Founder Field Notes: What a Robotics Founder Taught Me About Why Robots Don’t Scale (and What Actually Works)

I recently had a long, candid conversation with a robotics founder who’s deep in the trenches - not building flashy demos, but trying to get robots deployed in real industrial environments.

No press releases. No humanoid hype. Just real customers, real contracts, and real headaches.

The conversation stuck with me because it surfaced something I’ve seen over and over again in robotics, but rarely articulated clearly:

Robotics doesn’t fail because the tech doesn’t work.
It fails because deployment is hard, customers are risk-averse, and reality is messy.

Here are the biggest lessons I walked away with.

1. The Real Use Cases Are Weird (and That’s the Point)

One of the live opportunities they’re working on right now?

A robot that opens trailer doors.

At first glance, that sounds almost comically simple. But when you dig in, it’s exactly the kind of task robots should be doing:

• Thousands of trailers in operation
• Real worker safety risks (cargo falling, injuries)
• High repetition, low prestige, real liability

This wasn’t a moonshot idea. It was a very practical response to a real operational problem.

The lesson:
The best robotics use cases are often unsexy, niche, and overlooked - but painful enough that customers are willing to pay.

2. Demos Are No Longer the Bottleneck

A few years ago, building a convincing robotics demo could take months.

Now?
With modern learning-based approaches, this founder can spin up a task-specific demo in days, sometimes after teaching the system only a couple dozen repetitions.

That changes the game:

• Faster customer feedback
• More parallel exploration of use cases
• Lower cost of saying “let’s try this”

But it also shifts the bottleneck somewhere else.

Which leads to the real problem…

3. Robotics Is Not a Product Problem First. It’s a Sales and Trust Problem.

The hardest part of robotics isn’t the model architecture.
It’s getting permission to operate in the real world.

Industrial environments are dangerous, expensive, and unforgiving. Customers don’t let robots into their workflows unless they deeply trust the team deploying them.

That trust comes from:

• On-site maturity
• Safety awareness
• Operational credibility
• Willingness to work through ugly edge cases

Technology matters. But trust is the gatekeeper.

4. NRE and “Chunky” Revenue Is the Only Way In

This founder was blunt about something many robotics teams try to avoid:

Early customers don’t want clean SaaS pricing.

They want:

• Development contracts
• NRE
• Pilots
• Custom integrations
• Weird purchase / lease / hybrid deals

And honestly? That’s fine.

Trying to force Silicon Valley subscription logic onto industrial customers is one of the fastest ways to get laughed out of the room.

The lesson:
NRE isn’t a failure of business model maturity.
It’s how robotics earns the right to exist in production environments.

5. There Are Two Robotics Graveyards, Not One

Most people talk about the “prototype graveyard” - robots that never leave the lab.

But there’s a second, quieter one:

Robots that technically work in the field, but never scale.

Why?

• Technical issues never fully disappear
• Customers demand near-perfection
• Many organizations are culturally unwilling to tolerate ongoing iteration

In robotics, problems don’t vanish. They just become rarer.

And many customers aren’t okay with that.

6. Customer Selection Matters More Than Industry Selection

One of the most important insights from the conversation:

Robotics adoption depends less on industry and more on customer mentality.

The customers who succeed with robots tend to be:

• Growth-oriented
• Competitive
• Willing to invest ahead of certainty
• Looking for differentiation

They’re not always the biggest players.
But they’re the ones hungry enough to experiment.

Robotics doesn’t scale sector by sector.
It scales customer by customer.

7. Hardware Is Still the Hard Part (and the Differentiator)

There’s a lot of talk about software-defined robotics. But this founder was clear:

Industrial hardware is still brutally hard to get right.

Reach. Payload. Battery life. Reliability. Integration.

Many robots fail not because the AI is weak, but because:

• They can’t operate long enough
• They can’t reach far enough
• They aren’t robust enough for real environments

Once you have serious hardware, though, something interesting happens:

Applications start to multiply quickly.

8. Systems Integration Isn’t a Compromise. It’s a Strategy.

The emerging strategy here wasn’t “one robot, one task.”

It was:

• Build a modular, capable platform
• Reconfigure it per customer
• Keep reconfiguration costs low
• Let customer scale pay back customization

This is how industrial robotics has always scaled.

What’s changed is that modern ML makes reconfiguration much cheaper and faster than it used to be.

9. Data Becomes the Moat - But Only After Deployment

Everyone wants to talk about data moats and world models.

But the ordering matters:

Sales → Deployment → Data → Simulation → Scale

Without deployments, there is no proprietary data.
Without data, there is no long-term defensibility.

Robotics companies don’t win by having the best model on day one.
They win by being allowed into enough real environments to learn faster than everyone else.

Final Thought

What struck me most about this conversation wasn’t a specific technology insight.

It was the realism.

No illusions about clean revenue.
No fantasy of instant scale.
No belief that one perfect product solves everything.

Just a clear-eyed view of how robots actually make their way into the world:

Messy contracts.
Slow trust-building.
Customer-specific pain.
Relentless iteration.

It’s not glamorous.
But it’s how robots ship.

And the teams that internalize this early are the ones most likely to still be standing when the hype cycles move on.

Written by Bogdan Cristei and Manus AI