Solutions · agents · useful things

I build AI people use.

I design practical AI systems, agent teams, and business applications that move from idea to working proof fast - without giving up the human judgment that matters.

  • 29 public repositories
  • Minneapolis, Minnesota
  • Human control by design

Independent projects. Clear boundaries. Practical possibilities.

Selected public work

The collection.

Nine customer-neutral projects, reconciled from the live portfolio. Open a public prototype or go directly to its public source.

Live prototype

Freight Triage Control Tower

Turns flagged shipment email into owned, SLA-backed triage work with queues, item detail, aging, and resolution views.

  • Power Apps
  • Dataverse
  • Copilot Studio
Live game

Prompt Escape

A multiplayer prompt-engineering escape room that turns clearer instructions, constraints, and creativity into a playable team challenge.

  • TypeScript
  • GitHub Copilot
  • AI learning
MIT accelerator

SellerSphere

A voice-first seller reference for Dynamics 365, Copilot Studio, Microsoft 365 context, and Azure AI Speech with clear implementation guardrails.

  • Dynamics 365
  • Copilot Studio
  • Azure AI Speech
Starter kit

Agent Team Starter Kit

A configurable agent team using GitHub Issues as the work queue, specialized personas, outcome gates, Azure Functions, and OpenAI.

  • Multi-agent
  • GitHub Actions
  • Azure
Community solution

Customer 360 for Entities

An adaptive D365 and Dataverse web resource that unifies cases, orders, entitlements, contacts, and operational context on core forms.

  • Dynamics 365
  • Dataverse
  • JavaScript
Personal lab

Home Assistant Estate Panel

A local-first React control surface and migration playbook for replacing a subscription smart-home stack with Home Assistant.

  • React
  • TypeScript
  • Local first
Community accelerator

Warranty + Claims Operations

A Power Platform starter with Copilot Studio agents, Canvas Apps, flows, and Dataverse tables for warranty and claims workflows.

  • Power Platform
  • Copilot Studio
  • Dataverse
Hackathon prototype

Deal to Glory

A voice-first multi-agent flow that turns a spoken seller update into structured CRM, attainment, and recognition actions using synthetic data.

  • Multi-agent
  • Voice
  • TypeScript
Hands-on lab

Copilot Studio + MCP Lab

A hands-on MCP server lab showing how to deploy a tool and connect it to both GitHub Copilot and Microsoft Copilot Studio.

  • MCP
  • Copilot Studio
  • Azure
Builder profile

From rough idea to working system.

I work across the whole path: framing the outcome, shaping the architecture, building the interface, wiring the platform, and making the handoff usable.

01 / Agentic systems

Agents that do work, not just chat.

Orchestration, tools, grounding, MCP, evaluation, feedback loops, and explicit human gates.

02 / Business apps

AI inside the flow of work.

Dynamics 365, Dataverse, Power Apps, Power Automate, and Copilot Studio experiences grounded in real operational jobs.

03 / Product craft

Make the idea tangible.

Responsive prototypes, voice-first interfaces, control towers, web resources, and demo systems people can test instead of imagine.

04 / Operating design

Build the control plane around the model.

Clear ownership, evidence, observability, security boundaries, portable skills, and reviewable actions turn one-off AI moments into reliable ways of working.

Digital resume

Technologist. Builder. Systems thinker.

I am Bill Whalen, a Minneapolis-based technologist at Microsoft. I work at the seam between people, AI, and business systems: finding a stubborn workflow, framing the real outcome, and making the idea tangible enough to test.

The artifact might be a multi-agent team, a Dynamics 365 accelerator, a voice-first interface, an operating system for field work, or a game. The common thread is practical: working software, clear boundaries, and an adoption path.

How I work

Working principles

  1. 01

    Start with the outcome.

    Define what changes for the person doing the work before choosing the model or platform.

  2. 02

    Build the proof.

    A working prototype creates better decisions than another round of abstract architecture.

  3. 03

    Keep humans at the consequential gates.

    Automation should increase leverage without hiding uncertainty, ownership, or irreversible action.

  4. 04

    Package what survives.

    Turn validated patterns into accelerators, skills, and handoffs other builders can use responsibly.

Technologies I reach for

  • Microsoft Copilot
  • Copilot Studio
  • Azure AI
  • Dynamics 365
  • Power Platform
  • Dataverse
  • React
  • TypeScript
  • JavaScript
  • PowerShell
  • GitHub Actions
  • MCP
  • Home Assistant
Release architecture

One public story. More ways to read it.

The approved direction keeps future project pages, field notes, feeds, and agent discovery aligned to the same reviewed public release record.

Architecture defined

Project pages

A reusable public detail-page pattern for the problem, build, synthetic walkthrough, architecture, prerequisites, limits, and approved links.

release record → project page
Not published

Field notes + feeds

Release-centered articles, RSS, and Atom are planned surfaces. No feed endpoint or subscriber-address store is active on this site.

release record → article / rss / atom
No endpoint

Read-only MCP

A separately hosted MCP service may later expose only approved public releases. GitHub Pages cannot host that server, and none is connected here.

release record → public read operations

Architecture only. This portfolio does not currently publish a release catalog, feed endpoints, or an MCP server. Those surfaces require separate release and hosting approval.

Connect

Bring me the messy workflow.

Especially if it sits between AI, people, and a business system - that is usually where the interesting work begins.

esc