Gemini Turns 14 Messy Bullets Into a Finished Self-Assessment

AI WorkStack · Article version of our video · Published 11 November 2026

The result first: one employee's year — fourteen messy bullets, typed on a phone across ten months — becomes a finished year-end self-assessment: a real Google Doc with sections, strong lines and an evidence-based AI-impact box, built and exported inside Gemini, without opening Google Docs. Here is the whole workflow, with the exact prompt.

The one setup step (and the one technique that runs everything)

Setup: the first time you do this, Gemini asks you to connect the Google Drive app — one tap. The document Gemini builds lives in your Drive, which is also how you export it at the end.

Technique: you don't edit the document — you instruct it. Every instruction starts with the words "in the document", and every edit hands you back an updated version of the Doc. If an edit seems to go nowhere, that missing phrase is almost always why.

Step 1 — Paste the bullets, get a real Doc

Paste your year's bullets into Gemini with this prompt (use it with your own bullets, and keep your real numbers exactly as written):

Turn these work notes into a self-assessment Google Doc. Sections: Summary, Key Results, Teamwork, Growth. Keep every number exactly as written. If a bullet has no number, keep it short.

Gemini returns a Google Doc with those four sections, built from your bullets — every number preserved as written, and bullets without numbers kept short instead of padded out.

Step 2 — Fix the order, strengthen the lines

Now instruct the document, in the document: move your best result to the top of Key Results, and tighten the weak, vague lines into specific ones. The order rule that matters most: your best result belongs in the first three lines of the document — that is what a reader actually reads. Each instruction returns an updated version of the Doc, so you can see every change land.

Step 3 — The "AI-driven impact" box: evidence, not hype

Many review forms now include a box about AI-driven impact, and it is the box most people fill with claims. Fill it with measurements instead: what you used, on which work, and what it changed that you can point to — a time saved, an error rate, a turnaround. The box grades measurements, not claims. If you cannot point to it in your own notes, it does not go in.

Step 4 — The human pass, then export

Gemini writes it; you check it. Read the Doc once like a manager would, then run the 60-second final pass:

• Numbers match — every figure in the Doc is the figure in your notes.
• Claims are true — anything you cannot point to in your own notes gets deleted.
• Best result first — it sits in the first three lines.

Then open the Doc and export it into your review form. In the video, the whole job — for a one-page bullet list of this size — took about 9 minutes end to end.

The 5 rules to save

1. Result first — your best result in the first three lines.
2. Instruct, don't edit — "in the document…" or the Doc never changes.
3. Gemini writes it, you check it — anything you can't point to in your own notes gets deleted.
4. The AI-impact box grades measurements, not claims.
5. Sixty-second final pass: numbers match, claims true, best result first.

Common mistakes

• Opening Google Docs and hand-editing. The workflow lives in Gemini: instruct the document, and let each edit return an updated version.
• Forgetting the words "in the document". Without them, Gemini may answer in chat instead of changing the Doc.
• Letting the AI improve your numbers. The prompt says "keep every number exactly as written" — and the final pass exists to verify it actually did.
• Filling the AI-impact box with hype. "Used AI to be more productive" earns nothing; a measured before-and-after does.
• Skipping the human pass. The drafter is fast, not accountable. You are the one signing the review.

FAQ