All work

AI-assisted engineering workflows

Most of the coordination cost on a project comes from four tools that do not talk to each other: designs in Figma, plans in Jira, decisions in Confluence, and the truth in the repository. I built workflows that read all four and tell me where they disagree — and every one of them ends with me reviewing the output.

Role
Workflow designer & primary user
Period
Current, ongoing
Applied to
Live client delivery
Review model
AI review stage, then human approval

Cross-tool daily digest

Reviews Jira, Confluence, the Git repositories and Figma, then summarises what changed since the last run. It replaces the half hour every morning spent reconstructing what happened while I was somewhere else.

Figma to Jira

Takes a selected epic and a Figma frame and prepares implementation tasks for the disciplines actually involved — backend, frontend, DevOps and whatever else the design implies. The output is a draft, and it goes through a separate AI review stage before it reaches mine.

Design coverage gap analysis

The most useful of the four. It treats Figma as the project's current product source of truth, compares the designs against both existing Jira work and the source code, and identifies design elements that have not been implemented. For each gap it checks whether an appropriate Jira task already exists, and drafts one where it does not.

I perform the final review. The workflow does not autonomously change the delivery plan — the point is to surface the gap, not to quietly fill it.

Jira to initial implementation

Uses an existing Jira task to prepare an initial code implementation, again with a separate AI review stage before human review. It gets the boring 60% written; it does not get to decide the remaining 40% is fine.

Where I stand on this

These are practical everyday tools, not a claim to special status. Nothing here deploys to production autonomously, and no unreviewed AI-generated code or task is accepted automatically. What they genuinely buy is the removal of repeated coordination work, and visibility into the gap between what a product was designed to do and what has actually been built — which is a gap that, left alone, is normally discovered by a user.