A Tale of Two Sizes
Giants are not what we think they are. The same qualities that appear to give them strength are often the sources of great weakness.
There is one strange detail in the David and Goliath story that nobody mentions in the Sunday-school version and that Malcolm Gladwell illustrated so well in his book. He had a shield-bearer who walked in front of him. Why does the biggest, strongest, most feared warrior in the Philistine army need someone leading him onto the field?
Because, almost certainly, he could barely see.
The condition is called acromegaly. A pituitary tumour pumps out growth hormone in adulthood, and you get the height, the brow ridge, the enlarged hands and jaw, the giant of the children’s-book illustration. You also get visual field defects, optic-nerve compression, slow reaction time, weakness in the limbs that look monstrous from a distance. Goliath was led onto the field like a man being led to his execution. He just didn’t know it.
David was a slinger, and slingers in the ancient Near East were the artillery of their day. Long range, high accuracy, faster than any sword. Goliath was a melee specialist showing up to a ranged fight, half-blind, holding someone’s hand.
Where Small Starts Winning
Apart from just making for a great visual metaphor, I think it’s a great analogy for how the battlefield of Business Strategy just changed. The factors at play ostensibly favour Goliath, but when you apply just a tad more scrutiny the probability of success overwhelmingly favours David. Apply this to the modern day and you see a rapidly evolving landscape, where things are already starting to break…
Let’s take some examples I’ve seen recently of the so-called “Lean AI Unicorns”:
Legal Services - Crosby
“Crosby [is] an entirely new type of law firm where a suite of AI agents and 30 lawyers collaborate to speed up reviews for commercial contracts like services agreements, data processing agreements and NDAs.” - Forbes, May 2026
The slinger version of Big Law: a small hybrid team of lawyers and AI agents reviewing contracts at a median turnaround of 58 minutes and flat per-document pricing, against the traditional firm’s weeks-long review at $500–$1,000 an hour. Cursor used Crosby for 2,000 contracts and cut review time by ~50%; Crosby has now done 13,000 contracts and is growing 400% since October 2025. Same deliverable (Big Law-quality MSAs, NDAs, DPAs, with malpractice insurance) in a fraction of the time and cost. Legal services used to mean floors of associates but Crosby’s small size means they don’t have to worry about the legal or reputational ramifications of adapting existing workforce through training or lay-offs.
Software - Lovable
I think most of you know this company though for those who don’t: The promise of Lovable is an entire website made with AI. In software, Lovable looks like a normal dev‑tools company from the outside - production apps, real customers – until you see that they got to roughly nine‑figure ARR with fewer than fifty people. The user still gets working software; the part that vanished is the stack of engineers, managers and process in between.
Customer Service - Decagon (and Sierra)
Decagon builds AI agents that handle support for Hertz, Duolingo, Notion and others - the same refunds, cancellations and order issues that used to require offshore teams. Bilt cut its support staff from hundreds to 65 by routing work through Decagon, which went from about $10M to well over $30M ARR in a year and is now valued around $4.5B. It’s a small company doing the work of a BPO floor – a direct threat to the Accenture, TCS, Infosys, Teleperformance - whose models revolve around renting out human seats.
When the Inputs Change, So Does the Game
Most of the large company operating models I’ve consulted on or witnessed as an investor are running the same disease (here I categorise large company as anything as 500+ employees). The thing that made their size a huge advantage is the thing that is will kill them… if they don’t rapidly change.
For roughly a hundred years, large companies grew through a physics where four advantages compounded into a moat. Call the package the Scale Stack. Talent: the people at the core of any great company. Distribution: the channels, the shelves, the salesforce. Capital: you could outspend any entrant for as long as it took. Operational learning curves: the longer you ran, the cheaper each unit got. The MBA syllabus from 1925 to 2015 is mostly footnotes on how to stack these four into something akin to Strategy.
Each one used to be a help at larger scale… and now each one is a potential hindrance:
Talent is now either an asset or a liability, and this is arguably the most important of the four. In an organisation built around the operating model AI actually requires – small teams with high autonomy, short loops, talent next to the model with many individual contributors – more headcount is genuinely an advantage. In an organisation built around layers, sign‑offs and politics, every additional person is a drag. As Brian Chesky puts it, every person you add levies a “communication tax” on everyone else. At a giant, that tax falls hardest on your best people: the ones whose output AI now multiplies, but whose calendars are owned by the org. The lean team can ship in a week what it would take the larger team two months of governance and approval cycles.
Distribution still works in regulated procurement and enterprise sales (the places where the buyer is risk-averse and the product touches money). Everywhere else, Conway’s Law has teeth. Your distribution apparatus cannot reach the audiences forming on Substack, Discord, or whatever surface emerged six months ago. The team that would reach them does not exist on your org chart, and the governance process to create that team takes longer than the audience does.
Capital still wins for GPUs at scale, regulatory licences, and two-sided market liquidity. Everywhere else, the rate-of-return discipline that funded your scale punishes the exact kind of low-confidence, high-variance, fast-cycle experimentation that AI now makes ten times cheaper. The slinger ships forty experiments and keeps two while you hold a quarterly review that demands a business case for the first one.
Operational learning curves are a liability because the institutional knowledge that compounded into the lowest unit cost has hardened into institutional inertia. Christensen named the strategic version of this twenty-five years ago and called it the innovator’s dilemma. The operating-model version is starker. The structure that perfected exploitation cannot, at the same time, run exploration. Companies have tried to fix it with ambidextrous orgs, Spotify squads, dual operating systems, founder mode… etc.
The Barbell, and What to Do About It
But this isn’t a doom story. Plenty of giants will fix this. The ones that do will not do it through another round of blanket lay-offs - cutting headcount at random doesn’t change the operating physics, it just makes a slow company a slightly cheaper slow company. The ones that survive will do it through serious organisational change: smaller teams, fewer layers, talent next to the model, decisions made where the work happens.
So if you’re reading this and feeling anxious, I’d gently suggest the lay-off headlines are the wrong thing to fixate on. They’re noisy, they’re hyped and ignoring context, and they’re mostly a lagging indicator. The real signal - and the real strategic threat to incumbents - is the slinger. The 30-person law firm. The 50-person software company. The customer-ops startup eating the BPO floor. Watching them tells you far more about where work is going than another quarterly restructuring memo.
And if you’re thinking about where to work next, the calculus has flipped. Joining a smaller, AI-native company used to mean trading stability for upside. Now you’re also trading inertia for agility. Lower cost base, same product, faster cycle time, fewer layers between you and the customer - that’s not a worse deal, that’s the deal that’s winning. The economics simply favour a company that can serve the same outcome at a lower cost, and the org design favours one that can turn on a sixpence when the next twist comes.
The honest summary is this: in the AI era, most companies will be smaller. Some giants will adapt and survive at the top of the barbell. Many slingers will thrive at the other end. The middle will keep shrinking. If you’re a leader, the work is to redesign your company before the market redesigns it for you. If you’re an operator, the work is to make sure you’re standing somewhere on the barbell that’s still load-bearing - ideally close enough to the model, the customer, and the decision that you’re part of the leverage, not part of the tax.
Goliath’s problem wasn’t that he was big. It was that he was big in a fight that no longer rewarded being big. The good news is, unlike Goliath, we get to see the slinger coming.
Perhaps the cycle opening now is not just the cycle of the “small”, but of those who can organize their own time.When timing becomes strategic, agility beats weight. The new advantage may be rhythm, not scale.
Every time I read your writing, I feel like my eyes open more and more!
On one hand, I find it so exciting to visualize this momentum boost toward decentralised innovation, with an influx of smaller studios and startups tackling industries and problems from different perspectives.
But on the other hand, I remember feeling totally overwhelmed trying to keep up with the flood of AI products that seemed to keep popping-up in the market all at the same time.
It'll be interesting to see the solutions we come up with for mapping out and navigating such a fast growing landscape. A picture of the moving stairs from Hogwarts just popped into my mind.