It is the first week of the quarter. You write 5 OKRs. By week 3 you cannot remember 2 of them. By week 8 you have replaced the originals with whatever is on fire today. This tutorial replaces the “rewrite OKRs every quarter and forget them” cycle with a 90-minute setup plus a 15-minute weekly review: AI generates candidate key results, pressure-tests them, and runs the Friday check-in. The audience is individual contributors, founders, freelancers, and managers who set personal goals but lose track.
TL;DR
- Format (from John Doerr’s framework): an objective is “I will [achieve X],” measured by 2-4 key results. Keep 3 objectives, 3-4 key results each, scored 0.0-1.0 at quarter end.
- Cadence: quarterly, not annual. Annual goals create a false sense of time abundance; a 13-week clock forces action.
- Target average is 0.6-0.7, not 1.0. Per Google’s scoring guidance, a clean sweep of 1.0s means you sandbagged. Set each key result so you are genuinely unsure you will hit it.
- AI’s job: generate key-result candidates, critique measurability, run the weekly check-in. Your job: the direction (a values call), the trade-offs, and the courage to drop a stuck key result mid-quarter.
- Make the system persistent: keep the whole workflow in one ChatGPT Project, Claude Project, or Gemini Gem so the model already knows your objectives every Friday (see the tool table below).
What this covers
A reusable workflow to draft, sharpen, and review personal OKRs — 3 objectives, 3-4 key results each, weekly check-ins, mid-quarter pivot. AI handles candidate generation, measurement-quality critique, and the weekly check-in template. You keep the values judgment, the trade-offs you make, and the courage to drop an OKR mid-quarter. The system is designed to ship a quarter where you can name your OKRs without looking at the doc.
Pick the tool that remembers your quarter (June 2026)
The single biggest reason personal OKRs die is that the Friday check-in starts cold: you paste your objectives in again, or you do not, and the habit collapses. Use a persistent workspace so the model already holds your quarter. All three options below carry your objectives, custom instructions, and uploaded files across every chat for the full 13 weeks.
| Feature | ChatGPT Projects | Claude Projects | Gemini Gems |
|---|---|---|---|
| Plan needed | Free works; smoother on Plus ($20/mo) | Pro ($20/mo, $17 annual) | Free works; better on Google AI Pro ($19.99/mo) |
| Default model (June 2026) | GPT-5.5 | Sonnet 4.6 / Opus 4.7 | Gemini 3.1 Pro |
| Persistent memory | Project-only memory across all chats in the project | Custom instructions + uploaded files, read every chat | Persistent instructions; persona-shaped |
| Live doc sync | Re-upload to refresh | Re-upload to refresh | Connects to Google Drive — edit your OKR Doc and the Gem sees it |
| Best for | One project per quarter, everything in-thread | Document-heavy: drop in last quarter’s retro as a file | If your OKRs already live in a Google Doc |
Pick one and stick with it for the quarter. If your goals tracker is a Google Doc, a Gem that reads it live removes the copy-paste step entirely. If you want the model to reason over last quarter’s retro as a file, Claude Projects reads attachments on every chat. For most people, a single ChatGPT Project named “Q3 2026 OKRs” is enough. Deep-dive guides for each tool are in the Related section at the end.
Who this is for
Individual contributors with company OKRs who also want personal ones, founders setting quarterly self-goals, freelancers and consultants without a manager to set them, and managers who want a template their reports can copy. Best when you can name 1 thing you would regret not doing this quarter. Worst when you have not chosen a north star — OKRs amplify the direction you already point in.
When to reach for it
Start of a quarter, after a major project ships and you have capacity for new goals, when joining a new team and writing personal ramp-up OKRs, or 4 weeks before a planned career move. Also useful when 2 quarters in a row have drifted — OKRs reset the drift.
Before you start
- Write your one-year direction in one sentence. “Become a senior engineer specialized in distributed systems” works. “Grow my career” does not.
- List the 3 areas where measurable progress matters most this quarter — usually some mix of work, skill, and personal.
- Block 90 minutes for the draft session. OKR writing is not a 20-minute task.
- Have last quarter’s OKRs open if you had any. The graveyard of dropped key results teaches more than the ones you hit.
- Confirm your weekly check-in slot — same time every Friday, 15 minutes blocked.
Step by step
Open your persistent workspace from the table above and paste each prompt as a separate message so the model keeps the running context. Steps 1-5 are the 90-minute setup; steps 6-7 are mid- and end-of-quarter.
- Direction and themes. Prompt:
My one-year direction is [paste]. This quarter I want to make progress on [3 areas]. Propose 3 objective candidates per area — each as a 1-line statement that an outsider could read and tell whether I succeeded.Pick 3 objectives total, one per area. Aspirational but not impossible. - Key result candidates. Prompt:
For each objective, propose 5 candidate key results. Each must be: a number plus a date, observable by someone other than me, and not just a list of tasks I plan to do.Cut to 3-4 key results per objective. Reject any KR that is really an action item. - Pressure test. Prompt:
Here are my 3 objectives and 12 key results. For each KR, score 1-5 on: measurability, ambition, and personal control. Flag any below 3 and propose a sharper version.Rewrite anything flagged. The “personal control” axis surfaces KRs that depend entirely on other people. To make a soft KR a real stretch, use the OKR ÷0.7 rule: take the number you are confident you can hit and divide by 0.7 (confident in 100 → set 140), so a 0.7 score still means real progress. - Trade-offs. Look at your weekly hours. If your KRs collectively need 25 hours a week and you have 8 hours of slack, cut a KR. Prompt:
I have 8 hours a week of dedicated OKR time. Which KR should I drop and why? - Weekly check-in template. Prompt:
Generate a 5-minute weekly check-in template covering: KR-level progress percent, blockers, one decision needed, one experiment for next week.Save it as a saved prompt (ChatGPT), a project instruction (Claude), or pin it to your Gem, so every Friday it runs with one click. - Mid-quarter review at week 6. Reassess. Some KRs are clearly ahead; some are clearly stuck. Drop, replace, or rescope. Do not hide a stuck KR for the back half.
- End-of-quarter retro. Score each KR 0.0-1.0 on Google’s scale, where 1.0 means fully delivered. Below 0.4 means the KR was wrong or unworkable; above 0.9 means it was too easy. Aim for a 0.6-0.7 average across all key results — a clean sweep of 1.0s means you played it safe. Feed the retro back into next quarter’s step 1.
First-run exercise
- Set OKRs for the current quarter even if you are mid-quarter. The compressed runway teaches discipline faster than starting fresh.
- After the draft, send your 3 objectives to one friend or peer. Ask them to predict your score at quarter end. The conversation surfaces hidden assumptions in 10 minutes.
- Run the first weekly check-in 7 days later, regardless of progress. Skipping the first check-in trains you to skip all of them.
- After week 3, audit: can you name all 3 objectives without looking? If not, the wording is too long. Rewrite shorter.
Quality check
- Each objective is 1 line and would be intelligible to an outsider who knows your role. “Improve scope” is not intelligible; “Lead the migration off Redis for the analytics service” is.
- Each key result has a number and a date. “Get better at X” is not a key result. “Ship 4 deep-dive blog posts by Sep 30” is.
- No key result is a task list. “Read 3 books” is a task; “Apply 2 specific frameworks from books to my work and document the outcome” is a result.
- Personal-control score on every KR is at least 3. KRs that depend entirely on other people belong to those people, not you.
- Weekly check-in is on calendar before quarter starts. Otherwise the check-in slips into “when I have time” and never happens.
- Mid-quarter review is a hard date, not a wish. Block it now.
How to reuse this workflow
- Save the objective prompt, KR prompt, pressure-test prompt, and check-in prompt as 4 snippets. New quarter swaps direction and areas — keep the structure.
- Maintain a “personal OKR archive” with every quarter’s drafts, weekly check-ins, and end-of-quarter scores. Patterns emerge — the KRs you consistently miss are signals about life shape, not effort.
- Annually, look at the 4 retros side by side. The direction sentence usually needs an update once a year.
Recommended workflow
Direction in 1 line → 3 objectives → AI proposes KR candidates → pressure test → trade-off cut → weekly check-in slot booked → execute → week 6 mid-quarter review → end-of-quarter retro → feed retro into next quarter. For tactical task breakdown beneath the KR level, the AI task breakdown tutorial walks through milestone-week-day decomposition. If the framework itself is new to you, Google’s public re:Work guide to OKRs is the canonical primer the 0.0-1.0 scoring here comes from.
Common mistakes
- Writing 7 objectives because everything feels important. Three is a forcing function; seven is a wish list.
- Treating key results as task lists. “Read 5 books” measures input, not output.
- Skipping the personal-control check. KRs that depend entirely on a promotion committee or a hiring market are bets, not OKRs.
- Letting AI write the direction. Direction is a values judgment; AI offers smooth average direction sentences.
- No weekly check-in. Without it, OKRs drift into the doc-graveyard.
- Hiding the stuck KR until week 11. The damage compounds; honest mid-quarter cuts are cheaper.
FAQ
- Should personal OKRs match company OKRs?: Overlap is healthy; identity is suspicious. If your personal OKRs are just your work OKRs, you have no personal direction.
- What about lifestyle OKRs like fitness?: Yes, with measurable KRs. “Lose 8 pounds by Sep 30” works; “Be healthier” does not.
- How many key results per objective?: 3 to 4, matching Doerr’s 2-to-5 guidance. Five is usually too many to track; two leaves the objective brittle.
- What if a KR becomes irrelevant at week 4?: Drop it. Document why in the weekly check-in. Carrying a dead KR is worse than dropping one.
- Should I share my personal OKRs?: Share with one accountability partner. Public sharing creates pressure that can warp the direction.
- What if I score 1.0 on everything?: The KRs were too easy. On Google’s scale a perfect sweep signals sandbagging, not excellence; push harder next quarter and aim for a 0.6-0.7 average.
- Which AI tool should I use to run this?: Any of the three in the table above. Pick by where your goals already live: a Gemini Gem if they are in a Google Doc (it reads the doc live), Claude Projects if you want to attach last quarter’s retro as a file, or a single ChatGPT Project for an all-in-thread setup. All keep your objectives in context across the full quarter.