The AI Lean Canvas Strategist skill for Claude

You download the zip at the top of this page, put it in your Claude environment, and from that moment Claude builds canvases using this logic.

The problem the skill solves is old and stubborn. Anyone with an idea wants to build, and anyone who has to plan writes a business plan nobody reads. Both routes hide the same question: which assumption breaks this idea if it turns out to be wrong? That is exactly what the Lean Canvas was made for. It forces you to put problem, customer, proposition and revenue model side by side on one page, so the weak spots become visible before you put months of work into it. Our strategist adds the discipline that the canvas often misses in practice: assumptions are explicitly labelled with an ASSUMPTION prefix, and the three most dangerous ones each get a concrete experiment.

This skill is written for entrepreneurs who have just had an idea, want to pivot, are launching a new product line or preparing an investor pitch, and for anyone who feels their business model is not quite right yet. For that pitch, incidentally, the canvas stops at the content: if the nine blocks hold up, you have your story, and the Pitch Deck Builder NL turns that story into eight slides, each with a core line and the counter question next to it. It is part of the skill library NL we offer for free through the AI and automation service, with no account and no sales email afterwards.

One thing you should know beforehand: the tone is deliberately not nice. The style rules in the file dictate that the skill is bluntly honest about weak blocks, not diplomatic. An unfair advantage you do not have does not get dressed up, it gets written down as what it is: an assumption you still have to prove.

Assumptions hide in prose. In a flowing story, the sentence that recruiters are happy to pay €99 a month reads like a fact. On a canvas, in a block with an ASSUMPTION label in front of it, it suddenly becomes what it actually is: a guess you can test for a hundred euros of ad budget before you spend a year building. The form does the work here. What sits short and side by side, you can question; what is written out, you believe.

The order does the rest. It fills in the canvas in a fixed order, and that does not start with your solution but with the segment and the problem. Anyone who first has to write down three problems in the customer's own words, alongside the existing alternatives, sometimes discovers that the real alternative is doing nothing, and that is a different conversation from a competitor comparison. Only after that is the solution allowed onto the canvas, with each feature referencing the problem number it addresses. A feature with no problem number is a hobby.

In diff form: how the same information changes as it moves from business plan prose to canvas discipline.

That is the whole philosophy of the method in three red lines and three green lines: the same business, but you can test the second version while you can only believe the first.

Next comes the method. First five input questions: the idea in one paragraph, the problem and for whom, the stage you are in (idea, first customers, scaling or pivot), what you have already tested or learned, and the intended revenue model, even if you are not sure of it yet. Whatever you do not know gets that block filled with an ASSUMPTION label and marked as needing validation.

Then the nine blocks in the fixed order from the figure above, each with its own notes: the segment split into early adopters and mainstream, the problem with existing alternatives alongside it, the UVP tested against the 5 second rule, the solution linked to problem numbers, channels made concrete instead of vague, revenue and costs with a lifetime value indication and a guess at acquisition costs, key metrics with one north star, and an honest unfair advantage.

After the canvas comes the part that gives the skill its name: always a short list of the three riskiest assumptions, one about the problem (does this segment really feel it this badly), one about the solution (does this actually solve the problem) and one about growth (can this channel and revenue model scale). Each assumption comes with a concrete one to two week experiment: a round of interviews, a smoke test, a prototype or a fake door test.

The style rules are short but directive: write in the second person, a maximum of one page per canvas, bullets instead of long paragraphs, bluntly honest about weak blocks instead of diplomatic, and assumptions always with an explicit ASSUMPTION prefix. The file closes with a source list of four titles: Running Lean by Ash Maurya, Business Model Generation by Alexander Osterwalder, The Lean Startup by Eric Ries and The Four Steps to the Epiphany by Steve Blank. More on that below.

The engine underneath it comes from Eric Ries. The method is strongly inspired by The Lean Startup, from 2011, with the core idea that you test assumptions before you make large investments. In that way of thinking, the canvas is not a plan but a collection of hypotheses, and the build measure learn loop decides which hypothesis goes first. That is why the skill never stops at a filled in canvas: the three riskiest assumptions with their experiments are the bridge from paper to practice.

The customer conversation comes from Blank. Steve Blank described the customer development process in The Four Steps to the Epiphany, from 2005: get out of the building, talk to real customers, and only believe what gets confirmed outside the door. The experiments the skill proposes, from rounds of interviews to fake door tests, stand in that tradition. And because customer interviews have their own trap, people are polite and say what you want to hear, there is a separate Mom Test Coach skill for those conversations.

And two blocks deserve their own magnifying glass. The Unique Value Proposition is one line on the canvas, but a whole model sits behind it: anyone who wants to work out the proposition seriously, with a customer profile, value map and fit analysis, uses the Value Proposition Canvas skill NL. And the Key Metrics block refers to the AARRR framework; anyone who wants to measure their funnel from acquisition to revenue finds it worked out in the AARRR Pirate Metrics Coach skill.

Even without installing the skill you can improve your plan with these principles: put your idea on one page, write down for each block whether it is knowledge or hope, and think up a test for your biggest guess that costs less than a week of building.

After that, using it is simple: describe your idea and ask for a canvas. First it asks its five input questions, and then it fills in the nine blocks in the fixed order. If you want to learn to write this kind of instruction file yourself for your own work, the structure of a good skill is explained in writing SKILL.md NL, and the wider foundation on working with AI is in the knowledge base.

There are also situations where a different tool makes more sense. For an established business with proven revenue and partners, the original Business Model Canvas is often a better fit, because the four replaced blocks are relevant there. If your question is not about the whole model but only about the proposition, why does a customer choose me, then you are better off with the Value Proposition Canvas, which zooms in on exactly that part. And do not expect market figures from the skill: it structures what you know and labels what you do not know, but researching the market remains your work. The experiments tell you how.

The most important limit is the same as with every canvas: filling it in is the start, not the result. A Lean Canvas that disappears into a drawer after being filled in has become a one page business plan. The three experiments are there to be carried out, and the canvas should then be updated with what you learned. Hence the name of the method too: iterating from plan A to a plan that works.

A sharp canvas tells you which channel to test first, but as soon as that channel is your website, a different craft begins: getting found by the people with exactly your problem block. The free SEO scan shows in a few seconds where your site stands. In the guide which strategy model when NL you can see how this model relates to the others. And if you want to talk further about what AI can do for your business beyond that, from separate skills to full automation, we simply do that in a conversation.