By Ekansh Mittal
Earlier this year, I wrapped up a deeply rewarding chapter leading product and commercial growth at Spotta, a UK-born AI and IoT company focused on early pest detection to improve agricultural productivity. I’ve now moved on to Energy Research Labs, where I’m helping productise and generate traction for energy-saving architectures tailored to intensive indoor agriculture – new tools for old problems, one might say.
But before diving into what’s next, I want to take a moment to reflect on the last four years – what I’ve learned while scaling in one of the toughest sectors out there, agritech – and how those lessons might help others navigating the messy, mission-critical world of deeptech.
Why Agritech Is the Ultimate Stress Test for Innovation
Agritech doesn’t tolerate fluff. It’s a brutal environment where technical elegance is irrelevant unless it delivers measurable field impact. A “great” product that can’t survive dust, humidity, or operator neglect isn’t a product – it’s a liability.
At Spotta, we deployed AI-powered pest monitoring systems across 300,000 hectares of citrus and cotton farmland in Brazil, date farms in the Middle East, and pine forests in Europe. These systems targeted pests that threatened billion-dollar agroforestry industries – where infestation levels were surging and conventional methods like blanket spraying were increasingly ineffective and environmentally destructive.
We introduced low-energy, AI-based sensors trained to detect both visual and acoustic signatures of specific pests – pushing detection upstream before colonisation could begin. It worked. We saw 80% reduction in pesticide use, 35% operational cost savings, 20% drop in infestation rates within a single deployment cycle
But none of this success came from simply throwing technology at the problem. It came from something far less glamorous: disciplined use of the Lean Canvas.
Lean Canvas: A Compass, Not a Checkbox
I’m often surprised by how many deeptech companies skip this step – or treat it as a one-time slide deck filler. For us, it was the central operating document. Every assumption, every pilot, and every iteration was tested against it.
We identified:
- The real problem: not just pest infestation, but data latency – farmers were reacting weeks too late.
- The early adopter: large-scale citrus growers already losing millions to preventable yield loss.
- The key metric: time to detection and intervention – not the number of bugs trapped.
Each pilot wasn’t a demo – it was a validation checkpoint to test a critical box on the canvas.
We didn’t build dashboards until we confirmed that field managers could (and would) act on alerts. We didn’t invest in channel sales until we discovered that growers trusted agronomists far more than traditional sales reps. We didn’t pursue hardware leasing until we understood the cash flow cycles of Brazilian farms.
This is what I call customer-informed constraint design. Start with fewer features – but make each one frictionless to adopt. Layer in only what the field demands – not what the lab dreams of.
Translating Innovation Into Adoption
Deeptech, especially in sectors like agriculture or public infrastructure, lives or dies by early trust. You’re not just selling a product – you’re challenging a long-standing method. And the bar isn’t just performance – it’s reliability under real-world chaos.
That’s where so many high-potential technologies fall short. They over-index on core tech and underinvest in workflow integration. At Spotta, we succeeded not because our AI was better, but because we embedded it in ways that required zero behavioural change.
- Alerts came in formats teams were already used to.
- Devices were modular, rugged, and matched existing grove layouts.
- We didn’t ask the industry to adapt for us – we adapted to their environment.
It’s a principle I’m carrying forward at Energy Research Labs, where we’re now tackling pest detection in date palms, pine forests, and indoor vertical farms. The sectors differ, but the principle holds: build adoption before you scale ambition.
Scaling Deeptech Is About Precision, Not Velocity
We’re often told to “fail fast.” But in agritech and cleantech, failures are expensive. A misstep can ruin a crop cycle, destroy trust, and close doors that take years to reopen.
That’s why I’ve become a strong advocate for structured, iterative scaling.
- Start with high-touch deployments.
- Let early users shape the product.
- Turn them into evangelists.
- Then – and only then – scale.
It’s slower. It’s messier. But it’s what actually works.
Final Thoughts
Looking back, the biggest misconception I see in deeptech is the belief that impact comes from invention. It doesn’t. It comes from translation – the disciplined, often invisible work of making complex ideas usable, measurable, and reliable in the field.
That’s what I’ve learned from the agritech trenches. And that’s what I intend to keep doing – at Energy Research Labs, and beyond.
Build slowly. Validate obsessively. Scale precisely.
And always, always, carry a Lean Canvas.