Written by Ivo Remmelg, Venture Scout at Superhero Capital, hardware founder and angel investor.

 

A few years ago, turning an idea into working software took a team and several months. Today, one capable person with AI tools can often ship a working version over a weekend.

I see this from two sides: as a founder building a hardware company, and as an angel investor looking at startups every week. And one thought keeps coming back to me:

If you built it in a week, why can’t your customer build it in a week too?

This cuts both ways. Getting started has never been easier. You don’t need a big team or much money to prove an idea works. But standing out has never been harder, because everyone has the same tools.

So where is the real value now? I think the answer is less obvious than “just focus on sales.”

 

1. A working demo is not a working business

Think of building a house. With AI, you can put up the walls and roof in a weekend, and from the street it looks finished. But nobody can live in it yet: there’s no plumbing, no wiring, no permit, and no one to call when a pipe bursts.

Products are the same. The weekend version proves the idea works under normal conditions. The real product has to handle unusual cases, stay secure, connect to the customer’s other systems, get updates, and keep working for years, with someone accountable when it fails.

AI shrinks the first 20% of the work dramatically, and the remaining 80% much less. And that remaining 80% is where the value sits, precisely because it’s still hard.

So when a startup tells me “we built this in two weeks,” my follow-up question is now: “And how long until a paying customer relies on it every day without calling you?”

 

2. Customers were never really blocked by build cost

“Why can’t the customer build it themselves?” assumes that building cost was what stopped them. It rarely was. Companies could always hire an agency to build almost anything. They usually didn’t, because of:

• Attention. A hotel wants to run a hotel, not an IT department.
• Ownership. Whoever builds it has to maintain it for years.
• Accountability. When it breaks at 2 a.m., who fixes it? Who is liable?
• Experience. Building version one is easy. Knowing what goes wrong after 500 installations is not.

In practice, the customer’s real alternative is rarely “build it ourselves.” It’s “do nothing.” Your toughest competitor is often the status quo.

The concern is valid for thin products, though: simple internal tools, basic web apps, light wrappers around an AI model. Those really can be replaced by the customer or by a cheaper copy next month.

 

3. Sales matters, but sales alone isn’t the moat

My first instinct was that sales and go-to-market make the difference. That’s partly true. But sales skills can be hired and copied too.

What can’t be copied quickly are things that accumulate over time, like a snowball that only grows by rolling. No tool can create one in a weekend:

• Data nobody else has. For example, years of real-world performance data from hundreds of sites.
• Being built into the customer’s operations. Once systems are integrated and staff are trained, switching becomes painful, even to a cheaper alternative.
• Trust and references. When buyers are flooded with lookalike products, they fall back on one question: “Who else uses this, and did it work?” Trust becomes more valuable when building gets cheap.
• Certifications and approvals. Safety, fire, electrical, building codes. Slow, expensive, and a real barrier.
• The ability to carry the customer’s risk. Financing equipment, guaranteeing outcomes. That takes capital and confidence in your product.
• Owned distribution. Relationships with the installers, resellers, partners, engineers and consultants who actually decide what gets specified.

So the precise version is this: sales and go-to-market are the tools you use to build these snowballs. The snowballs are the moat.

 

4. Customers increasingly pay for outcomes, not features

When features are cheap to build, they’re also cheap to copy. If your value is “we have feature X,” a competitor adds feature X next month.

A promised result is much harder to copy: lower energy bills, fewer complaints, no upfront investment. That’s one reason models like Hardware-as-a-Service make sense to me. The customer isn’t buying a device. They’re buying an outcome, with someone else carrying the risk. Promising that credibly takes field data, a track record and financial strength, which are all snowballs.

 

5. Hardware: faster, but the slow parts are still slow

This isn’t only about software. Hardware development has sped up enormously. Design, simulation, firmware and prototyping that took years can now take months, sometimes weeks.

But some steps still run on physical reality and other people’s schedules:

• Certification labs
• Supply chains and component availability
• Moving from 10 units to 10,000 reliable ones
• Learning how a device behaves after five winters in a real building
• Physical installation in buildings that are all slightly different
• Real-estate sales cycles that often take a year or more

AI accelerates design far more than it accelerates validation, certification and field experience. That’s good news for hardware companies with an installed base. Their moat is much harder to fake than in software.

 

6. What this changes when evaluating companies

As an investor, I now put less weight on:

• How fast the product was built
• How many features it has
• How impressive the demo looks

And more weight on:

• How many paying customers rely on it daily, and for how long
• What data they collect that nobody else has
• How painful it would be for a customer to switch away
• Whether reference customers would vouch for them
• Whether they own a distribution channel or rent someone else’s
• What certifications, approvals or capital requirements stand in a newcomer’s way

The one test I keep coming back toBuilding is no longer the bottleneck. A working product alone isn’t an advantage anymore. The lasting value is in what still takes years to accumulate: data, trust, integration, certifications, distribution, and the ability to carry risk for your customer.

So here’s the test I now apply to every company I look at, including my own:

If a well-funded competitor with the best AI tools started today, what could they still not replicate within 12 months?

Whatever the answer is, that’s your real value. If the answer is “nothing,” that’s worth knowing early.