TL;DR
Most OKRs die the same way: written in week 1, ignored by week 4, and rediscovered as a guilt artifact in week 13. This tutorial fixes the cadence, not just the wording. You will use an AI chat model (Claude Opus 4.7, GPT-5.5, or Gemini 3.1 Pro — all current as of June 2026) to draft 3 objectives with 3 measurable key results each, pressure-test them for slip risk and sandbagging, run a 15-minute Monday check-in that marks every KR green / yellow / red, and produce an honest end-of-quarter retro. The AI does the drafting and grading math; the weekly habit does the rest. Copy-ready prompts are below.
What this covers
The pain: you wrote OKRs in week 1, never looked at them in week 4, and by week 10 the quarter ended with three measured outcomes and seven aspirational paragraphs. This workflow treats OKRs as a living doc you read every Monday, not an artifact you write once and never reopen. The trick is the check-in cadence: a 15-minute weekly walk through every key result that surfaces a red line in week 4, when you still have nine weeks to recover, instead of week 8, when you do not.
We will also lean on the actual scoring grammar Google publishes in its re:Work goal-setting guide: a 0.0–1.0 scale, 3–5 objectives with about 3 key results each, and a target average of 0.6–0.7 because a 1.0 average means the targets were too soft.
Who this is for
- Team leads who set quarterly OKRs for an engineering, product, or design team.
- ICs whose company asks for personal OKRs and wants them to ladder to the team’s.
- Founders aligning a small team against three priorities.
- Chief-of-staff roles drafting cross-functional OKRs that span several squads — the same cadence catches misalignment before it costs a sprint.
Pick the AI model first
Any frontier chat model can draft a competent OKR set. The differences that matter for this workflow are context window (can it hold last quarter’s doc plus this quarter’s strategy notes?) and honesty under pressure (will it tell you which KR is sandbagged?). All three below were current as of June 2026.
| Model | Plan that unlocks it | Context window | Why pick it for OKRs |
|---|---|---|---|
| Claude Opus 4.7 | Claude Pro $20/mo, Max $100/$200 | 1M tokens | Strongest at the pressure-test step; pushes back on vague KRs without being prompted hard |
| GPT-5.5 (Thinking) | ChatGPT Plus $20/mo | ~320 pages in-app on Plus; full 1M only on $200 Pro | Cleanest tables and scoring math; default if your team already lives in ChatGPT |
| Gemini 3.1 Pro | Google AI Pro $19.99/mo | 1M tokens | Best if your OKR doc lives in Google Docs/Sheets; reads the whole quarter folder in one paste |
For most teams the recommendation is simple: use whatever you already pay for. The drafting quality gap between these three on a one-page OKR doc is smaller than the gap between running the weekly check-in and not running it. If you want a deeper model comparison, see our ChatGPT vs Claude vs Gemini guide.
Pick the doc surface second
OKRs live or die on whether the team reopens the doc every Monday, so put it where the team already works.
| Surface | Cost (as of June 2026) | Good for | Watch out for |
|---|---|---|---|
| Google Docs / Sheets | Free / Workspace seat | Fastest start; pairs with Gemini 3.1 Pro | No native scoring rollup; you maintain the table by hand |
| Notion | Free / Plus / Business $15 seat (AI bundled in Business since early 2026) | Teams already in Notion; AI writing and rollups in one place | Custom Agents bill credits since May 4 2026; standard AI writing does not |
| Dedicated OKR tool (Mooncamp, Weekdone, Lattice) | ~€8/user/mo and up | 20+ person orgs that want check-in reminders and rollups built in | Overkill for a single team of 5; another login to maintain |
Switching surfaces mid-quarter breaks the habit. Pick one before you draft, not after.
Before you start
- Have last quarter’s OKRs and actual scores open. No last quarter? Gather any quantitative outcome you can defend.
- Know the company’s top 3 priorities for the quarter. If you cannot name them, do that first; team OKRs that contradict company priorities die by week 6.
- Decide committed vs aspirational up front. Per What Matters, committed KRs are pass/fail and must hit 1.0; aspirational KRs are stretch goals where 0.7 counts as a win. Label each KR so the end-of-quarter score is not an argument.
- Block 90 minutes for drafting and 15 minutes every Monday for the check-in. Both go on the calendar before you start.
- Lock the scoring rubric: 0.0–1.0, with the Google green/yellow/red bands below.
The scoring bands you will use every Monday
These are Google’s published bands, which keep the Monday check-in to a 5-second judgment per KR:
| Score | Color | Meaning |
|---|---|---|
| 0.7–1.0 | Green | On track; what we’re doing is working |
| 0.4–0.6 | Yellow | At risk; we may need to change approach |
| 0.0–0.3 | Red | We likely won’t meet this KR |
For an aspirational KR set, a quarter-average around 0.7 is the target. Consistently averaging 1.0 means you sandbagged; consistently below 0.4 means the targets were unrealistic or the plan never got resourced.
Step by step
- State strategic context in 3 sentences. Example:
Team is platform infra; company priority is reducing onboarding time; our quarter must contribute to that without breaking 99.9% uptime. - Ask AI to draft 3 objectives that ladder to a company priority, each with 3 measurable KRs. Reject any KR whose verb is “improve”, “drive”, or “support”.
- Demand a baseline and a target for each KR. Example:
p95 onboarding latency from 8.2s to under 4s, notmake onboarding faster. - Pressure-test the set. Ask AI which KR is most likely to slip and which is sandbagged. Adjust until both answers feel honest.
- Label committed vs aspirational, publish the doc, and pin it to the Monday agenda. Every Monday: 15 minutes, walk every KR, mark green / yellow / red.
- When a KR turns red two weeks running, re-plan — do not re-write. Ask AI for 3 interventions that could move the number; pick one and commit.
- End of quarter: paste the final numbers and score. Ask AI for a retro that separates “missed because of execution” from “missed because the target was wrong”. That distinction is the input to next quarter.
Copy-ready prompts
Save these four as snippets. Each new quarter, swap only the bracketed context.
Drafting prompt
You are an OKR coach. Strategic context:
[3 sentences: my team, the company priority we ladder to, a constraint we cannot break].
Draft exactly 3 objectives, each with exactly 3 key results.
Rules:
- Every KR has a baseline number and a target number.
- Reject verbs like "improve", "drive", "support". Use a measurable metric instead.
- At least one KR per objective must be a leading indicator I can steer mid-quarter.
- Label each KR [committed] (must hit 1.0) or [aspirational] (0.7 = success).
Output as a one-page table: Objective | KR | Baseline | Target | Type | Leading/Lagging.Pressure-test prompt
Here is my OKR set: [paste table].
1. Which single KR is most likely to slip, and why?
2. Which KR looks sandbagged (target too easy for a quarter)?
3. Which objective does NOT clearly ladder to the company priority I gave you?
Be blunt. I would rather fix it now than score it in week 13.Monday check-in prompt
Here is my OKR set with this week's numbers: [paste table + current values].
For each KR give a 0.0-1.0 score and a green/yellow/red using these bands:
0.7-1.0 green, 0.4-0.6 yellow, 0.0-0.3 red.
Then list only the KRs that are red OR yellow two weeks running, and for each
propose 3 concrete interventions for next week.Retro prompt
Final OKR numbers for the quarter: [paste table + final values].
Score each KR and the quarter average. For every miss, classify it as
"execution miss" (right target, wrong path) or "wrong-target miss"
(unrealistic or misframed from the start). End with 3 changes for next quarter.First-run exercise
Pick a quarter that already ended. Reconstruct what your OKRs should have been from the outcomes you actually remember, then run the drafting prompt on that hindsight context and compare it to what you wrote at the time. Two things show up: the KRs you would write now are sharper than what you wrote then, and you forgot at least one outcome that should have been an objective. Both findings make next quarter better, and the exercise calibrates how aggressive the AI’s default targets are versus your real velocity.
Quality check
- Every KR has a baseline number and a target number. No baseline means no honest scoring.
- Every objective ladders to a company priority in one sentence. If you cannot draw the line, the objective dies by week 6.
- At least one KR per objective is a leading indicator. Lagging-only OKRs cannot be steered mid-quarter.
- The set has 3 objectives, not 5. Five objectives means zero priorities and a quarter of context-switching. (Google allows 3–5; for a single team, 3 is the focused upper bound.)
- Each KR can be measured in 5 minutes from a real source. If measuring takes half a day, the weekly check-in will not happen.
- Each KR is labeled committed or aspirational, so the end-of-quarter score is a number, not a debate.
How to reuse this workflow
- Keep a
quarters/folder with the doc for every quarter. Diff two adjacent quarters; the pattern of misses is the real signal. - Track an “OKR hit rate” across quarters. A quarter-average below 0.3 means targets are too aggressive; above 0.8 means sandbagged.
- After 4 quarters, paste the four retros into the next drafting prompt. The model calibrates to your team’s actual range instead of generic stretch targets.
- Share the drafting prompt with one peer team. Comparing drafts catches misaligned priorities before the quarter starts.
Common mistakes
- KRs with verbs like “improve” or “drive”. They expand forever and cannot be scored.
- Writing the doc in week 1 and never reopening it. The weekly check-in is the entire workflow; without it, OKRs are decoration.
- Five objectives for one team. Three is the focused upper bound.
- No baseline number. End-of-quarter scoring becomes a debate instead of a number.
- Re-writing OKRs in week 7 because they slipped. Re-plan the path to the same KR; do not change the KR after the fact.
- Sandbagged targets. The team learns that hitting 1.0 means the target was too easy, and stops reaching for stretch.
- Mixing committed and aspirational without labels. A 0.6 is a pass for an aspirational KR and a failure for a committed one. Label them.
FAQ
- Which AI model should I use to draft OKRs?: Any of Claude Opus 4.7, GPT-5.5, or Gemini 3.1 Pro (all current as of June 2026) drafts a solid set. Use whichever you already pay for; Claude tends to push back hardest on vague KRs, GPT-5.5 produces the cleanest scoring tables, and Gemini 3.1 Pro is handiest if your doc lives in Google Workspace.
- What if the company has no OKRs?: Build team OKRs that ladder to a written company priority. If none is written, draft one for the team’s owner to react to before you set yours.
- How aggressive should KR targets be?: For aspirational KRs, 0.7 is “ambitious target hit”; if you regularly average 1.0 the targets are too soft. Committed KRs are different — they are pass/fail and must hit 1.0.
- What if a KR depends on another team?: Mark it
depends on Xand list a separate ask in the doc. Do not let a dependency quietly hide a slip. - How long should the OKR doc be?: One page. If it does not fit on one page, the team will not read it weekly.
- Can I pivot mid-quarter?: Once per quarter, with an explicit note on why. More than once means the strategy itself is unstable, not the OKRs.
Related
Tags: #okr #Planning #Quarterly #Tutorial