Home / Blog
Project-delivery field notes and practical AI workflows. The newest pieces may live on Substack or in selected Medium publications; the hub keeps one chronological reading map.
AI can summarize what a project system captures, but it cannot reliably recover delivery evidence, ownership, blockers, handoffs or dependencies that were never structured or captured.
Read the article →The tasks were still visible. Dependencies, blockers, ownership and cross-functional handoffs were not, turning a tooling constraint into a delivery-visibility problem.
Read the article →A build can satisfy every acceptance criterion and still fail the question the user came to answer. Here’s the 20-minute AI pass I run now.
Read the article →Trello was still working. The project had simply become too complex for what the board could make visible, and the switch to Jira changed more than task tracking.
Read the article →Two teams agreed on every feature in the product. The disagreement was in the limits attached to those features — and nobody owned the decision.
Read the article →A requirement can be specific, structured, approved by five people, and still leave the most important decision unmade.
Read the article →The requirement was correct in every artifact. The platform had a different definition of the same word, and nobody had written that down anywhere.
Read the article →We shipped role-based team permissions on a product one person uses alone. Every reviewer signed off.
Read the article →One prompt. Two jobs. Only one of them worked, and the broken half was invisible.
Read the article →RAID has four letters and most risk registers run on two. A 15-minute assumption audit.
Read the article →Ask it to improve its own draft and you get polish. Ask it to review someone else's and it takes the thing apart.
Read the article →Building the register was never the hard part. Keeping it describing the project you're actually running is.
Read the article →Every RAID log has four letters. Most only use three.
Read the article →It isn't a skill problem. It's that you're in the document.
Read the article →Building it was never the hard part.
Read the article →Six patterns that look like agile delivery and produce none of it, with the first fix for each.
Read on Medium →Stop writing clever prompts. Keep a brief per workflow instead.
Read the article →Most people accept the first draft or throw it out. The usable work is in round two.
Read the article →Four failure modes from a food traceability project, and the elicitation process we rebuilt because of them.
Read the article →An AI note-taker gives you a tidy summary. It doesn't give you owners, deadlines, or a decision anyone can be held to.
Read the article →The hard part was never the writing. It's keeping the report honest when AI makes “everything's fine” sound so convincing.
Read the article →The prompt I use to turn a wall of specs into a one-page brief — and the one mistake that makes most AI summaries worse than useless.
Read the article →Stop asking which AI is “best.” The real skill is matching the tool to the task.
Read the article →What used to take me hours the night before a steering meeting — and how I do it now.
Read the article →Follow The Tech Manager's Playbook on Substack for the weekly issue.
Subscribe on Substack →