Writing

Thought leadership and co-authored work on AI, delivery, and organisational capability.

Start here: a home for retrospectives that keeps objectives in view

This article introduces Home in NABS Retros 0.11.0, a VS Code extension that runs retrospectives inside the team’s own repository: one place to start, with Objectives and New Retro as the two main actions and the Retrospective Maturity Curve shown front and centre. Home flags open objectives that need review, suggests the next retro date from the team’s cadence and checks set-up, on the basis that reviewing objectives every retro is how a team moves past Reflection.

From ritual to reinvention: giving the retrospective a guided path

This article returns to the Retrospective Maturity Curve (Ritual, Reflection, Realignment, Reinvention) and introduces NABS Retros, a VS Code extension being built at Net Advantage that keeps each retro as Markdown in the team’s repository, uses GitHub Copilot to draft evidence-based summaries and turns action items into GitHub issues. Version 0.8.0 adds a guided dashboard that walks the team through six stages mapped to the curve, with soft gating, separate facilitator and participant views, and nothing sent on the team’s behalf.

Read on LinkedIn

AI Suggests, People Decide: Tying retrospectives to the objectives that matter

This article describes the Objectives feature in NABS Retros, which ties each retro to the organisation’s strategic objectives and their acceptance criteria, including automatic measures drawn from Git and GitHub data. GitHub Copilot suggests an evidence-only assessment, but the team records the verdict, and if it disagrees with Copilot it must write down a reason that is kept alongside the AI’s view for leaders to read.

Read on LinkedIn

Keeping Leaders in the Loop: From the room to the people who decide

This article covers Close & share, the final stage of a NABS Retros retro, which archives the whole retro to a GitHub Discussion so that its objectives, Copilot’s assessment, the team’s verdict and reasons, action items and insights to share upwards reach leaders without them opening VS Code. It also explains how archiving resumes safely after an interruption, how retro files are removed only through a staged, reviewable deletion, and the feature’s current limits.

Read on LinkedIn

Management Bandwidth: Five Early-Warning Metrics That Show You’re the Bottleneck

This article treats management bandwidth as an operational constraint and sets out five early-warning metrics for routine agile software delivery decisions: iterations to understanding, mean time to decision, number of reminders, frequency of broken commitments, and quality debt accrual rate. Each metric is scoped to day-to-day work the manager already controls, with thresholds that mark when they have become the bottleneck, and a private sprint tally that flags two or more metrics in the yellow for two consecutive weeks so the manager can offload work before the team has to say so.

Read on LinkedIn

From Mills’s Copilot to GitHub Copilot: A 1970s Idea in My Day-to-Day .NET Work

Harlan Mills’s “copilot” in the surgical team (as Fred Brooks told it) was a second skilled programmer supporting the lead developer - not a luxury most teams could afford. This article connects that 1970s idea to GitHub Copilot in modern .NET work: comparing OAuth approaches, driving consistent package updates across repos, and using MCP for SDLC tasks so the developer in the chair stays fully responsible while shipping faster.

Read on LinkedIn

Core Capabilities for an Effective AI Implementation

A leadership guide to evolve your organisation from data to intelligence

Co-authored with Amanda Schreyer.

This whitepaper outlines seven core capabilities that turn static data into dynamic intelligence: data and content curation, AI-enhanced search and retrieval, intelligent matching and recommendation, context-aware data entry, AI-driven process automation, continuous learning and governance, and amplifying talent. It is written for leaders who want to scale AI responsibly without chasing models for their own sake.

Download the PDF