The AI Empathy Map Coach skill for Claude

The problem this skill solves is uncomfortably familiar. Ask any business owner who their customer is and you get an answer that sounds watertight and is based on nothing at all. The campaign gets built, the landing page goes live, and the message doesn't land. Not because the work is bad, but because it is aimed at a customer who mainly exists inside someone's own head.

The empathy map forces you to pull that picture apart: what does the customer really see around them, what do they hear from colleagues and friends, what do they think and feel, what do they say and do, where is their pain and what are they really looking for. And with every answer, the question that matters most: how do you know that?

It is written for business owners, marketers, product people and sales, as the file itself says. Anyone building a campaign, landing page, product or pitch who wants to sharpen their target group first. It is part of the skill library NL we make freely available through our AI and automation service, with no account and no sales email afterwards.

One thing you should know up front: the skill does not do customer research for you. It structures what you know, flags what you don't know, and tells you where to go and look. A map full of assumptions is an honest starting point, not a final destination, and the skill says so itself: gaps are valuable, they are your research questions.

The first problem is that saying and knowing get mixed up. In almost every marketing meeting, assumptions get stated with the confidence of hard data. Our customer cares about sustainability. Our customer has no time. Ask where that comes from and you rarely get shown an interview or a dataset, usually just a feeling. The skill accepts a feeling like that, but it gets the label ASSUMPTION attached. That changes the discussion: it stops being about who is right and becomes about which assumption should be tested first.

The second problem is that customers themselves cannot be fully trusted either. Not out of bad faith, but because people report what they would like to do rather than what they actually do. That is why the Say and do quadrant in the canvas is a single box: the tension between those two is itself information. There is an explicit instruction for this: find the gap, flag it, because the difference between what customers say and what they do is gold.

The third problem is that the canvas is usually filled in as a box-ticking exercise. A team sits round the table for an afternoon, sticks sticky notes into six boxes and goes home satisfied. Everything on those sticky notes comes from the same heads that already misjudge the customer. That approach is explicitly refused: never fill in a template without questioning the user, the file states. It asks concrete questions per quadrant that let you discover for yourself where your own knowledge runs out.

Shown as a diff: the difference between how a target group is usually written up and how the skill does it, with lines taken from the example in the file.

Then comes the structure. Step zero is the mandatory starting block: who are we looking at, what do they need to do, and what is the decision moment. The file is blunt about it: without this block the empathy map is worthless. Then come the six quadrants, each with concrete sub-questions. See covers providers, adverts, competitors and website experiences. Hear covers colleagues, podcasts, authorities and the unwritten rules of the industry. Think and feel covers beliefs, doubts, dreams and what keeps someone awake at night. Say and do covers observable behaviour and the gap between words and actions. And as the closing pair, Pain and Gain: the obstacles and risks on one side, the outcomes they are really looking for on the other.

The method has five rules. First ask follow-up questions, never fill it in blindly. Use real language, written the way the customer would say it themselves. Give evidence for every claim: where does this come from, data, interview, observation or assumption, and mark assumptions explicitly as ASSUMPTION. Look for the gap between saying and doing, because that is where the gold is. And finish with insights: three to five concrete insights plus three action steps, for example a landing page hook, ad copy, a product improvement or a sales pitch.

The output has a fixed structure you can copy straight away: the header with who, goal and moment, then per quadrant the points with their source in square brackets, a separate line for the gap between saying and doing, and to finish the core insights and concrete actions. Two worked starting examples come with it: Mark, the self-employed painter from Tilburg who wants to improve his site, and Lisa, 38, HR manager at an SME with eighty staff who has to choose a new HR tool at the moment management is asking for cost savings and the current tool costs €30,000 a year.

The file closes with five things the skill never does, more on that below, and with a proper source credit: the Empathy Map Canvas by Dave Gray at XPLANE, original version 2009, revised version July 2017, described in his book Gamestorming and his article Updated Empathy Map Canvas. The canvas is open source, and the file notes that it is used worldwide by, among others, IBM, IDEO and Salesforce.

That order is not a detail. A team that jumps straight to sticking notes into the quadrants produces observations without direction: interesting material, no decision. By first fixing which decision or task is central, every quadrant becomes a filter. What the persona sees and hears only matters insofar as it touches that one decision. And by ending at pain and gain, the canvas forces a ranking: which frustration weighs heaviest, which outcome is really being sought. That is where the insights the skill closes with come from.

The canvas is open source and, according to the credit in the file, is used worldwide by, among others, IBM, IDEO and Salesforce. Anyone who wants to check the origin themselves: the skill points to Gray's book Gamestorming and his article Updated Empathy Map Canvas on Medium. We deliberately do not write more about the source here than the file itself does.

The empathy map rarely stands alone. In the design world it is the best-known tool of the empathise phase of design thinking, and anyone who wants to run through that whole process can turn to the Design Thinking Coach skill. Anyone who would rather reason from the job a customer wants done than from the person can take the Jobs to be Done Analyst skill: the same empathy, a different starting point. The two complement each other rather than replace each other.

After that, using it is simple: tell it who your customer is and what you are dealing with, and the skill starts with the starting block. New to skills? Read what Claude skills actually are NL first and the step-by-step installation guide NL. Want to write a skill yourself for your own way of working afterwards? The structure of a file like that is explained in writing a SKILL.md NL.

There are also situations where you are better off leaving it, or at least weighing it differently. If you have genuinely had zero customer contact, no conversation, no review, no data point, then almost every line of the map becomes an ASSUMPTION. It handles that correctly, and flags everything, but be aware of what you then have: not a picture of your customer, but a tidy list of research questions. That is worth something too, as long as you don't call it a customer picture. Valuable as a starting point, dangerous if you build a campaign on it directly. The honest advice then is: use the map to pick your three most important assumptions and test those first, for example with a few customer conversations.

The second limit is that an empathy map describes, it does not validate. A pain sitting in the quadrant does not mean someone will pay for the solution. Other tools from the same library exist for that question: the Value Proposition Canvas NL to hold your offer up against the customer's jobs, the Lean Canvas Strategist to test the whole business model on one sheet, and the Business Model Canvas Coach for the fully worked-out version of that. All three run on the customer knowledge the empathy map delivers.

And the last limit is the most important: the map is a snapshot. Customers change, markets shift, and last year's map describes last year's customer. Anyone who genuinely wants to steer by it repeats the exercise whenever something changes about the decision moment, and keeps the old versions to see which assumptions turned out to be wrong.

A sharp empathy map tells you what your customer is searching for and in which words. The next question is whether that customer finds you the moment they search. The free SEO scan shows in a few seconds where your site stands. And if you want to talk through what AI can do for your business beyond this, from individual skills to full automation, we are happy to just have that conversation.