Building Value
The new rules of company-building
For most of the last twenty years, building value in a company meant building the things you could point at on a slide. The product. The pipeline. The revenue. The retention curve. The growth rate. The logo wall. The forecast. Operators measured the company by the artefacts the company produced, and investors paid for those artefacts at multiples that depended on how predictable they looked.
That model isn’t dead, but it’s becoming dramatically less interesting. Every one of those artefacts is now under pressure from AI.
Product surfaces are easier to clone in a weekend than they ever were to build. Outbound pipelines can be generated, sequenced, and personalised by software that costs less than a junior SDR.
Revenue can be manufactured short-term by anyone with a credit card and a sharp positioning angle. Even growth rates, the metric the venture industry treated as gospel, are increasingly a function of how aggressively a company is willing to subsidise distribution rather than how strong the underlying product is.
It’s worth being honest about what venture-backed company-building actually is underneath the storytelling.
At its core, it’s a financial arbitrage. You raise capital at one valuation, you deploy that capital across a small number of legs, fund the product, fund the team, fund the distribution, and you try to find a strategic buyer or a public market that will pay a meaningfully higher valuation for what you’ve assembled.
Fund the product, fund the team, fund the distribution, exit at a higher price.
Startups are a lot more of a financial equation than most operators are willing to admit, and the venture industry has spent four decades optimising every step of that equation. The problem now is that the steps themselves are getting harder to defend. When the cost of building a credible product collapses to a weekend, the spread between what you paid for that capability and what an acquirer is willing to pay you for it shrinks.
When distribution can be bought, the spread on that leg compresses too. The arbitrage is still real, but the margin in each leg of it is thinning, and the financial logic of the venture model only works when at least some of those legs are still expensive enough to create a defensible spread.
If you’re a founder right now, the uncomfortable question is this:
When every visible thing about your company can be imitated, accelerated, or compressed by a competitor with access to the same models, what is actually compounding inside your business?
The answer almost nobody is paying attention to is the only one that matters. The thing that compounds is the organisation itself.
The know-how it accumulates. The identity it has built. The brand it has earned in the minds of customers, operators, and the market. The judgement it has concentrated. The taste it has developed. The shape of its decision-making. The character of the people inside it, and the standard they hold each other to when no one is watching.
Building value in the age of AI is no longer about the company’s outputs. It’s about the company’s substrate.
The collapse of the things we used to call moats
It’s worth being precise about why the traditional value drivers are softening, because the conversation in venture has been somewhat lazy about this.
Product moats are collapsing fastest. The cycle time from “interesting demo” to “competitive product” has gone from months to weeks. Anyone with a Claude or GPT subscription, a competent designer, and a clear specification can ship a credible v1 of most SaaS products in a fraction of the time it took the incumbents. The defensibility that used to come from “we built this faster than anyone else could” is genuinely vanishing for the categories where the underlying capability is becoming a commodity.
Pipeline moats are sort of collapsing too. Five years ago, a company with a thousand high-quality, well-qualified leads in its pipeline had something real. Today, the cost of generating that pipeline at near-equivalent quality is approaching zero. Anyone can scrape, enrich, segment, personalise, and sequence at scale.
What’s harder is converting, retaining, and expanding, and conversion increasingly depends on trust, brand, and the specific human relationships your team has built. None of those are downstream of automation.
Revenue moats are real but unevenly so. ARR is no longer a clean signal of company health. A company can manufacture short-term ARR through aggressive land-grab pricing, AI-generated demand creation, and burn-funded expansion, while the underlying retention story quietly rots. Investors who have only ever pattern-matched on revenue growth are about to learn something painful about the difference between revenue you earned and revenue you bought.
Growth rate moats are the most overrated of all. A 3x year-on-year growth rate sounds impressive until you realise that in many AI-adjacent categories, the underlying market is itself growing 10x year-on-year. You weren’t growing faster than the market. You were the market growing past you, and you happened to be standing in the way.
None of this means these metrics don’t matter. They do. They’re necessary. They’re how the business gets funded, how operators stay employed, how investors stay confident. But they’re no longer where the durable value lives. They’re outputs, not the source.
The source has moved. The source is now the organisation itself.
Know-how and identity as the new substrate
When the visible layers of company-building become easier to copy, the invisible layers become disproportionately valuable. The know-how a company has accumulated is one of those layers. Identity is the other.
Know-how isn’t documentation. It’s not a wiki page or a Notion handbook. It’s the compressed judgement of dozens of decisions a team has made together about what good looks like in their specific context. Why the second version of a feature worked when the first one didn’t. Why a particular customer segment churned and another didn’t. What a deal feels like when it’s about to slip. What a hire looks like in the second interview when they’re going to be a problem in month nine. None of that knowledge transfers cleanly. None of it is in a model’s training data. None of it can be acquired by spinning up a competitor with the same headcount and the same tech stack. It can only be built by a specific group of people, working on a specific problem, for long enough that the patterns become muscle memory.
Identity is what holds that know-how together. It’s the answer to the question “what kind of company are we?” that gets used, often unconsciously, every time a difficult decision needs to be made and the playbook doesn’t quite cover it. A company with a strong identity makes decisions faster, more consistently, and with less senior oversight than a company without one, because the identity is doing the work of telling everyone what to do when the rules run out.
Most companies dramatically underestimate how much of their performance, especially under pressure, is downstream of identity rather than strategy. A team that knows who it is can move at speed without losing coherence.
A team that doesn’t has to relitigate first principles every time the environment changes, and in the AI era, the environment is changing every quarter.
If you’re a founder, the question is no longer “what is our product strategy?” The question is “what kind of company are we becoming, and is the organisation we’re building actually capable of executing on the kind of company we say we are?” Most aren’t. Most have a stated strategy that lives in the deck and a real strategy that lives in how people actually behave on a Tuesday afternoon, and the gap between those two is where value either compounds or quietly leaks.
Brand is the rarest resource
The third layer of the substrate, alongside know-how and identity, is brand, and brand has quietly become one of the most undervalued and most defensible assets a company can build in the AI era.
The argument I keep coming back to is this. In a world where the cost of producing content, products, pitches, and noise has collapsed to roughly zero, the bottleneck is no longer production. It’s attention.
And the honest answer to “who is paying attention?” in most categories is, nobody is paying attention to anything that matters. Feeds are flooded. Inboxes are full. LinkedIn posts blur into each other. AI-generated essays compete with AI-generated comments under AI-generated headlines. Every category is louder than it has ever been, and the signal-to-noise ratio is collapsing in real time.
In that environment, brand is no longer a marketing function. Brand is the only mechanism by which a company earns sustained, voluntary attention from the specific audience it needs. Customers. Operators. Investors. Acquirers. Future hires. The entire population of decision-makers whose choices determine whether your company compounds or stalls. If those people don’t know who you are, don’t trust who you are, and don’t have a clear emotional shorthand for what you stand for, no amount of AI-augmented distribution will move them, because they have already learned to filter out everything that hasn’t earned their attention the hard way.
Brand isn’t your logo, your colours, your tone of voice on social, or the polish of your website. Those are surface artefacts that anyone can now generate in an afternoon.
Brand is the cumulative perception of your company in the minds of the people who matter, built over years through what you ship, who you hire, what you stand for publicly, what you refuse to do, and how you behave when no one is forcing you to behave well. It is genuinely difficult to fake, because faking it requires sustained behaviour over time, and sustained behaviour is the one thing AI cannot manufacture on your behalf.
The companies that are quietly winning this cycle are the ones whose brand has become an unfair advantage in every other part of the business. They close hires faster because the candidate already trusts them. They close customers faster because the buyer already wants to be associated with them. They close fundraising rounds faster because the investor already wants to be on the cap table. They close acquisitions on better terms because the strategic buyer already has a thesis about why they belong together. None of that shows up cleanly in the financials, and almost none of it is being modelled correctly by investors using last-cycle frameworks. But it is, in my experience, one of the highest-compounding assets in any modern business, and it is the asset that becomes most valuable precisely when everything else gets cheaper.
Brand, know-how, and identity together form the substrate of the company. The product, the pipeline, the revenue, the growth rate, those are the outputs that the substrate produces. Most operators have spent their careers optimising the outputs. The ones who win the next decade will be the ones who learn to invest deliberately in the substrate.
The talent war is unlike anything I’ve seen before
I’ve been operating in tech for long enough to remember several talent cycles. The post-2010 mobile era. The 2014-2017 enterprise SaaS surge. The 2020-2021 zero interest rate mode when anyone with a pulse and a Stripe account could raise a Series A.
None of them look anything like what’s happening right now.
What’s happening right now is that the gap between the top 1% of operators and the median is widening at a rate that is genuinely difficult to overstate. The reason is simple. AI doesn’t level the playing field.
It does the opposite.
AI empowers the most capable and driven people to be even more effective, by an enormous margin, while leaving median performance roughly where it was.
A great engineer with Claude Code is not 20% more productive than a great engineer without it. They’re three to five times more productive in raw output, and arguably ten times more productive in the highest-leverage tasks where taste and judgement get amplified by execution speed. A great salesperson with AI-augmented research, prep, and follow-up is running a different game than a median salesperson, even if the median salesperson is using the exact same tools. A great founder with AI-augmented thinking is operating at a tempo and breadth that previously required a team of analysts, and they’re doing it before breakfast.
The reason this matters is that the talent war is no longer about getting bodies in seats. The talent war is about getting the specific 5 to 50 people who can carry your company on their backs, because that’s roughly the number of people you actually need now to build something that would have required 200 to 500 people a decade ago. And those 5 to 50 people know exactly what they’re worth. They have offers from every other serious company in the market. They are being courted with comp packages, secondaries, scope promises, founder access, and structural incentives that didn’t exist three years ago.
If you’re a founder and you haven’t internalised this, you’re going to lose. Not in some abstract future. You’re going to lose this quarter, in deals you didn’t know you were in, for hires you thought you were going to close.
The implication is that hiring is no longer one operating function among many. It is the single highest-leverage activity a CEO does.
Not because the cliché says so, but because the structural reality of AI means that one exceptional hire is worth more than ten merely good ones, and the cost of getting it wrong, in dilution, in distraction, in cultural drift, in the opportunity cost of the great hire you didn’t make instead, is genuinely company-defining.
I would go further. I think the CEOs who win this cycle will be the ones who treat talent acquisition as the most important strategic process in their company, full stop. Not the most important HR process. The most important strategic process, ranking above product, ahead of fundraising, before partnerships. Because the kind of company you can build is downstream of the kind of people you can recruit and retain, and in this market, both halves of that sentence are getting harder.
Choosing the right people, on both sides of the table
The reframe I want to offer is that the talent question isn’t really about being chosen. It’s about choosing well, on both sides.
For founders, choosing well means understanding that hiring is a two-way commitment with a multi-year asymmetric payoff, and that the candidates worth committing to are the candidates who can absorb meaningful scope, exercise meaningful judgement, and compound their effectiveness over time inside your specific organisation. Most hiring processes screen for the wrong things. They screen for credentials, polish, fluency in the existing categories. They don’t screen for the specific shape of intelligence and ambition that fits the specific shape of the company being built. The result is companies full of impressive people who somehow can’t get anything done together.
The hiring decisions that matter most aren’t the ones at the top of the funnel. They’re the ones at the bottom, where the decision to extend an offer is a decision to give a specific human being meaningful authority, scope, and economic participation in your company’s future. If you’re not willing to do that, you’re not actually hiring. You’re staffing. The two are very different things, and the great people can tell which one is happening from the first conversation.
For operators choosing where to spend the next chapter of their career, choosing well means understanding that you’re committing years to a specific founder’s vision and a specific organisation’s shape, and that the recruiting process is unusually bad at revealing either one. Recruiting shows you the pitch. It shows you the mission, the talent density, the imagined future, the language the company uses about itself. It rarely shows you the real structure of power. It almost never shows you how the organisation actually behaves when it’s under strain, when your work becomes inconvenient, when you ask for something they didn’t want to give, when the founder’s belief in your potential needs to convert into title, authority, scope, economics, or resources.
The asymmetry of information runs against the candidate, hard, in every recruiting process. The company has done this hundreds of times. The candidate has done it five. The company knows exactly which emotional buttons to press. The candidate has only the most general framework for what to look for.
Both sides have an enormous amount riding on getting this right. The founder is choosing who they’re going to trust with the operating leverage of their company. The operator is choosing who they’re going to trust with the next chapter of their professional life. Both decisions are made in compressed timeframes, with imperfect information, against the backdrop of competing offers and fast-moving markets. Anyone who tells you this is a clean, rational process is lying to you or hasn’t been close enough to it.
What ambitious people actually want
If you want to choose well as a founder, you have to understand what the people you’re trying to recruit are actually optimising for, which is rarely what they say in the interview.
Ambitious people tend to value a small number of things intensely. They want to feel rare, seen, irreplaceable, the quiet refusal of being interchangeable that drives most high performers more than money does. They want to feel destined, like the work they’re doing is bending toward something inevitable rather than incremental. They want to feel they’re not missing out, that the room they’re in is the room where the compounding is actually happening. They want to feel they have something to prove, that the credentials and the polish and the validation of their first decade are about to be tested against something real. They want proximity to how the business runs, not because they’re status-obsessed but because they want their judgement to actually matter. And some of them, the rarest ones, want to sacrifice for something that means more than the paycheck, a mission sharp enough that the wrong people would refuse to work there.
Most candidates can’t articulate which of these they’re actually starving for. The strongest founders can read it off them in a thirty-minute conversation, and they’ve already built a company that delivers the specific emotional payoff that specific person needs. The weakest founders pitch the same thing to everyone, then wonder why they keep losing the candidates who matter to companies that seem, on paper, less impressive.
The cash piece is worth dwelling on. Cash closes people.
It rarely converts them. Anyone who has lost a great hire to a competitor for an extra fifty thousand a year has misdiagnosed what was happening in the conversation. The candidate wasn’t choosing the cash. They were choosing the company that made them feel something specific, and the cash was the proxy they could justify to themselves and their partner.
If your hiring process can’t deliver an emotional payoff sharper than “competitive comp and good culture,” you’re going to lose every contested deal in this market.
Organisational shape as a moat
Here’s where I want to push the thesis further than most of the discourse currently goes.
In this era of AI, team sizes will be smaller. Concentration of great talent will be higher. The capability to do more with less has officially arrived, and it has consequences that almost no one in venture has fully internalised yet.
For most of the SaaS era, the implicit assumption was that scaling a company meant scaling headcount. You raised more, you hired more, you built more, you sold more. The org chart was a leading indicator of seriousness. A Series B company with sixty people felt more legitimate than one with twenty, almost regardless of what the twenty were actually shipping.
That assumption is breaking. The companies I find most interesting right now are running half the headcount of their direct comparables and shipping at twice the velocity. The reason isn’t that they have better tooling. Everyone has the same tooling. The reason is that they made an early bet on concentration over distribution. They chose to hire fewer people, hold a higher bar, give those people more scope, more authority, more economics, and let them operate inside a tighter structure where the cost of an average hire would be visible to everyone immediately.
That bet is the new moat. Not the product. Not the category. The shape.
A company with twenty exceptional people, each operating at three to five times their pre-AI productivity, with clear scope and meaningful ownership, is a fundamentally different kind of organisation than a company with a hundred good people doing the same nominal work. They make decisions faster, because there are fewer people to align. They ship faster, because there are fewer handoffs. They retain better, because the people inside them know exactly how rare the company is and exactly how exposed they would be in a more diluted environment. They recruit better, because the talent density itself becomes the pitch. And they compound faster, because every additional hire either raises the bar or drops it, and at small numbers, drops are visible immediately.
The capability to do more with less isn’t just a cost story. It’s a culture story, a recruiting story, a velocity story, and a defensibility story all at once. The companies that will define the next decade are going to be smaller, sharper, more concentrated, and more selective than the companies that defined the last one. They will be built in shapes the previous era couldn’t have produced, because the previous era didn’t have the leverage that AI has now made available to the most capable operators.
The implication for founders is that the most important architectural decision you make this year is probably not your tech stack or your go-to-market motion. It’s the shape of the organisation you’re building.
How small can you stay while still being credible at scale?
How concentrated can you make the talent?
How much scope can you give each person before the model breaks?
How sharply can you define the kind of person who belongs and the kind of person who doesn’t?
How much of your operating leverage you’re willing to bet on a small number of exceptional people, instead of distributing it across a larger number of merely competent ones.
The shape question is genuinely hard, and it’s the question almost no one is asking explicitly. Most founders default to the old shape because the old shape is what investors recognise, what their advisors lived through, what their peer set is doing. But the founders who break out in this cycle are going to be the ones who realise that the old shape is the constraint, not the answer.
The opportunity
When I look at where the durable value is being built in the AI era, almost none of it is in the places the discourse is focused on. It’s not in the model layer, where the capital intensity is brutal and the differentiation window is months. It’s not in the application layer, where the cost of imitation is collapsing toward zero. It’s not in the categories everyone is naming and renaming on a weekly basis.
The durable value is being built in the institutions.
In the companies that have figured out how to attract a small number of exceptional people, give them a shape they couldn’t have inside a more conventional organisation, and let them compound their judgement together over years. That’s the part that doesn’t get easier when AI gets better. That’s the part that gets harder, because every other piece of company-building gets faster, and the institution underneath has to absorb more change, more pressure, more ambiguity than it ever had to before.
The opportunity is to ask what kind of company hasn’t been possible before, what kind of person has been waiting for that kind of company to exist, and what kind of shape would let those people finally do the work they were always capable of.
If you’re a founder, the work is to design that shape deliberately, to recruit into it with the seriousness the talent war now demands, and to convert your belief in the people you’ve chosen into structure they can actually feel in their day-to-day.
If you’re an operator, the work is to look past the pitch and the polish and the language, and to ask, with your eyes open, whether the company in front of you is actually built in a shape that will let you become the version of yourself you’re trying to become.
The companies that win this cycle will not be the ones with the best products, the fastest growth, or the most aggressive distribution. Those things will matter, but they will not be the moat. The moat will be the institution itself. The know-how. The identity. The brand. The judgement. The shape. The people. The standard those people hold each other to when no one is watching.
That’s where value is being built in the age of AI. Almost everything else is downstream.