Deep tech startups — particularly in robotics and automation — face a unique set of challenges that distinguish them from software-native companies. The physics of hardware development, the complexity of regulatory and safety validation, the capital intensity of manufacturing, and the long sales cycles associated with enterprise industrial customers all create a crucible that tests founders in ways that building a SaaS product simply does not.

Over the past several years, we have had the privilege of partnering closely with founders across our portfolio at the earliest stages of their companies. We have watched them navigate fundraising, recruiting, customer discovery, and the inevitable engineering setbacks that come with building physical products. We wanted to capture some of the lessons we have observed — hard-won insights about what it actually takes to build a robotics company from scratch.

Lesson One: Customer Discovery Is Not Optional — It Is Existential

The most common mistake we see early-stage robotics founders make is spending too long in the lab before having meaningful conversations with prospective customers. The temptation is understandable. The technical challenges are absorbing, the founders are often world-class engineers with deep subject-matter expertise, and it feels productive to make progress on the hardware before "bothering" customers with an unfinished product.

The problem is that customer conversations at the earliest stage are not sales calls — they are discovery sessions. The information a founder gains from spending a week in a warehouse, a factory floor, or an agricultural operation talking to operations managers and frontline workers is worth months of lab development. You learn what problems actually cause pain, which workarounds people have invented that reveal latent demand, what the real decision-making process looks like, and who the internal champions for automation tend to be.

Several of our portfolio founders have told us that their first customer discovery conversations fundamentally changed the product they were building. In one case, a team building autonomous inspection robots for industrial facilities discovered through customer conversations that the highest-value application was not the general inspection use case they had been targeting but a very specific corrosion detection application in a particular class of petrochemical facility. Refocusing on that specific problem increased the willingness to pay by a factor of three and accelerated the time to first commercial deployment by more than a year.

The lesson: schedule customer visits before you think you are ready. The conversations will make you better, not just better at selling but better at building.

Lesson Two: Hire for Character Before Credentials

Building a deep tech startup at the seed stage is, by definition, an exercise in doing more with less. You cannot staff an early robotics company with a full team of specialized engineers across every domain — mechanical, electrical, software, computer vision, controls. You need people who can span multiple disciplines, learn quickly, and tolerate the ambiguity and occasional chaos of an early-stage company.

The founders in our portfolio who have built the strongest early teams consistently prioritize character attributes over pedigree. They look for people who are curious and self-directed, who have built things outside of work (side projects, open-source contributions, competition robots), and who have demonstrated the ability to figure things out when there is no established answer. A mechanical engineer who spent weekends building custom CNC machines and has a GitHub full of personal projects is often more valuable at the seed stage than a credentialed specialist from a large robotics company who has spent their career executing against clear specifications.

This does not mean credentials are irrelevant. For certain highly specialized roles — RF systems design, FDA-grade embedded software for medical devices, certain areas of controls theory — depth of expertise genuinely matters and cannot be substituted. But for the generalist engineering and operational roles that make up the majority of an early team, character and trajectory matter more than resume.

The other dimension of this lesson is the importance of founding team chemistry. Co-founders who have complementary skills, deep mutual trust, and a shared work style can move faster and recover from setbacks more effectively than technically stronger teams with fractious dynamics. We have seen technically impressive founding teams implode due to co-founder conflict, and we have seen technically modest teams achieve remarkable things because of their ability to work together through adversity. Chemistry is not sufficient, but it is necessary.

Lesson Three: Capital Efficiency Is a Virtue — But Hardware Has a Floor

The venture capital ecosystem has absorbed many lessons from software startups about capital efficiency, and there is a risk that those lessons get applied too bluntly to hardware companies. In software, the marginal cost of an additional user is near zero. In robotics, every prototype costs money to build, every sensor integration requires time and components, and every field test incurs real costs in travel, logistics, and hardware repair.

The founders in our portfolio who have navigated this best are those who have been disciplined about which things to build in-house versus procure, which technical investments create lasting competitive advantage versus which are commodity components that can be bought off the shelf, and which development milestones are truly necessary before a first customer pilot versus which are perfectionist tendencies masquerading as prudence.

Capital efficiency in robotics is not about spending as little as possible — it is about spending in the right places. A team that spends $500,000 on three high-fidelity customer pilots that generate real usage data and committed pilot-to-production conversion agreements has been far more capital efficient than a team that spends the same amount building a perfect prototype that has never been seen by a customer.

Related to this is the question of when to invest in manufacturing scale. The answer is almost always later than founders want. The drive to optimize and scale manufacturing is natural for engineering-minded founders, but investing in manufacturing infrastructure before you have validated that you are building the right product at the right price point is one of the most capital-destructive mistakes a hardware startup can make.

Lesson Four: Enterprise Sales Requires Dedicated Focus

Selling to enterprise industrial customers — large manufacturers, logistics operators, healthcare systems — is a distinct discipline that is genuinely difficult and deeply different from any other form of sales. The procurement cycles are long. The number of internal stakeholders who need to be aligned is large. The technical validation requirements are rigorous. And the risk tolerance of the operations leaders who own the decision is often very low — they are being asked to introduce unfamiliar technology into production environments where downtime costs real money.

Founders who try to manage enterprise sales relationships on top of product development and fundraising almost always underperform relative to those who hire a dedicated enterprise sales leader early. The timing of this hire is a judgment call, but the founders in our portfolio who brought in an experienced enterprise sales executive at or before their first pilot deployment consistently achieved faster progression from pilot to production than those who deferred the hire.

The profile of the right early enterprise sales leader for a robotics company is specific. They need enough technical literacy to speak credibly with engineers and operations managers. They need experience navigating multi-stakeholder procurement in industrial or healthcare settings. And they need the patience and strategic thinking to manage relationships that may take eighteen months from first conversation to signed contract. This profile is rare and commands a premium in the market — but the investment is almost always worth it.

Lesson Five: Expect the Hardware to Surprise You

Every hardware founder we have ever spoken to has a story about the hardware surprising them. A sensor that performed perfectly in the lab developed erratic behavior in the field due to electromagnetic interference from nearby industrial equipment. An actuator that was rated for 10 million cycles started exhibiting fatigue behavior at 2 million cycles under the specific load conditions of the customer environment. A battery that tested fine in a temperature-controlled lab failed in an outdoor agricultural application during summer heat.

The lesson is not that hardware is unpredictable — it is that you cannot fully predict hardware behavior until you have run it in the real environment for which it is designed. This argues strongly for getting hardware into field test conditions as early as possible, even with early-stage prototypes that are far from production-ready. The earlier you discover the surprises, the cheaper and less disruptive they are to address.

Building a culture of rigorous testing and transparent failure analysis into a company's DNA from the earliest stage is one of the most valuable things a founding team can do. Companies that treat hardware failures as learning opportunities to be analyzed and shared openly develop better products faster than those that treat failures as embarrassments to be minimized.

Key Takeaways

  • Customer discovery should begin before the product is ready — those conversations will change what you build for the better.
  • Hire for character, curiosity, and trajectory at the seed stage; credentials matter less than the ability to figure things out.
  • Capital efficiency in hardware is about spending in the right places, not minimizing spend — customer validation is always worth the investment.
  • Enterprise sales in industrial markets requires dedicated focus and the right sales leadership hire, made earlier than most founders expect.
  • Get hardware into real field conditions as early as possible — the surprises are cheaper to address before they become production issues.

Conclusion

Building a deep tech robotics company is hard. The physics, the capital requirements, the enterprise sales cycles, and the talent competition all make it harder than building most software businesses. But the founders who navigate these challenges successfully are building some of the most valuable and enduring companies in the world. We are committed to being their earliest partners and their most engaged supporters throughout that journey. Talk to us if you are building in this space, and check out our portfolio to see the kinds of companies we back.