Two Spruce Partners

The short version

Last week, I built a complete positioning and messaging framework for a client — a B2B company building an AI project management platform — in about five hours. A single long afternoon.

Done the usual way, that is a two-month engagement. A good positioning consultant charges $25,000 or more for it, and you wait weeks for the readout.

I want to be precise about that claim, because “AI did my positioning in an afternoon” is exactly the kind of line that deserves suspicion. The output was not a thinner, faster version of the real thing. It was as rigorous and grounded in evidence & established method as what I would have produced by hand in eight times the time.

Two things made that possible, and they are the whole point of this post. First, the real bottleneck in positioning was never the frameworks. It was reading all the evidence, and that cost just collapsed. Second, I did not just ask an AI to jump in and start drafting the positioning & messaging. Before any positioning work began, I had it research the strongest practitioners in the field and build the method on their published work. This was the key.

I also turned the whole process into a reusable Claude skill. It is free, it lives on an open skills marketplace, and by the end of this post you will know how it works and how to run it yourself.

Why it was fast: the bottleneck was never the frameworks

The frameworks for positioning have been public for years. You can read the best books on it in a weekend. The reason good positioning stays rare and expensive is not a shortage of method — it is the evidence the method has to run on.

Real positioning is supposed to start from what customers actually did and said: why they switched, what almost stopped them, the exact words they used. Collecting that means reading the sales calls, the win/loss notes, the onboarding threads in full or scheduling a month of research calls and coding them against a model.

That is the step that collapsed. My client has been using an AI meeting notetaker from the start, so has years of transcripts of sales calls with actual customers.

I had Claude read every recorded customer conversation in full going back six months, and code each one against a demand model: the frustrations that pushed people to look, the pull that attracted them, the anxieties that stalled them, the habits that held them in place. What used to be a month of listening and tagging became part of one afternoon — and it was more complete than the hand version, because nothing got skimmed to save time.

The frameworks did not get faster. The evidence they need got cheap. That is the real unlock, and it is why the result came out better rather than just quicker.

Why it’s good: I researched the experts before writing a word

Here is the move that made the difference, and it is the part I recommend most solidly.

Before I wrote a single line of positioning, I did not ask the AI what it thought about my client’s market. I asked it to go find out who the strongest positioning and messaging practitioners in the world are, read their published work, and lay out what each one actually prescribes. The method came before the work.

That reframes what the AI is for. Left to its own devices, a language model will happily produce confident, generic positioning — the beige, on-trend copy you have read a hundred times. That is the failure mode everyone worries about, and it is real. But the same model is extraordinary at a different job: reading a field’s best thinking, synthesizing it faithfully, and then executing it with discipline.

So the method I ended up with is not my opinion, and it is not the AI’s opinion. It is a composite of four named practitioners’ published methods, run in a deliberate order, with their own tests applied at each step. When the work flags a differentiator as weak, it is weak by April Dunford’s table-stakes test — not because a model felt that way. When it says a deal died in anxiety, that is coded against Bob Moesta’s forces, not guessed.

That is why the output is solid enough to build a company’s messaging on. The rigor is borrowed from people who have earned it, and the AI’s contribution is doing their work thoroughly and consistently, at a speed and cost that used to be out of reach. Standing on named shoulders is what separates this from AI slop.

The four thinkers I built on

I chose four practitioners by three rules: currently active or still clearly load-bearing, relevant to B2B, and — most important — covering each other’s blind spots rather than saying the same thing four times.

April Dunford — the positioning decision. Author of Obviously Awesome and Sales Pitch, and still the definitive B2B positioning practitioner. Two of her ideas do most of the work. Start from competitive alternatives — what a customer would actually do if you did not exist, not the rival your CEO likes to name. And treat differentiated value as the hard part: ask “so what?” over and over until you land on money, risk, or time. Most teams stop two rungs short.

Bob Moesta — the demand evidence. Jobs to be Done, in the Clay Christensen lineage, but with the practical instruments: the switch interview and the Four Forces. A switch only happens when push (frustration with today) plus pull (the appeal of the new) outweighs anxiety (what could go wrong) plus habit (inertia). This is the piece most teams skip, and skipping it is why deals die in “no decision.” Everyone over-invests in making the new thing attractive and leaves anxiety and habit untouched.

Anthony Pierri & Rob Kaminski (Fletch PMM) — the expression. The sharpest current thinkers on turning positioning into actual copy, anchored on the homepage. Their formula: for [segment] trying to [job], we are better than [alternative] because [differentiation]. They take aim at a specific failure — the belief that enterprise buyers want lofty, aspirational messaging. Plenty of well-known B2B homepages read that way, and you can finish reading them without learning what the company does. They also give you the competitor-swap test: put a rival’s logo on your headline; if it still reads as true, your copy describes the category, not you.

Andy Raskin — the narrative. His breakdown of a now-famous sales deck became the standard model for strategic narrative: name the undeniable shift in the world, show there will be winners and losers, tease the promised land, present your capabilities as the gifts that get people there, then prove it. Plus one rule that fixes a lot of bad decks — never tell buyers their current approach was foolish. They chose it. Say the shift is what made it stop working. The shift is responsible, not them.

Who I left out, on purpose. Ries and Trout are foundational but written for a mass-advertising era; Dunford supersedes the method. Geoffrey Moore is about adoption sequencing, not messaging. StoryBrand is too formulaic for enterprise and overlaps Raskin. And category design — inventing a new category and teaching the market to want it — I excluded deliberately, and turned that exclusion into one of the skill’s guardrails. Category creation needs capital, patience, and an evangelism budget. For most companies, choosing an existing category with real search demand beats inventing one.

Positioning is a decision. Messaging is what you say about it.

The single most useful idea in the whole method is about order, not content.

Positioning is a decision. Messaging is the expression of that decision. Narrative is the story you tell about the decision. In that order — and every stage should narrow the options for the next one.

Most messaging problems are really sequencing problems. Teams start with narrative because narrative is the fun part. But a story invented before the evidence has nothing to constrain it, so it can go anywhere — and a company that starts there tends to end up describing its product several different ways, because no underlying decision was ever written down to hold the line. Start from evidence and a decision, and the words finally have something to be accountable to.

That is also why the four thinkers are not four competing options. They are a pipeline, and each one closes the gap the previous leaves open. Dunford will not hand you words. Fletch will not tell you what to research. Moesta is a research lens, not a messaging output. Raskin gives you a story but not a target segment. Run in order, they cover one another.

The gates between those stages are what keep the work honest. That is the next piece — how the method becomes a skill.

How the method becomes a skill

A Claude skill is just a folder: a main SKILL.md file plus reference files. Claude reads the main file when your task matches, and pulls in a reference file only when that particular stage is active. That keeps the entire method available without stuffing all of it into the model’s attention at once.

positioning-and-messaging/
  SKILL.md                     the sequence, the gates, the guardrails
  references/
    framework-document.md      the deliverable spec
    moesta-demand.md           switch interviews, Four Forces, question bank
    dunford-positioning.md     components, table-stakes test, value ladder
    fletch-execution.md        the formula, homepage structure, specificity tests
    raskin-narrative.md        five elements, deck structure
    audit-mode.md              critiquing existing material

Three design choices are worth calling out, because they are what make it reliable rather than a prompt with good intentions.

Gates between stages. Each stage has an explicit condition for moving on. You cannot leave the decision stage until every differentiator has survived the table-stakes test. You cannot leave expression until someone outside the company can read the headline and correctly say what the product does. The gates are what stop the work from quietly turning into a writing exercise.

The sequence is enforced but not locked. If you want to jump straight to narrative, the skill tells you exactly what you are skipping and what risk that carries — and then it does what you asked. A method that refuses to bend gets abandoned. One that explains the cost of bending gets trusted.

Guardrails as named failure modes. The skill carries an explicit list of the ways positioning goes wrong: inventing a category, starting from narrative, drifting into internal jargon, ending up with several heroes, citing unsourced statistics, letting value claims stop at features, positioning against the wrong competitor, over-serving pull. Each one is checked at every stage. “Don’t invent a category” is a far more useful instruction to an AI than “think carefully about your category,” because it names the specific mistake to watch for.

What it produces: a document you can actually use

An early version of the skill produced good thinking and no usable artifact — a common way for positioning work to fail. The fix was to specify the output as carefully as the method. Good analysis that nobody can pick up and use is a lost project.

The deliverable is one document in three layers, ordered by how often each part gets read rather than by how foundational it is.

1. The decision — short and stable. The problem, the real competitive alternative, what only we can claim, who it is for, the category, and what we are explicitly not. This barely changes.
2. The expression — applied and living. The narrative, the copy for each surface, a proof library, a quote bank, objection handling, and the vocabulary list. This is where the day-to-day work happens.
3. The evidence — reference. The demand research, the competitive scan, the sources, and a change log.

In front of all three sits the one page: the one-liner, the category, the segment, the three claims, the real alternative, and the strongest proof. It is written last, and it is the part people actually memorize. It is also the section most frameworks are missing.

The real test of the artifact is simple: can a salesperson who missed the workshop open it and write an on-message email in five minutes? If not, the thinking never leaves the building.

Two craft details turned out to matter more than I expected, and the skill enforces both.

Every quote carries a clearance status, internal-only by default. Recording consent for a sales call is not permission to quote someone in your marketing. Treating those as the same thing is how companies end up implying relationships and endorsements they do not actually have.

And put the caveat inside the claim, not beside it. Sales will use your strongest number no matter what the footnote says. A claim written as “one customer reclaimed 25–30% of their team’s time” survives being repeated; “25–30% time saved” does not — the qualifier falls off the first time someone says it out loud. The skill writes numbers so the guardrail travels with them.

What a rigorous run reliably surfaces

Here is the part I did not expect: running a real evidence pass is uncomfortable, and the discomfort is the value. When you actually read all the calls and apply the tests, the method surfaces a handful of hard truths that anecdote-driven positioning never reaches. These showed up in my work, and they show up broadly — so I will state them as general principles, because that is what they are.

Losses cluster in anxiety and habit, not value. When you code why deals stall, the pattern is rarely that the buyer did not see the value. It is that they could not get sign-off to send data outside the company, or could not tell who owns the switch internally, or were quietly held by a homegrown tool they had already built. Moesta’s point lands hard here: if your messaging has no answer to “what do we have to give up, migrate, or relearn,” it loses to no-decision — no matter how attractive you make the new thing.

Differentiation decays, fastest in AI. Run Dunford’s table-stakes test against competitors’ live marketing — the actual current pages, not your memory of them — and claims that were sharp eighteen months ago are routinely now free, bundled features of the incumbent your buyer already pays for. Date your competitive scan, and re-run it before every campaign.

Buyers do not use your vocabulary. The clever internal terms a company coins for its own concepts almost never appear, unprompted, in how customers actually talk. People say ordinary things — chasing, out of date, falls through the cracks. Keep a two-column list: the words they use, and the words only you use. The second column is fine in a strategy doc and poison on a website, because it forces the reader to learn your language before they can understand your product.

Check what your proof actually measures. It is easy to prove adoption of a feature and quietly call it proof of the outcome you are positioning around. They are not the same claim, and a careful read catches the gap. Sometimes the honest answer is that you have an evidence gap — and saying so beats papering over it.

The durable differentiator is often architectural, and invisible on a feature grid. The thing that survives every test is frequently not on any feature list at all. It is a structural choice about how the product works, and it falls out of the value ladder, not the roadmap. Those are the differentiators worth building a message on, because they are the hardest for a competitor to copy by shipping one more feature.

What it didn’t do

If I stopped there this would read like an ad, so here is the honest boundary. The method is powerful, and it is not magic. Four things it did not do.

Sales calls are not switch interviews. I instructed Claude to read all recorded sales calls from the last six months. Recorded sales calls are vendor-led conversations where the customer is reacting to a pitch, not being interviewed neutrally about their decision. So Moesta’s push and pull are partly contaminated, anxiety is captured well, and habit is captured worst — because nobody on a sales call ever asks a prospect why they didn’t adopt something. What you get is a strong, evidence-grounded hypothesis set. Not validated research. The skill is explicit about the difference.

It made me retire two statistics. Including one this very process introduced from a secondary source before checking it. Industry-statistics roundups cite each other and often trace back to nothing; two aggregators reported different numbers for the same underlying report. The skill now forces live web verification for exactly this reason — and it still caught me once.

It surfaced tensions it could not resolve. In at least one case the evidence pointed one way and the business pointed the other, and no amount of method decides that for you. A good process makes the real decision legible. It does not make the decision for you.

If you want to build your own skill

The positioning skill is specific, but the way it is built generalizes to almost any professional method you want an AI to run reliably. Six things I would tell anyone encoding their own expertise this way.

Encode the sequence, not just the content. The most valuable thing in this skill is not the summary of four methods — you can get those from the books. It is the insistence on running them in a particular order, with a gate between each step.

Specify the deliverable as carefully as the method. Decide up front exactly what the finished artifact looks like, section by section. Good thinking with no usable output is the most common way these projects quietly fail.

Name the failure modes explicitly. “Don’t invent a category” beats “consider your category carefully” every time. An AI follows a named, specific rule far more reliably than a vague caution.

Force external verification. The skill requires live web search for the table-stakes test, because a model’s memory of what competitors claim is always stale. If a step depends on current facts, make it go and check.

Let the method bend. Build in the ability to say what is being skipped, and then do what the user asked. A rigid method gets abandoned the first time it is inconvenient.

Note where your sources disagree, rather than blending them. Fletch deliberately keeps Jobs-to-be-Done vocabulary out of their framework; my skill uses Moesta’s method for research and Fletch’s formula for expression, and says so out loud. Honest seams are more useful than a smooth blend that hides the tradeoffs.

The skill is yours

I have published the skill on an open skills marketplace, and it is free. [Note: as of 9/23/26, this still isn’t approved and available on this marketplace, so please reply here or on Linkedin if you’d like a copy of this skill]. Anyone can install it and run the same process I ran for my client. If you are working on positioning or messaging — a homepage that does not explain itself, a deck that describes the product five different ways, a launch you want to get right — this is a real starting point, not a toy.

A few honest expectations before you run it. You need actual evidence — recorded calls, win/loss notes, real customer language, or better yet, a dozen customer interviews — for the evidence stage to earn its keep. The output is a strong, well-grounded first draft of your positioning, not a finished decision handed down from on high; you still have to make the calls it surfaces. And the further reading below is not optional. The skill is a way to run these four thinkers’ methods; the thinkers themselves go far deeper than any summary can, and they are worth your time.

Get the skill: COMING SOON

The sequencing, the gates, and the deliverable spec are mine. The methods are theirs, and the credit belongs to them:

• April Dunford — Obviously Awesome and Sales Pitch, plus the Positioning podcast. The positioning decision and the table-stakes test.
• Bob Moesta — Demand-Side Sales 101 (with Greg Engle); the Four Forces developed with Chris Spiek. The demand evidence and switch interviews.
• Anthony Pierri & Rob Kaminski — Fletch PMM. Homepage-level expression and the competitor-swap test.
• Andy Raskin — “The Greatest Sales Deck I’ve Ever Seen” and his writing since. The strategic narrative.

If you build something with it, or improve on it, I’d love to hear about it.