# workdna *AI-era work intelligence, for the person doing the work.* > Work intelligence that classifies what you actually do (innovation, maintenance, glue, tech-debt, firefighting), so you can see your trajectory, your automation risk, and the skills worth growing into. Employee-first by design, no leaderboards, ever. ## The problem Most tools count activity. PRs merged. Lines changed. Story points closed. None of these explain whether the work being done is actually building new value or just keeping the lights on. In an AI-leveraged world, that gap is the difference between roles that scale and roles that get automated. Engineers, PMs, designers, analysts, and operators have no way to see what kind of work fills their week. Their managers see even less. By the time the answer matters (a layoff, a re-org, a career pivot), it is too late to redirect. ## How it works 1. You connect your work surface (today: GitHub OAuth). 2. workdna pulls the last 90 days of commits and PR reviews. 3. Claude classifies each unit semantically into one of five categories with a one-sentence rationale and a confidence score. 4. Your personal dashboard shows the breakdown, the trajectory over time, an automation risk score per commit, and a list of suggested skill shifts. Every classification is clickable. You see the reasoning. You can disagree. ## The five categories - **innovation** -- new capability, design, or product surface - **maintenance** -- bug fixes, dependency bumps, keeping the lights on - **glue** -- code review, mentoring, unblocking, design docs (the threads that hold teams together) - **tech-debt** -- refactoring, infrastructure repayment, internal cleanup - **firefighting** -- incidents, hotfixes, reactive work ## Four refusals workdna is built on The full manifesto, signed and dated, is at https://tryworkdna.com/principles. 1. **Employee-first.** The personal dashboard is the product. Team views aggregate personal views; they never replace them. 2. **No leaderboards.** Managers see team-level distributions, not names. Not a toggle, not a tier, not a workaround. 3. **Semantic, not metadata.** Classification reads the actual work (diff, commit message, PR description) and explains itself. Jira labels describe whatever someone typed. 4. **Credit invisible work.** Reviews, mentoring, unblocking, design docs are first-class contributions, not footnoted in an "other" bucket. ## Roadmap Starts with software engineers via GitHub because commits are the densest, most structured signal of knowledge work. The same classification layer maps cleanly onto: - **next** -- product (Linear, Jira) - **later** -- design (Figma) - **later** -- analysts (notebooks, dbt) - **later** -- ops (calendar, docs) Engineering is the right place to prove the method. It is not the place where workdna ends. ## Live example A sanitized public example using OpenStatus's real commits is available at https://tryworkdna.com/openstatus -- 489 commits, three contributors, the last 6 months classified end-to-end. ## A note from the founder I have spent 5+ years building software at big enterprise tech and at an early-stage startup, always inside a team. In 2020 at Samsung, I watched layoffs hit engineers, PMs, and designers whose work was mostly repetitive. In 2023 at Turvo, after ChatGPT, another round. Different reason on paper, same pattern underneath. The people who lost their jobs were doing duplicative or automatable work and had no way to see it coming. That is why I built workdna. Human work needs to be measured very differently in the AI era. Counting commits and lines of code was never enough. It is definitely not enough now. -- Kunal Mahato ## Get in touch Replies come from the founder, not a queue. Use the contact form at https://tryworkdna.com/#contact, or email via the link there. ## Links - Site: https://tryworkdna.com - Live example: https://tryworkdna.com/openstatus - Principles: https://tryworkdna.com/principles - Source code: https://github.com/mahatokunal/work-intelligence - llms.txt manifest: https://tryworkdna.com/llms.txt