Start Where the Corpus Is One Person Deep: Exhaust Density as the Qualifying Trait
Say you can compile someone's whole professional exhaust into a wiki that answers questions they didn't know to ask. Who do you sell it to first, and with what offer? The answer isn't an industry — it's a trait. And the loudest objection to it, “I don't want AI writing my documents,” turns out to be correct, and to be the pitch.
The short version
- The market for a compiled-archive product isn't segmented by industry. It's qualified by exhaust density — past work that is voluminous, structurally reusable, and currently irretrievable.
- The solo twenty-year consultant is the cleanest customer of all, and the winning entry product is read-only: it writes nothing. The objection “don't let AI touch my documents” is correct, and it's exactly why this sells.
- The three retrieval tiers — lookup, scope discovery, proactive volunteering — are an adoption ramp. You sell tier one; tiers two and three arrive free as the corpus fills in. Then climb: solo → boutique → corporate.
Here's a question I kept circling, because it's the one that decides whether a good idea becomes a business. Suppose you can take someone's entire professional exhaust — every proposal, every document, every scrap of client work from the last twenty years — and compile it into something they can talk to. Not a chatbot bolted onto a help centre. A wiki over their own life's work that can find the thing, connect the thing, and eventually volunteer the thing before they think to ask. Fine. Who do you sell that to first?
The instinct is to answer with an industry. Legal. Accounting. Architecture. Pick a vertical, tune the pitch, go. I think that instinct is wrong, and the reason it's wrong is the most useful thing in this piece. The buyer you want isn't defined by what they do. They're defined by what they've accumulated.
1. The qualifying trait isn't an industry — it's exhaust density
What makes this product valuable is the exhaust: twenty years of client work someone has already done and can no longer reach. That's not a property of a sector. Two accountants can sit at opposite ends of the spectrum, and a data consultant and a management consultant and a solo architect can sit at the same end. So stop segmenting by what they sell and start segmenting by the shape of their archive.
The qualifying trait isn't an industry — it's exhaust density. The segment is anyone whose past work is voluminous, structurally reusable, and currently irretrievable.
Three axes, and you can eyeball all three in a five-minute conversation. I want to be honest about what kind of instrument this is: it's a trait, not a score. I'm not going to hand you a number or a threshold, because a fake metric would be worse than none. This is a shape you learn to recognise, the way you learn to recognise a good prospect for anything.
| Axis | The question to ask | What a strong signal looks like |
|---|---|---|
| Volume | Is there a lot of it? | Decades of documents, proposals, reports, decks, emails — a career's worth, not a project's worth. |
| Structural reuse | Does the work rhyme? | Every engagement resembles a prior one. Templates, repeated shapes, the same clause and the same pricing move showing up across years. |
| Irretrievability | Is it currently trapped? | All that volume and reuse, and yet finding the right past piece today means remembering it existed and digging by hand. |
Irretrievability is the axis people skip, and it's the one that carries the value. Volume without reuse is just a big pile. Reuse without volume is a tidy folder you already navigate fine. It's the combination — a large, rhyming body of work you can't get back to — that turns retrieval from a convenience into a genuine unlock. The pain isn't that the work doesn't exist. It's that it exists and stays out of reach, so people rebuild what they've already built, over and over.
I'm deliberately not going to re-explain how the archive gets compiled — the read path versus the write path, why capturing everything was never the bottleneck and compiling it is. That's a whole argument of its own and I've made it elsewhere. Here I only care about the go-to-market consequence: exhaust density is the variable you qualify on.
2. The solo consultant is the cleanest customer of all
Once density is your filter, one buyer stands out so far ahead of the rest that it's almost unfair: the person who has been working for themselves for twenty years. Not because their exhaust is bigger — a corporate has more — but because it's cleaner. Four reasons, and they stack.
One brain made every decision, so the corpus is coherent — there's a single point of view running through two decades of work, which is exactly what makes it compile into something that hangs together. No employer owns the exhaust, so there's no IP boundary to negotiate. There's no political question about whose knowledge gets compiled, because it's all theirs. And the pain is felt personally, every single time they rebuild a proposal they know for a fact they've written three times before.
I'll admit the bias in the open: I am this customer. That's not a coincidence I'm hiding — it's why the product keeps demonstrating itself on me while I build it. A solo operator with a dense, coherent, personally-owned archive is the segment of one that the whole Marketplace of One logic points at. I won't rehearse that framework here; the short version is its own line: start where the corpus is one person deep.
3. The biggest objection is the pitch
Now the part that matters commercially, because it inverts the thing every AI seller has been trained to do. There's a very common objection from exactly the people you most want as customers. It goes: “I can't really think of anything in my business to automate with AI. And honestly I don't want AI writing my documents — that's my job. That's my differentiation.”
The standard move is to treat that as resistance and overcome it. Don't. It isn't ignorance. It's correct. The writing is their judgment — it's the part that shouldn't be delegated, the part that differentiates them, the whole reason someone pays them instead of someone else. And once you accept that the objection is right, the entry product designs itself, because it has to concede the point completely.
“I don't want AI writing my documents” isn't resistance to overcome. It's correct — and it's the pitch.
So the entry product writes nothing. It's read-only. Your filing cabinet learns to talk; you keep the pen. It answers “where's my template for this?” and “show me the version of this I did for a client in aged care” and “what did I say about pricing on that job in 2019?” — and it touches none of the making. It automates the one part of the job they hate (finding things) and leaves untouched the part they are (making things).
This is why it slips past the defences that kill most AI adoption. Nobody in history has felt replaced by their own archive becoming searchable. There's no identity threat in “where's my template.” There's no labor-hours business case, so there's no politically unstable story about saving hours by cutting people — and I'll leave the full argument about why labor-hours ROI is unstable to one side; it's a separate doctrine. There's simply nothing here to sabotage. You've accidentally described the only AI product an AI-skeptical knowledge worker will adopt without a fight, because it agrees with them about the one thing they were never going to give up.
4. The three retrieval tiers are the adoption ramp
Read-only lookup is the entry, not the ceiling. And the elegant part — the part that makes this a business rather than a feature — is that the same product deepens on its own. The three tiers of retrieval turn out to be an adoption ramp, where each tier earns the trust that the next one gets to spend.
| Tier | What it does | What it earns |
|---|---|---|
| 1. Lookup | “Show me my 2019 pricing template.” You know it exists; the system fetches it. | Value on day one, trivially demonstrable. This is what they buy. |
| 2. Scope discovery | “What have I done in aged care?” You have a question, not a filename. | Arrives free. Starts quietly changing how they scope new work. |
| 3. Proactive volunteering | Mid-conversation about a new client, it surfaces the old engagement that rhymes — unprompted. | The moment they stop calling it “search.” They never asked you to sell it. |
They buy tier one. It's concrete, it's safe, it pays for itself the first afternoon. Tier two then arrives for free, because it's the same corpus asked a slightly looser question, and it starts to shift how they think about a new job before they've consciously noticed. Then one day tier three fires: they're talking through a prospect, not querying anything, just thinking out loud — and the system volunteers the 2019 engagement that rhymes, with the pricing that worked and the clause that saved them. That's the moment the thing stops being a search box in their head and becomes something closer to a colleague.
I'm not going to make the case here for why that top tier is worth so much more than the other two, or how you'd put a number on it — that's a separate argument about retrieval and value that deserves its own room. The point for go-to-market is narrower and, I think, lovelier: you never have to sell tiers two and three. The corpus sells them for you, on schedule, as its edges fill in. You sell the filing cabinet. What they stay for is that their past selves start showing up to the meeting.
They came for the filing cabinet. They stay because their past selves showed up to the pitch meeting.
5. The corporate version is the same asset with a different heart
Everything above is the solo case. The corporate case is the same machine wired to a different emotion. For the solo, the wiki is self-extension — it's them, more of them, reachable. For a firm, it's survival of turnover.
Think about what a departure costs a firm today. Someone leaves and twenty years of context walks out the door with them. The documents stay behind, but they go mute — nobody left can interrogate them. The corpus outlives the author and stops answering. Compiled, it keeps talking: why did we price it that way, who's done this before, where's the deck that won. That's institutional memory turned into a balance-sheet asset — the kind of compounding thing a firm funds as capital rather than expenses as a tool. It's a board-level purchase, not an IT one.
But be clear-eyed: it's also the messier sale. The moment more than one person's exhaust goes into the corpus, you inherit provenance across authors, permissions, and the quiet politics of making one person's knowledge collectively addressable. All the frictions the solo doesn't have, the firm has in bulk. Which is exactly why you don't start there.
6. The ladder: prove on solos, climb to boutiques, let corporates arrive
Put it together and the sequence writes itself. It's a ladder, and the rungs are chosen for a reason.
The go-to-market ladder
- Solos first. Cheap to serve, coherent corpus, personal pain, no politics — and delighted ones become evangelists. This is where you prove the thing and where the density trait is at its purest. Named example: the twenty-year independent consultant rebuilding a proposal they've written three times before.
- Boutiques next. The natural wedge is the professional-services firm that already hand-cranks bespoke analysis for its clients — assembling, by hand, exactly the kind of compiled knowledge asset it has never built for itself. They own none of the compounding value they generate. That gap is the opening.
- Corporates last. Let them arrive once the category exists and the solo and boutique wins have made the case legible. Same asset, bigger cheque, and by then you're selling something a board already half-understands. Named example: the firm that loses twenty years of context every time a senior person resigns.
None of this is a new grand strategy. It's the plainest reading of a rule I keep coming back to: start where the corpus is one person deep. Prove it where the archive is cleanest and the buyer feels the pain in their own hands, then climb toward the harder, richer sales once the pattern is undeniable. The pilot for all of it is a solo operator with a dense, coherent, irretrievable archive — and if you want to know how that pilot is going, I'm the one running it.
Where this leaves your go-to-market
If you're building or selling anything that turns a body of past work into something people can query, take three moves from this. First, qualify on exhaust density, not industry: score the prospect on volume, structural reuse, and — the one everyone forgets — irretrievability. Second, lead with read-only. Don't fight the “keep AI away from my writing” objection; agree with it, and sell the archive that talks back without ever picking up the pen. Third, let the tiers do your selling: ship lookup, and let scope discovery and proactive volunteering earn their own trust as the corpus fills in.
The temptation everywhere else in AI is to point it at the work someone already does and make that work faster. That's the horse, and there's no terminal value in a faster horse. This is the other kind of move — the compounding asset that answers the question you didn't know to ask, and gets denser every time you feed it. You just have to sell it to the right person, with the right first offer, in the right order.
So find the dense, coherent, unreachable archive. Sell its owner a filing cabinet that talks. And then get out of the way while their own past shows up to do the rest of the selling.
Scott Farrell writes on AI strategy, knowledge architecture, and the economics of building compounding intellectual assets at LeverageAI. If you've ever rebuilt a proposal you knew you'd written three times before, your exhaust is dense enough — the next move is letting it talk.