Power Platform & Agentic AI Architect

Rob Manning

Thirteen years building on the Microsoft stack, from Dynamics CRM to multi-agent solutions on Dataverse. This site is where I build in the open and compare notes with other architects.

Connect on LinkedIn Explore live demos
13+
Years experience
4
MS certifications
AB-100
Solution Architect
AB-100 PL-600 AB-400 PL-200
Agentic layersince 2026
Copilot StudioMicrosoft FoundryMCPA2AMulti-agent orchestration
Power Platformsince 2019
CanvasModel-drivenPower AutomatePower PagesPipelines & ALM
Dataverse & Dynamics 365since 2013
Schema designSecurityC# pluginsAzure integrations

Each layer built on the one beneath it.

Journey

Built on evolution,
always adapting

The journey started in November 2013 as a Systems Engineer at a Microsoft Partner, working across infrastructure, Exchange, Active Directory, and early Dynamics CRM. By 2015 the work had shifted fully into Dynamics 365 development. When Power Platform emerged in 2019, the Dataverse foundation was already established, and the move into architecture wasn't a role change; it was the next natural layer. The same pattern held when agentic AI arrived: by 2026 the work had expanded into agentic-first, multi-agent solutions across Dataverse, Copilot Studio, and Microsoft Foundry.

The growth from engineer to architect has come from evolving continuously alongside the platform, the clients, and the craft.

Additive Power is where that building happens outside client work: a full 3D print shop modeled in Dataverse, with every layer above wired in. See the live demos
  1. Nov 2013
    Systems Engineer
    Microsoft Partner — infrastructure, Exchange, Dynamics CRM
  2. Jun 2015
    Dynamics 365 Developer
    C# plugins, JS customizations, Azure integrations, Power BI
  3. Since 2019
    Power Platform Architect
    Enterprise architecture, CoE, ALM, full platform delivery
  4. Since 2026
    Agentic AI Business Solutions Architect
    AB-100 certified — agentic-first, multi-agent business solutions

Learning by Building

Live Power Platform Demos

Real deployments running on a personal developer tenant. Built to document patterns, not just describe them — interact directly.

🖨️
Demo Scenario: Additive Power — 3D Print Shop Operations
All demos share a single Dataverse solution built around a fictional 3D printing business — 5 custom tables, 2 security roles, synthetic data across 4 customer channels (B2B, B2C Premium, Retail, Walk-In). The solution uses a apdemo_ publisher prefix, ALM-ready packaging, and field-level security — the same patterns used in enterprise client environments.
⚡
Print Shop Kiosk App
Power Apps · Canvas · Dataverse · Role-Based
▶ Video
5-screen kiosk app for a 3D print shop. Check printer status, look up or create customers on the spot, open sessions, monitor active jobs, and close sessions with automatic duration tracking, covering printer sessions, job records, walk-in contact registration, and filament consumption.
→ 5-screen workflow built around a realistic print shop environment
→ Walk-in contact creation — new customers registered directly from the kiosk without leaving the app
→ Role-based access with FrontDesk and Operator security roles
→ Live session tracking with automatic duration calculation
→ Dataverse-native: printer status, customer lookup, filament consumption
→ Runs on iPad, Android tablet, and desktop without code changes
▶ Watch Demo →
5 screens · session tracking · job records · field-level security roles
📊
3D Print Operations Dashboard
Power BI · Microsoft Fabric · Direct Lake · Dataflow Gen2 · DAX
Live
Four-page operational report covering Operations Overview, Walk-In Sessions, Filament Tracker, and Printer Health, built on Microsoft Fabric's Direct Lake architecture. Data flows from Dataverse through Dataflow Gen2 into a Fabric Lakehouse, where a Direct Lake semantic model reads Delta Parquet files directly from OneLake without Import refresh cycles. Custom DAX measures cover reprint rate, session utilization, filament consumption, and printer availability, with option set integer values resolved via dimensional lookup tables in the semantic model.
→ Four-page report: Operations Overview, Walk-In Sessions, Filament Tracker, Printer Health
→ Direct Lake architecture — Dataflow Gen2 into Fabric Lakehouse, no import refresh cycles
→ Custom DAX measures for reprint rate, session utilization, and filament consumption
→ Option set values resolved via dimensional lookup tables in the semantic model
→ Report published and embedded via Power BI Embedded A-SKU
📊 Open Report →
4 pages · Direct Lake · Dataflow Gen2 · 9 DAX measures · 3 lookup tables · Fabric Lakehouse
🔄
Session Close Handler
Power Automate · HTTP Trigger · Dataverse · Secure Proxy
Live

Runs a live Power Automate flow via secure proxy. The flow creates a synthetic printer session backdated 45 minutes, then immediately closes it — stamping the end time, calculating duration, resetting printer status to Available, conditionally creating a linked Print Job record, and updating contact metrics. Full 8-step chain exercised on real Dataverse data every run.

8-step flow · EndTime stamp · Duration calc · Printer status reset · Job creation
🤖
Additive Power Assistant
Copilot Studio · Knowledge Base · Generative AI · Claude Sonnet 4.6
Live
Ask the Additive Power assistant questions about the print shop. It answers using a grounded knowledge base covering printer status, active sessions, filament inventory, and job history — with tabular responses and source citations.
→ Natural language queries — printer status, sessions, filament, job history
→ Tabular responses with source citations for operational queries
→ Suggested prompts guide visitors to key demo scenarios
→ No authentication required — publicly accessible chat embed
🤖 Open Demo →
Knowledge base · Generative AI · Tabular responses · Source citations · No auth embed
🗄️
Production Console
Model-Driven App · Dataverse · Views · Business Rules · Dashboard
▶ Video
Back-office production console built directly on Dataverse. No custom UI required. Manages the full printer fleet, session history, job records, and filament inventory through configurable views, forms, and a real-time operations dashboard. Security roles scope access by function: operators see sessions and jobs, admins see everything including filament reorder thresholds and printer maintenance flags.
→ Dataverse-native back-office console — no custom UI required
→ Full fleet management across printers, sessions, jobs, and filament inventory
→ Configurable views, forms, and real-time operations dashboard
→ Role-scoped navigation: Operator and FrontDesk access enforced at the security role level
▶ Watch Demo →
4 views · operations dashboard · field-scoped roles · business rules · Dataverse-native
🧠
Agentic AI Layer
MCP · Microsoft Foundry · Azure Functions · Dataverse
▶ Video · AI
A custom MCP (Model Context Protocol) server deployed as an Azure Function exposes live Dataverse data as callable AI tools: printer availability, active session queue, filament inventory levels, and job history. A Microsoft Foundry agent connects to the MCP server, enabling natural language queries against real operational data with no UI interaction required. The same Dataverse tables driving the Canvas App and Power BI report become the grounding data for an AI agent that can reason, summarize, and surface actionable insights on demand, bridging the Power Platform stack with the Microsoft agentic AI ecosystem.
→ Custom MCP server deployed as an Azure Function exposing four live Dataverse tools
→ Microsoft Foundry agent connected via OpenAPI — natural language operational queries with no UI required
→ The same Dataverse tables powering the Canvas App and Power BI report become AI grounding data
→ Certificate-authenticated Entra ID integration bridging the Power Platform stack with the Microsoft agentic AI ecosystem
▶ Watch Demo →
4 MCP tools · Azure Functions · Microsoft Foundry Agent · Dataverse grounding · Natural language ops queries

Data Model

Solution Architecture

Five Dataverse tables designed the same way I build production solutions — publisher prefix, security roles, ALM packaging, and financial fields excluded at schema level.

Publisher: apdemo Solution: APDemoPortfolio 5 Tables 2 Security Roles
👤
Contact
Standard · extended
🖨️
Printer
apdemo_ custom
⏱
PrinterSession
apdemo_ junction
📋
PrintJob
apdemo_ custom
🧵
Filament
apdemo_ standalone
View Full Schema & Relationships →

Microsoft Certified

Certifications

Agentic AI architect certified, plus all three Power Platform certification tiers — functional, developer, and architect.

AB-100
Microsoft Agentic AI Business Solutions Architect
Microsoft · Active
✓ Expert Level
PL-600
Microsoft Power Platform Solution Architect
Microsoft · Active
✓ Expert Level
AB-400
Microsoft Power Platform Developer
Microsoft · Active
✓ Associate Level
PL-200
Microsoft Power Platform Functional Consultant
Microsoft · Active
✓ Associate Level

Technology

Tech Stack

Tools and platforms I work with daily across the Microsoft ecosystem.

Power Platform
Power Apps (Canvas)
Power Apps (Model-Driven)
Power Automate
Power Pages
Power BI
Dynamics 365
Sales
Customer Service
Field Service
Customer Engagement
Dataverse
Azure & DevOps
Azure DevOps CI/CD
Azure Functions
Logic Apps
Service Bus
Entra ID / AAD
Static Web Apps
Development
C# / Server Plugins
PCF Controls
JavaScript / TypeScript
REST / OData APIs
Liquid Templates
Architecture
Solution Design
Power Platform CoE
ALM & Governance
DLP Policy
ISV Packaging
AI & Agents
Microsoft Foundry
Copilot Studio
Model Context Protocol (MCP)
Azure OpenAI
Prompt Engineering
AI Agents

Community

Let's Connect

Power Platform architect based in Atlanta. Always interested in connecting with others building on the Microsoft stack — architecture questions, pattern discussions, or just comparing notes.

Connect on LinkedIn