AI for OKRs: Set, Track, and Review Goals with AI

How to use AI to set better OKRs, track progress automatically, and run quarterly reviews that actually change behavior — not just fill a spreadsheet.

Mark Rachapoom

March 26, 2026·7 min read

OKRs (Objectives and Key Results) are one of the most widely adopted goal frameworks in tech — and one of the most widely misused. Teams write OKRs in January, paste them into a spreadsheet, and look at them again in December to notice they forgot about half of them. This is not goal management; it's goal theater.

AI doesn't fix organizational alignment problems. But it significantly reduces the friction in three specific parts of the OKR process: writing good key results (most teams write bad ones), tracking progress without manual overhead, and running reviews that produce learning.

Why Most OKRs Fail#

Before the AI part, a quick diagnosis. OKRs fail for a few consistent reasons:

Too many objectives. Five objectives is already too many. Three is better. The constraint forces real prioritization.

Vague key results. "Improve customer satisfaction" is not a key result. "Increase NPS from 32 to 45 by June 30" is. Key results should be measurable, have a deadline, and leave no ambiguity about whether they were achieved.

No progress tracking. If you only look at OKRs quarterly, there's nothing to course-correct in between. Weekly check-ins on key result progress are what turn OKRs into behavior change.

No accountability. Every OKR should have a single owner. "Sales team" is not an owner. "Marcus" is.

AI helps with writing and tracking; the organizational accountability part is up to you.

Using AI to Write Better Key Results#

The most common mistake in OKR writing is confusing activities with results. "Launch the new pricing page" is an activity. "Increase trial-to-paid conversion rate from 12% to 18% by Q2" is a result.

AI can help you make this distinction:

"Review these draft key results and improve them.
For each one:
1. Does it measure an outcome, not an activity?
2. Is it specific enough to know unambiguously when it's achieved?
3. Does it have a deadline?
4. Is it ambitious but achievable?

If a key result fails any of these tests, rewrite it.
If you can't rewrite it without more context, explain what information you need.

Draft KRs:
- [paste your draft key results]"

Running draft OKRs through this check catches most of the common problems before they get locked in for the quarter.

Generating KR options:

When you have an objective but aren't sure how to measure it, AI can generate candidate key results:

"I have this objective: 'Make our enterprise customers wildly successful'
Generate 5 candidate key results that would measure meaningful progress 
toward this objective. Focus on outcomes, not activities."

You pick the ones that resonate and fit your current strategy.

Setting Up OKR Tracking in DenchClaw#

OKRs that live in a shared Google Sheet are better than OKRs in no system. But OKRs connected to your actual data — where progress updates automatically from your CRM and metrics — are dramatically better.

Create an OKRs object:

"Create an OKRs object with fields:
- Quarter (enum: Q1 2026, Q2 2026, etc.)
- Type (enum: Objective, Key Result)
- Parent Objective (relation to OKRs, for key results)
- Owner (text)
- Target Value (number)
- Current Value (number)
- Unit (text: %, $, count, etc.)
- Start Value (number)
- Progress (computed from start/current/target)
- Status (enum: On Track, At Risk, Behind, Complete)
- Notes (richtext)
- Due Date (date)"

With this structure:

Objectives group their key results
Each key result tracks toward a specific number
Progress is visible and queryable
History is preserved quarter over quarter

Connecting key results to data:

For key results that can be computed from your CRM data, DenchClaw can update them automatically:

KR: "Increase pipeline coverage from 3x to 5x" → auto-calculated from deals in your CRM
KR: "Close 10 enterprise deals" → count of deals in 'Won' stage with value > $X
KR: "Reduce time-to-close from 45 to 30 days" → average of close_date - created_date for Won deals

Set up a weekly cron: "Update all OKR current values from the CRM data and flag any KRs that have moved to At Risk status."

Weekly OKR Check-ins#

The discipline that makes OKRs work is a weekly ritual — not a 30-minute meeting, but a lightweight check-in that takes 5 minutes.

Prompt for weekly check-in:

"Generate a weekly OKR check-in summary for week of [date].
For each key result in the current quarter:
- Current value vs. target
- Is it on track to hit the target by the due date?
- Has status changed from last week?
- Any new blockers?

Highlight any KRs that moved from On Track to At Risk.
Note any KRs that are ahead of target."

This runs automatically in DenchClaw each Monday and sends a Telegram message. You spend 5 minutes reviewing it; if something's off track, you address it during the week when there's still time to course-correct.

Quarterly Reviews#

End-of-quarter OKR reviews often feel ceremonial — going through the list and scoring 0% to 100% for each KR, with everyone knowing the scores in advance. A good quarterly review should produce learning that changes how you set goals next quarter.

Review prompt:

"I'm running the Q[N] OKR retrospective. 

Final scores for each KR:
- [KR name]: [final score]% — [brief note on why]

For the retrospective, help me answer:
1. What KRs did we hit and why? What can we systematize from this?
2. What KRs did we miss and why? What would we need to change to hit them?
3. Were any KRs set too easy (hit 100% easily)?
4. Were any KRs set too hard (impossible from the start)?
5. What's the most important learning that should change how we set Q[N+1] OKRs?

Don't just describe what happened. Extract lessons."

The output from this review becomes the input for the next quarter's OKR-setting session. The goal is a learning loop, not a grading exercise.

OKRs and the gstack Workflow#

DenchClaw's gstack workflow integrates with OKRs naturally. The Reflect phase of gstack — the weekly retrospective — generates data that feeds into OKR tracking. Engineering-specific OKRs (sprint velocity, deployment frequency, test coverage) are tracked by gstack and fed into the same OKR system.

This means your product OKRs and engineering OKRs are visible in the same place, queryable together, and connected to the actual work being done.