Knowledge is dead, Intelligence wins: The startups that will outlearn you are already coming

The old startup playbook said knowledge was power. AI just blew that up.
Startups that keep selling "expertise" will get quietly wiped out.
The next great companies won't know more - they'll learn faster.

For decades, if you knew something others didn’t, you could build a business around it. Expertise was scarce. Customers had no choice but to trust those who knew best.

That world is gone.

Large language models have torn down the barriers. Knowledge is now infinite, immediate, and largely free. What once required years of experience, industry secrets, or personal connections is now available to anyone with a keyboard.

Across nearly all established industries - health, finance, logistics, real estate, professional services and many others - information asymmetry is collapsing. And when knowledge becomes abundant, it stops being valuable.

Top investors like Greylock’s Jerry Chen have long warned that traditional moats built on expertise would weaken. In his 2017 essay The new moats, and more recently in The new new moats (2023), Chen argues that AI has only accelerated this collapse - pushing defensibility even higher up the stack, from data ownership to insight generation.

Founders who build businesses around "we know more" are already falling behind - even if they don't realise it yet.

Knowledge is no longer a moat. It’s an open field.


The birth of Intelligence

If knowledge is being commoditised, the obvious question is: what replaces it?
The answer: intelligence - systems that learn, predict, and improve faster than any human expert could.

For startups that stay stuck in the old model - simply packaging and selling static expertise - the outlook is bleak. Pure services businesses, like traditional law firms, consultancies, and advisory businesses, face serious pressure.

But services businesses are not necessarily doomed. Those who adapt with serious intent can still win - and win big. By combining deep domain expertise with AI-driven learning systems, services businesses can supercharge their offerings, move faster than human experts alone ever could, and create new defensible positions.

Just as AI is becoming intrinsic to every product - reshaping how we search, shop, diagnose, and build - it will also become intrinsic to every service. In the future, no serious professional service will operate without embedded AI at its core - not as a feature, but as a foundation.

Examples where the new winners will emerge:

• LegalTech: Platforms that predict negotiation outcomes, not just draft templates.
• Healthcare: Systems that refine and personalise treatments based on live patient feedback.
• Fintech: Tools that guide financial decisions based on evolving customer and market behaviour, not static reports.

As Elad Gil notes, survival will belong to companies that don’t just automate tasks - but build systems that learn from every transaction, every interaction, every outcome.

The future belongs to those who stop trying to know more - and start trying to learn faster.

Why Systems of Intelligence are the new moats

If static knowledge is no longer defensible, dynamic intelligence is. Systems of Intelligence - platforms that gather data, learn from it, adapt in real-time, and embed themselves deeper into customer workflows - create moats that grow stronger with every interaction.

As Jerry Chen frames it in The new moats and updates in The new new moats, "the next defensible companies won’t just own data - they’ll know what to do with it, creating systems that continuously learn across multiple domains."

Systems of Intelligence create compounding advantage because:

• They learn faster: Every user, transaction, and datapoint makes the product smarter.
• They predict and personalise: Customers don't just use the product - the product understands and anticipates their needs.
• They lock in workflows: The system becomes embedded in the customer’s operations. Leaving it means losing a strategic advantage.

The compounding happens quietly at first. Then it becomes overwhelming.

And by the time it’s visible to competitors, it’s too late.

How to tell if you're already falling behind

Many founders pride themselves on the expertise they have accumulated. And rightly so.
But in the new landscape, pride alone won’t defend you. You need to honestly ask:

• Is your product or service getting smarter with every customer interaction?
• Are you automatically improving outcomes for your users?
• Can you predict better results tomorrow than you could yesterday - based on what you learned today?

If the answer is no, you're not building a moat. You're just renting time.

And someone - somewhere - is already building the system that will replace you.

How to start building your System of Intelligence

You don’t need to invent a new AI model to play this game. You need to start embedding learning into your offering now.

Here’s how:

1. Combine 'systems of record' into a 'system of intelligence'
Most businesses already sit on valuable "systems of record" - databases of customer activity, product usage, transactions, interactions.

But on their own, these are just archives. The goal is to combine multiple systems of record into a live, adaptive workflow that turns raw data into real-time insight.

Example: At Duet, we’re applying this thinking to venture fundraising. Instead of simply giving founders access to static investment market data, we're combining multiple live datasets like Pitchbook - daily-updated investment activity records, and market movements - into a new dynamic workflow.
From there, we layer in AI-driven analysis:

• Evaluating emerging fund theses and individual partner preferences.
• Profiling client companies at a much deeper operational and strategic level.
• Predicting investor-to-company fit based not just on sectors or stage, but on subtle behavioural signals and shifting priorities at the investor level.

This combination creates a true system of intelligence: A living, evolving map of fundraising opportunities tailored specifically for each startup.

It’s a capability that simply wasn’t possible before - and it improves with every deal closed, every pitch made, and every investor movement tracked.

2. Prioritise data loops over features
Anyone can copy a feature. But if your product improves every time someone uses it, that's much harder to replicate.
Ask yourself:

• What are we learning from every user action?
• How are we feeding that learning back into the experience - automatically?

Example: Take Figma, the collaborative design platform. Figma didn’t just build design features. It learned from how users collaborated in real time - which actions slowed teams down, which workflows boosted productivity - and fed that back into new product optimisations. Their learning loops turned user behaviour into product acceleration, which in turn drove viral growth across teams and companies.

Figma didn’t just add features. It got smarter with every click.

3. Design for workflow integration, not just usage
A good product gets used. A great product becomes unavoidable.
You want your product to move from being a useful tool to becoming a critical system inside daily operations.

Example: Salesforce became dominant not because it was a better database, but because it embedded itself so deeply into sales workflows - lead tracking, pipeline management, deal forecasting - that companies simply couldn’t operate without it.

More recently, companies like Rippling have extended this idea: By unifying HR, IT, and finance workflows into a single adaptive platform, Rippling makes switching away practically impossible - while learning continuously from usage patterns across departments.

Workflow integration isn’t about convenience. It’s about creating dependence - and defensibility.


Move faster, learn smarter - or get left behind

The startup game in 2025 isn’t about who knows the most. It’s about who learns the fastest - and compounds that learning every day.

Building a system of intelligence isn’t optional anymore. If your product or service isn’t getting smarter with every interaction, you're standing still. And standing still is the fastest way to lose.

You don’t have to invent the next AI breakthrough. You just need to stop thinking like an expert - and start thinking like a system.

The companies that win from here won’t talk endlessly about what they know. They’ll quietly, relentlessly, outlearn and out-adapt everyone else - until there’s no competition left.

Outlearn, or get outplayed.

Let's talk.