Project Deep Dive

Joan

Microsoft-native, enterprise-first governed digital-work framework: a kernel and governance operating layer for work admission, policy, approvals, evidence, and platform control.

Governed digital workMicrosoft-native substrateWork OperationsPlatform Control

Flagship Case Study

The governed-work thesis behind Joan

Joan is not the Chief-of-Staff product. It is the Microsoft-native governance kernel for digital work: the layer that decides what work is allowed, what policy applies, how approvals happen, and what evidence closes the loop.

This project shows how I think about enterprise AI when Microsoft already owns the work surface. Joan is positioned as the governed digital-work layer that fits that substrate instead of pretending to replace it.

Where the gap is

AI work is spreading faster than governance. Teams need a way to admit work, apply policy, route exceptions, and prove outcomes before they scale digital labor.

What had to be true

The platform needs Work Operations, Workforce Builder, and Platform Control. Those planes make future workers manageable without claiming Joan already ships a live worker suite.

Why it matters

This is the enterprise version of the AI operating-system problem: make digital work governable, auditable, and Microsoft-native before asking senior buyers to trust it.

Positioning Lens

Joan

Joan matters because senior enterprise buyers will not trust AI work unless the operating model is explicit. Work admission, policy, evidence, exception handling, and platform control have to be designed before worker packs are treated as live products.

Signature platform decisions

Separated Joan from the Chief-of-Staff / digital employee direction so the Microsoft-native platform story stays clean.
Made work admission, approvals, evidence, exceptions, and closeout truth the core product language.
Positioned contract and compliance governance as the first domain direction without overstating digital-worker implementation status.

The Problem

What problem was worth solving?

I build things to solve a specific pain, not to decorate a portfolio. This is the pressure that made this project necessary.

Enterprise teams do not need another generic assistant or a shadow control plane competing with Microsoft. They need a governed work layer that can decide what work is admitted, what policy applies, which approvals are required, what evidence closes the loop, and how future digital work stays inside enterprise trust boundaries.

The Build

What I built to solve it

Joan is the Microsoft-native kernel and governance layer for governed digital work. The platform spine is Work Operations, Workforce Builder, and Platform Control: admission, policy, approval routing, exception handling, evidence and closeout truth, cost attribution, and Microsoft 365 / Graph / Azure substrate alignment. It is deliberately not positioned as a Chief of Staff app or a live digital-worker suite.

Architecture 1

Work Operations plane tracks admitted work, approvals, exceptions, evidence, and closeout truth

Architecture 2

Workforce Builder plane defines future worker and capability policy without claiming live worker-pack execution

Architecture 3

Platform Control plane aligns Azure, Microsoft Graph, Azure OpenAI, AI Search, Key Vault, and operator truth

Value

What changed because it existed

This is the clearest evidence of practical value, system leverage, and execution quality.

Platform boundary

Not Chief of Staff · not everything-agent

Joan is the governed digital-work platform layer. Joan4U is the separate Chief-of-Staff / digital employee direction.

Governance

Admission · policy · approvals · evidence

The core product idea is control over what digital work is allowed to happen and how it is proven after the fact.

Enterprise posture

Microsoft-native substrate

The platform is designed to use Microsoft-owned surfaces and Azure controls rather than rebuild a competing enterprise control plane.

First domain direction

Contract and compliance governance

The first active wedge is being shaped around governed contract and compliance work, without overstating worker implementation status.

Tech Stack

What it runs on

The stack matters here because it reflects design choices, constraints, and how the system was intended to scale or integrate.

Microsoft Agent FrameworkFastAPIAzure AI Foundry / Foundry Agent ServiceMicrosoft 365Microsoft GraphAzure Container Apps / App ServiceAzure OpenAIAzure AI SearchAzure Key VaultAzure PostgreSQL / Cosmos DBAzure Monitor / Application InsightsPolicy and approval workflow modelEvidence and closeout ledgerCost attribution model

Build Story

How the thinking unfolded

This is the reasoning path behind the output, not just the finished artifact.

Product realignment

Separated Joan from Chief-of-Staff positioning and clarified it as the Microsoft-native governed digital-work framework.

Kernel baseline

Defined Work Operations, Workforce Builder, and Platform Control as the core platform planes.

Governance model

Centered work admission, policy, approvals, exceptions, evidence, closeout truth, and cost attribution as first-class platform responsibilities.

Commercial wedge

Moved toward contract and compliance governance as the first domain direction while keeping worker implementation claims out of the public story.

Capability Signal

What this project demonstrates

Each project is a proof point. These are the capabilities it most clearly reveals.

Enterprise AI governance

Microsoft-native platform strategy

AI work operating model

Approval and evidence architecture

Product boundary discipline

Continue

Want to go deeper?

Ask the Latif AI Guide about the architecture, the commercial logic, or what this project says about how I approach hard problems.

Back to Projects