Threadbase
A proactive portfolio-intelligence platform for enterprise change leaders. It reads the reports, presentations, assessments, and transcripts teams already produce and surfaces changes, risks, dependencies, and conflicting accounts with every finding tied to its evidence.
My roleIndependent build for enterprise change portfolios
The Problem
Enterprise change leaders receive fragmented, contradictory, and stale reporting across their programs. The signal exists in the decks, assessments, and meeting transcripts their teams already produce, but no one reads all of it, and status tools only answer the questions someone knows to ask. The dangerous question is the one nobody asks.
The Product Decision
Threadbase speaks first. Instead of waiting for a user to query it, the product reads everything the portfolio produces, reconciles it into one governed portfolio view, and surfaces what changed, what needs judgment, and where programs conflict. Two rules govern the design:
- Evidence first. Every material finding links to the source page it came from. Incomplete evidence is shown as incomplete rather than presented as fact, and a true all-clear is distinguished from missing or stale information. The product never renders a failure as a zero.
- Human governance. Disputed or incomplete findings route to people for judgment. Reviewers accept, edit, or reject with recorded reasons, and every decision remains auditable.
What I Owned
Threadbase is an independent build, designed for the enterprise change portfolios I work in professionally. I own the product vision, the system design, the specification, the build direction, and the deployment.
What Is Built
I have built and deployed the core Azure workflow that turns portfolio materials into source-linked findings, routes disputed or incomplete information for human review, and reconciles approved findings into one governed portfolio view. That covers the full chain in production: document ingestion with malware and credential scanning, extraction of claims with page-level provenance, a governed review queue with deterministic ranking, publication gates with evidence-derived confidence bands, and a proactive leadership view that leads with what changed and what needs attention.
I am now extending the product into executive briefings and deeper cross-portfolio analysis: the intelligence views that let a leader move from a finding to its evidence in two clicks, and governed executive outputs built from the reconciled portfolio state.
Trust Controls
The platform treats trust as a product requirement, not a disclaimer:
- A finding publishes only after passing explicit gates, with a confidence band derived from its evidence.
- Identical documents in produce identical findings out, so runs are reproducible and diffable.
- Reporting periods, snapshots, and baselines make change over time a first-class object: the same two snapshots always produce the same change set.
- Every mutation to the portfolio state is audited, and relationship history is append-only.
AI-Native Delivery
The product is specified before it is built: a normative specification defines every guarantee, and each guarantee is covered by at least one machine-verifiable proof that runs against a clean checkout and the deployed environment. Coding agents implement against that specification under my direction; a single gate command is the proof of state, and CI runs the identical command.
Microsoft Ecosystem Fit
The platform runs on Azure Container Apps and PostgreSQL with Entra-based identity and managed identities throughout, with infrastructure as code in Bicep. Document understanding uses Azure AI Content Understanding, so dense corporate slides resolve into findings that still point at the exact source page.