The time of rented AI solutions is over, value is everything
A manifesto on what actually moves the needle when you put AI inside your business.
Axel Dekker
CEO
The time of rented AI solutions is over, value is everything
A manifesto on what actually moves the needle when you put AI inside your business.
Axel Dekker
CEO
The AI industry is having an awkward moment. The first wave of companies that slapped a chat interface on top of GPT and called it a product are running out of runway, their customers are doing the math, and their investors are finally asking the questions they should have asked eighteen months ago.
The pitch was "we're an AI company," but the reality is they're a thin layer of prompts paying margin to OpenAI on every call.
I've been here before, just not with AI. When we scaled Packaly, we hit the same fork in the road. You can stitch together other people's tools, ship something that looks like a product, and hope the platform fees stay friendly, or you can build the parts that actually compound. Most companies pick the first option because it feels faster, and most of them regret it within twelve months.
That's where the AI market is right now, and it's why I want to put down, plainly, what we believe at What's Next.
The wrapper problem isn't technical, it's economic
A wrapper is a product that sits on top of someone else's intelligence and adds a UI, a workflow, or a vertical-specific prompt. There's nothing wrong with that as a starting point, but there's a lot wrong with what it does to your business over time.
Three things compound badly when you're a wrapper. Your unit economics are someone else's pricing decision, so the day Anthropic or OpenAI raises token prices, your margin moves without your permission and you don't get a vote. Your defensibility is rented, because anyone with two engineers and a weekend can build the same UI on the same model, which isn't a moat so much as a parking spot. And your data flywheel doesn't exist, because every interaction your customer has is processed by a model you don't own, on infrastructure you don't control, generating context you can't use. The asset stays with the platform.
Now contrast that with a business that integrates AI deeply into its own operations and product. The interactions become proprietary data, the workflows become defensible processes, and the output quality improves because you're tuning the system to your specific customers rather than generic users. That's the difference between a feature and an asset.
What investors actually pay for
I spent enough years in shareholder rooms to know what gets a multiple and what gets a polite nod. Investors don't pay for AI, they pay for businesses where AI makes the underlying economics meaningfully better.
A logistics company that uses AI to cut routing costs by 18% is worth more because of that 18%, not because there's an LLM somewhere in the stack. A customer service operation that handles 70% of tier-one volume with AI agents is worth more because the cost-to-serve dropped, not because the deck has the word "agent" on slide three.
The companies getting marked up right now are the ones where AI is making something measurably faster, cheaper, or higher quality inside a real business. The companies getting marked down are the ones where AI was the business.
If you're an operator, that should change how you think about every AI initiative on your roadmap. The question isn't "what AI tools should we buy," it's "where in our business would owning an AI integration give us a structural advantage."
What we believe
This is the manifesto part, so take it or leave it.
We believe most off-the-shelf AI products are renting you capability you should own. The customer service bot, the sales copilot, the document analyzer, these aren't products so much as patterns, and the patterns are accessible to anyone willing to build them properly.
We believe the build-versus-buy question has been framed wrong for two years. The right question isn't "is it cheaper to buy or build," it's "does this capability need to compound for us, or is it commodity infrastructure we just want working." Email runs on commodity infrastructure and that's fine, but the AI that handles your customer relationships should not.
We believe every serious company with a real customer base has at least three places where AI integration would add genuine valuation, not because AI is magic, but because the company has proprietary context, proprietary data, and proprietary workflows that a generic tool will never reach. Those three things are exactly what makes integration valuable.
We believe the wrapper era is ending and the integration era is beginning. The next five years will be defined by companies who built AI into the bones of their business, and companies who paid SaaS fees to companies who did.
We believe operators are better positioned to win this than pure AI companies, because you already have the customers, the data, and the domain expertise. What you've been missing is the technical confidence to build rather than buy, and that's a fixable problem.
What this looks like in practice
I'll be specific because vague is useless.
When we work with a client, the first conversation is never "what AI tool do you need," it's "where in your business is the most expensive bottleneck, and what does ownership of that workflow look like in two years." Sometimes the answer is a customer service agent that lives inside their stack and learns from their tickets, sometimes it's a WhatsApp conversation system that handles thousands of customer touchpoints in a way no third-party tool can match, and sometimes it's a quote generator that knows their pricing logic better than their salespeople do.
The common thread is ownership. The data stays theirs, the workflow logic stays theirs, the competitive advantage stays theirs, and they get the AI capability while keeping the asset. That's a different model from what most of the industry is selling, and it's the one we've staked the company on.
Pick your side
If you're building a company in 2026, you're going to make this choice whether you make it consciously or not, and every quarter you defer the decision is a quarter you're choosing to rent.
I'm not going to tell you that buying AI tools is always wrong. Email clients and calendars are fine, and there's a whole class of commodity AI tooling you should absolutely buy because it doesn't need to compound for your specific business. But for the AI that touches your customers, your data, and your core operations, the right answer is almost always the harder one: build it, own it, and make it part of what you actually sell.
The companies that figure this out will own the next decade. The ones that don't will spend it explaining to their board why the SaaS line is going up faster than revenue.
We know which side of that we're building for.
The pitch was "we're an AI company," but the reality is they're a thin layer of prompts paying margin to OpenAI on every call.
I've been here before, just not with AI. When we scaled Packaly, we hit the same fork in the road. You can stitch together other people's tools, ship something that looks like a product, and hope the platform fees stay friendly, or you can build the parts that actually compound. Most companies pick the first option because it feels faster, and most of them regret it within twelve months.
That's where the AI market is right now, and it's why I want to put down, plainly, what we believe at What's Next.
The wrapper problem isn't technical, it's economic
A wrapper is a product that sits on top of someone else's intelligence and adds a UI, a workflow, or a vertical-specific prompt. There's nothing wrong with that as a starting point, but there's a lot wrong with what it does to your business over time.
Three things compound badly when you're a wrapper. Your unit economics are someone else's pricing decision, so the day Anthropic or OpenAI raises token prices, your margin moves without your permission and you don't get a vote. Your defensibility is rented, because anyone with two engineers and a weekend can build the same UI on the same model, which isn't a moat so much as a parking spot. And your data flywheel doesn't exist, because every interaction your customer has is processed by a model you don't own, on infrastructure you don't control, generating context you can't use. The asset stays with the platform.
Now contrast that with a business that integrates AI deeply into its own operations and product. The interactions become proprietary data, the workflows become defensible processes, and the output quality improves because you're tuning the system to your specific customers rather than generic users. That's the difference between a feature and an asset.
What investors actually pay for
I spent enough years in shareholder rooms to know what gets a multiple and what gets a polite nod. Investors don't pay for AI, they pay for businesses where AI makes the underlying economics meaningfully better.
A logistics company that uses AI to cut routing costs by 18% is worth more because of that 18%, not because there's an LLM somewhere in the stack. A customer service operation that handles 70% of tier-one volume with AI agents is worth more because the cost-to-serve dropped, not because the deck has the word "agent" on slide three.
The companies getting marked up right now are the ones where AI is making something measurably faster, cheaper, or higher quality inside a real business. The companies getting marked down are the ones where AI was the business.
If you're an operator, that should change how you think about every AI initiative on your roadmap. The question isn't "what AI tools should we buy," it's "where in our business would owning an AI integration give us a structural advantage."
What we believe
This is the manifesto part, so take it or leave it.
We believe most off-the-shelf AI products are renting you capability you should own. The customer service bot, the sales copilot, the document analyzer, these aren't products so much as patterns, and the patterns are accessible to anyone willing to build them properly.
We believe the build-versus-buy question has been framed wrong for two years. The right question isn't "is it cheaper to buy or build," it's "does this capability need to compound for us, or is it commodity infrastructure we just want working." Email runs on commodity infrastructure and that's fine, but the AI that handles your customer relationships should not.
We believe every serious company with a real customer base has at least three places where AI integration would add genuine valuation, not because AI is magic, but because the company has proprietary context, proprietary data, and proprietary workflows that a generic tool will never reach. Those three things are exactly what makes integration valuable.
We believe the wrapper era is ending and the integration era is beginning. The next five years will be defined by companies who built AI into the bones of their business, and companies who paid SaaS fees to companies who did.
We believe operators are better positioned to win this than pure AI companies, because you already have the customers, the data, and the domain expertise. What you've been missing is the technical confidence to build rather than buy, and that's a fixable problem.
What this looks like in practice
I'll be specific because vague is useless.
When we work with a client, the first conversation is never "what AI tool do you need," it's "where in your business is the most expensive bottleneck, and what does ownership of that workflow look like in two years." Sometimes the answer is a customer service agent that lives inside their stack and learns from their tickets, sometimes it's a WhatsApp conversation system that handles thousands of customer touchpoints in a way no third-party tool can match, and sometimes it's a quote generator that knows their pricing logic better than their salespeople do.
The common thread is ownership. The data stays theirs, the workflow logic stays theirs, the competitive advantage stays theirs, and they get the AI capability while keeping the asset. That's a different model from what most of the industry is selling, and it's the one we've staked the company on.
Pick your side
If you're building a company in 2026, you're going to make this choice whether you make it consciously or not, and every quarter you defer the decision is a quarter you're choosing to rent.
I'm not going to tell you that buying AI tools is always wrong. Email clients and calendars are fine, and there's a whole class of commodity AI tooling you should absolutely buy because it doesn't need to compound for your specific business. But for the AI that touches your customers, your data, and your core operations, the right answer is almost always the harder one: build it, own it, and make it part of what you actually sell.
The companies that figure this out will own the next decade. The ones that don't will spend it explaining to their board why the SaaS line is going up faster than revenue.
We know which side of that we're building for.
