The AI OKR Coach skill for Claude

The problem the skill solves is almost always the same. There is an ambition, everyone nods, and two weeks later there is a list of thirteen points nobody can say have been achieved. Often because half those points are not a goal at all but a task: new website live, sales training done, CRM cleaned up. Those are things you tick off, not outcomes you measure. By the end of the quarter there is a green tick next to a task that changed nothing. That is exactly the distinction it removes before the set goes to your team.

It is written for entrepreneurs, management teams and team leaders who want to translate a quarterly or annual plan into something you can steer on weekly. According to the frontmatter it surfaces automatically for questions like setting OKRs, quarterly goals, stretch goals, formulating team goals, how do I measure this goal, or simply: my goals are vague. It belongs to the skill library of 100 free skills we make available through the AI and automation service, with no account and no sales email afterwards.

What you should know upfront: the skill makes your direction measurable, it does not invent it for you. If you do not know where your business should be heading this quarter, you will not get an OKR set out of it, just a neatly worded misunderstanding. And it is explicitly not meant for to-do lists or task management. That is stated literally in the frontmatter, and it is exactly the line most teams cross.

The activity is written down as the goal. This is the classic mistake and by far the most common one. "Launch a new website" feels like a goal, because there is work in it and you can tick it off. But it says nothing about what is different afterwards. The matching Key Result from the skill reads: reach 30,000 unique visitors a month on the new site by 30 September. Still the same website, but now with an outcome attached. Every activity you submit gets rewritten into the outcome you are actually after, or it gets pulled out of the set and put where it belongs: your task list.

The Key Result cannot be scored. The test is simple and strict: can you give it a score between 0 and 1 at the end of the period? "Improve customer satisfaction" cannot, because who decides when it has improved. "Raise NPS from 32 to 50" can, because it contains a starting value, a target value and a measurement point. The skill wants to see all three elements in every Key Result.

There are too many goals. One to three Objectives, three to five Key Results per Objective. These are focus defaults; sufficient coverage, not the count alone, determines whether an exception is justified. So anyone who submits ten Objectives does not get ten back. That feels like a loss the first time, until the moment you notice that a team that truly finishes three things gets further than a team that starts ten.

The KPI is sitting inside the OKR set. KPIs measure the health of a running system: revenue, margin, uptime, occupancy. OKRs drive change. Put your standing monitoring numbers in your OKR set and at the end of the quarter you have mainly measured that the shop is still running. So it pulls those numbers out, with one exception: if you want to deliberately influence a KPI this period, it may join the set as a Key Result.

And then there is the question of who writes the set. OKRs are cascadable but not imposed top down. Doerr sums that up as alignment, not assignment: a team OKR may be linked to a Key Result from a higher Objective, but the team formulates it itself. If it lands in your inbox instead, it is an instruction with a new name.

In diff form, examples from the file show outcomes alongside a separately labelled leading activity indicator. These are editorial illustrations, not measured improvements.

Andy Grove developed OKRs at Intel in the 1970s, building on Management by Objectives. John Doerr learnt the method from Grove and introduced it at Google in 1999. He later popularised it more widely. References to YouTube, Spotify, LinkedIn and the ONE Campaign do not by themselves establish his personal role in each adoption. The core distinction remains: an Objective is what you want to achieve, a Key Result is evidence that you have achieved it.

The file distinguishes qualitative Objectives from measurable Key Results. Three to five Key Results is its default, not proof that a smaller set cannot cover an Objective. Committed targets aim for 1.0; around 0.7 is a common aspirational expectation, not a universal pass mark. The baseline-goal label describes a commitment and is different from a metric’s observed starting value.

The six steps cover context, separating activities from outcomes, choosing one to three Objectives, specifying measurable Key Results, proposing committed or stretch labels, and an execution check. The three context questions do not authorise guessing missing baselines. Supporting activity volumes are separately labelled leading indicators, and the default number of Key Results can be adjusted with a coverage explanation.

The output format is fixed. At the top, period, level and strategic theme in one sentence. Underneath, per Objective, a qualitative sentence in the present tense, with the Key Results underneath it, each with a starting value, a target value and a deadline, and a separate line stating which Key Results are baseline and which are stretch. At the bottom, the execution check with three parts: what has to be true, three sharp risks, and which actions start in the first two weeks. The file then contains a fully worked example from input to output, which you see in shortened form in FIG.01 above.

It closes with five rules under the heading what you do not do, and with five sources. Those five forbidden rules appear below in FIG.04, because that is exactly where a skill's value lies: not in what an AI can come up with, but in what you let it refuse.

Peter Drucker laid the foundation. The Practice of Management from 1954 introduced Management By Objectives: steer on agreed outcomes instead of on supervision. That idea, that a goal is something manager and employee agree on together, still sits in the rule that a team formulates its own OKRs.

Andy Grove made it measurable. At Intel in the 1970s he developed Management By Objectives further into OKRs, described in High Output Management from 1983. His addition is exactly the split this skill insists on: alongside the goal there is always the question of how you know you have got there.

John Doerr brought it outside. He learned the methodology from Grove and introduced it at Google in 1999, then popularised it more widely. References to YouTube, Spotify, LinkedIn and the ONE Campaign do not establish his personal introduction at each organisation. Measure What Matters from 2018 contains the two rules this skill uses most: the distinction between committed targets aiming for 1.0 and aspirational expectations around 0.7, and alignment, not assignment as the answer to whether you can impose OKRs.

Christina Wodtke made it workable. Radical Focus from 2016 is about the part where most teams get stuck: the rhythm after writing it down. This skill adds a first-two-week execution check as a practical proposal; not reviewing until week six can delay course correction, but does not prove that the quarter is lost.

Google re:Work supplied the practice. The public guide Set goals with OKRs is the reason many of the rules in this skill feel familiar to anyone who has worked at a technology company: cascading without imposing, a limited number of Objectives, and an explicit distinction between goals you commit to achieving and goals you deliberately set too high.

Even without installing the skill you can use these rules to clean up your own set: strip out the tasks, put a starting value and a date behind every number, and mark in advance what is baseline and what is stretch. If you want to fit the methodology into a wider plan, the OGSM Plan Builder skill puts goal, strategy and measures on a single page, or the Balanced Scorecard Coach skill, which asks the same question from four perspectives at once.

After that, using it is simple: type your ambition the way you think it, and ask for OKRs. The rougher the input, the better, as long as it contains real information: where you stand now, what is bothering you and which period you mean. It asks itself what is missing, sticking to its own rule of at most three questions. New to this kind of file? Start with what Claude skills are NL and the explanation of installing Claude skills NL. Want to write one yourself? The structure is in writing a SKILL.md NL. And the wider explanation of working with AI is in the knowledge base.

There are also situations where you are better off leaving it be. The first is stated literally in the frontmatter: not for pure to-do lists or task management. Ask it to plan your week and you get a framework applied to a problem that needs a different solution. The second is an organisation that has not chosen a direction yet. OKRs make a choice measurable, they do not make the choice for you. Anyone who still has to decide which market or offer to focus on gets further with the SWOT Analysis skill. The third is a period shorter than a month. A set with stretch goals and an execution check needs time to mean anything.

One more honest limit: the skill makes your set better, not your execution. A sharp OKR set that you close in week two and only reopen in week twelve does precisely nothing. That is also why Radical Focus is among the sources, and why the execution check asks which actions start in the first two weeks. If you want to lock in the rhythm, you can combine the set with the RICE Prioritisation skill for the order of the work, and with the WOOP Method Coach skill NL to name, per Key Result, the obstacle you are going to run into in advance.

And finally: the skill does not write your ambition for you. If you leave "where do you want to go this quarter" blank, no inspiration comes out of it. What happens instead is that the three context questions from step 1 force you to make that choice anyway, and that is often the most useful part of the conversation.

A sharp OKR set makes visible where things get stuck, but it does not solve it. If your set has a Key Result about more enquiries or more sales calls, the real work sits a level earlier: being found by the people looking for you. The free SEO scan shows in seconds where your site stands. And if you want to talk further about what else AI can do for your business, from loose skills to full automation, we simply do that in a conversation.