Don’t Sell AI Tools. Become an AI-Native Company Instead.

A memo from the front lines of portfolio management — where even the e-commerce ops girl is vibe coding.

— -

We recently held a management meeting across several of our portfolio companies. The group was diverse: cross-border e-commerce, comic and manga AI agents, cross-border intelligent agents, and pure-play tech development. One moment from that meeting has stuck with me ever since.

The girl who handles operations for our cross-border e-commerce business — not an engineer, not a product manager — had quietly started writing code with the help of AI. Not just prompting. Building. She had created a series of custom agent skills that saved her hours every week. No CS degree. No bootcamp. Just curiosity and a chat interface.

That moment crystallized a question we’ve been wrestling with ever since: If AI is already empowering the average ops worker to build tools, what does that mean for companies whose entire business model is selling AI tools?

— -

The Half-Life Problem

Here’s the uncomfortable truth about building AI-powered products today: the shelf life of your competitive advantage has never been shorter.

Twelve months ago, building an AI writing assistant, an AI customer service bot, or an AI data extraction tool felt like meaningful product work. Teams spent six months, sometimes longer, fine-tuning models, building integrations, and shipping interfaces. Then a major lab dropped a model update — and suddenly the core capability was available out of the box.

This isn’t a hypothetical risk. It is the current operating reality.

The pattern repeats itself with predictable cruelty:

• A startup builds a polished AI coding assistant → GitHub Copilot or Cursor absorbs the use case.
• A team creates an AI-powered summarization tool → every LLM interface adds it natively.
• A company sells AI agents for customer support → the model providers build agent frameworks and give them away.

Big Tech has essentially unlimited distribution, unlimited compute, and deep research teams shipping continuously. If your moat is “we wrapped an LLM around this workflow,” that moat is a puddle.

The investment thesis for selling AI tools — i.e., creating software products whose primary value proposition is AI capability — is becoming increasingly hostile for small companies and startups. You’re not competing on a level playing field. You’re competing against the same companies that provide your infrastructure.

— -

The Real Opportunity: Using AI as Your Workforce

Let’s reframe the question. Instead of asking ”How can we sell AI to other businesses?”, ask: ”How can we use AI to outcompete incumbents in an existing industry?”

This is the AI-Native company model, and it is fundamentally different from the AI-tool company model.

An AI-Native company doesn’t sell AI. It absorbs AI — into operations, into delivery, into every layer of how the business functions — and then competes in a traditional market by offering better service at lower cost. AI is the engine, not the product.

Think about what this actually looks like in practice:

• A cross-border e-commerce operator that uses AI agents for product research, listing optimization, customer service, and logistics coordination can run at a fraction of the headcount of a traditional competitor. They’re not selling AI software. They’re selling goods — just more efficiently than anyone else in their category.

• A content studio that uses AI for ideation, drafting, image generation, and distribution can produce at five to ten times the output of a traditional agency. Their pitch isn’t “AI-powered content.” It’s simply: faster turnaround, lower price, same quality.

• A professional services firm (legal, accounting, consulting) that embeds AI into every step of its workflow can serve more clients, reduce errors, and price aggressively while maintaining healthy margins. The client doesn’t care about the AI. They care about the outcome.

In each case, the AI capability is a structural cost advantage — not a product feature. And unlike a product feature, a cost advantage compounds quietly over time.

— -

Why This Works Better Before AGI

We are in a specific and temporary window. AGI — broadly capable artificial general intelligence that can autonomously perform complex, open-ended tasks across domains — is coming. Most serious researchers and technologists believe it is a matter of years, not decades.

But we’re not there yet.

In the pre-AGI period, AI dramatically amplifies human capability without replacing the need for human judgment, relationships, domain expertise, and trust. The bottleneck is not raw intelligence. It is execution, context, and distribution — things that humans still provide and that incumbents have built up over years.

This creates a temporary but significant arbitrage opportunity for companies willing to restructure themselves around AI as a core operational layer.

The window looks something like this:

Now → Near-AGI: Companies that internalize AI as workforce can dramatically undercut incumbents on price while maintaining quality. The incumbents are too slow, too risk-averse, or too organizationally complex to restructure fast enough. Small AI-Native teams can do what previously required headcounts an order of magnitude larger.

Near AGI → AGI: The competitive advantages begin to compress as AI tools become even more widely available and incumbents finally adapt. The companies that moved early will have accumulated data, customer relationships, and operational muscle memory that late movers won’t easily replicate.

Post-AGI: A genuinely different world. Hard to predict precisely. But the companies that navigated the transition will be better positioned than those that didn’t.

The opportunity is not infinite. It exists because large incumbents are slow. Exploit it now.

— -

The Market Dynamics Are Merciless on AI Tool Sellers

Let me be direct about why selling AI tools is the harder path.

The buyer is getting smarter. Two years ago, enterprises were excited to buy any “AI-powered” product. Today, procurement teams ask hard questions. They want ROI evidence. They’ve been burned by AI tools that underdelivered. And they’re increasingly aware that the underlying models are commodities.

The margins are being squeezed from both ends. Computing costs have dropped dramatically, which sounds good — but it also means the barrier to entry for competitors drops with it. When it’s cheap to build AI products, it’s cheap for everyone to build AI products.

You’re in a features war you cannot win. If your product’s value comes from AI capability, then every time the underlying model improves, you have to sprint to incorporate it and maintain your edge. Your roadmap is hostage to OpenAI’s, Anthropic’s, and Google’s release calendars. That’s an exhausting and precarious place to build a business.

Distribution is still king. The companies winning in enterprise AI software are overwhelmingly the ones that already had distribution — Salesforce, Microsoft, Google, ServiceNow. They didn’t build better AI. They plugged AI into relationships they already owned. Startups don’t have those relationships.

By contrast, if you’re an AI-Native operator competing in e-commerce, content production, logistics, or professional services, your competition is often legacy businesses that haven’t adapted. That’s a much more manageable fight.

The Uncomfortable Implication

This framework suggests that the best use of AI for most small businesses and startups is not to build products about AI. It’s to use AI to become dramatically better at whatever they were already doing.

The e-commerce company should become the most efficient e-commerce operator in their category, not sell e-commerce AI tools to other e-commerce companies.

The content studio should become the fastest, cheapest, highest-quality content producer in their niche, not sell content AI tools to other creators.

The logistics company should compress costs and improve reliability through AI, not pivot to selling logistics AI software.

In each case, the AI is the unfair advantage. But the business is still the business.

— -

Before AGI changes everything, there is a window — maybe two to four years, maybe more — where small, nimble, AI-Native companies can outcompete legacy incumbents in almost any industry.

The companies that will win that window are not the ones building the next AI writing tool or the next AI customer support bot. They are the ones quietly restructuring their operations around AI, serving their markets better and cheaper than anyone else, and building the relationships and data that will matter even after the window closes.

The ops girl who figured out vibe coding to solve her own problems understands this intuitively, even if she’d never frame it this way. She’s not selling AI. She’s using it to do her job better than anyone else.

That’s the playbook.

— -

These views are my own and reflect observations from working with early-stage companies navigating AI adoption.