Deep Runs — your AI intern

What Deep Runs is for

Some jobs are too big for a single chat and too boring to do by hand: judging 400 creators against a rubric, screening a lead list, triaging applicants. Deep Runs takes a spreadsheet and a task and works through every row — returning a verdict, a score, and a reason for each.

How to run one

1. Deep Runs → New deep run.
2. Upload a CSV or XLSX. Zami previews the columns and picks the one that names each row.
3. Write the task and criteria in plain language. Press Optimize to expand a short task into a full brief, and attach a criteria document if you have one.
4. Choose enrichment. On: the intern researches each row to fill gaps before judging (slower, deeper). Off: it reasons over the sheet as-is (fast).
5. Review and export. Watch the grid fill with verdicts — Qualified, Maybe, Rejected — plus scores, per-criterion pass or fail, and rationale. Open any row for its reasoning and sources. Pause, resume, retry failed rows, and export back to a sheet.

Getting good results

Be specific about what "qualified" means — list your must-haves. Start with a sample before running the whole list, especially with enrichment on, since time and cost scale with list size. On high-stakes decisions, treat the Qualified rows as a ranked starting point for a human.