

Governance and security strategy.
Feature testing, enablement, and adoption plans.
User support initiatives.
Custom tooling to enhance product capabilities.

Current Environment¶

55K Environments across Developer, Sandbox, Dataverse for Teams, and Production tiers.
200K Apps and 400K Flows in active use.
25K Agents driving automated processes.
7K Pages developed for dynamic use.

Environment Groups¶

Personal Productivity - Developer Environments (Lowest Risk)

Self or Auto-Provisioned
Single owner, no sharing
Most 1^st party connectors allowed; certain actions blocked
Inactivity-based lifecycle

Team Collaboration - Dataverse for Teams Environments

Self-Provisioned
Access tied to M365 Group label
Most 1^st party connectors allowed; certain actions blocked
Inactivity-based lifecycle

Enterprise Development - Sandbox & Production Environments (Highest Risk)

Request process for provisioning
Additional management policies & compliance reviews
All 1^st party connectors allowed by default, custom policies by request
Attestation-based lifecycle