ProBackend
robotics
1 hour ago6 min read

Why the Best Robotics Play Isn't Building Robots: Destro AI and the Orchestration Gap

Destro AI raised $8M on a thesis most founders would reject: stop trying to build the perfect robot and start orchestrating the messy, human-filled warehouses where those robots actually have to work.

A Contrarian Bet on Software Over Metal

Most robotics startups begin with a question that sounds smart in a pitch deck: "What can this machine do that a person can't?" Destro AI started with the opposite one. "Who needs to move what, when, and why hasn't it moved yet?" The distinction matters more than it sounds at first glance, and it's the reason a $8 million seed round landed in the lap of a company that doesn't manufacture a single bolt.

Destro came out of stealth on September 30, 2026, with funding led by Base10 Partners and Bonfire Ventures, plus additional backing from CoFound Partners. The company's product is an AI intelligence layer called Mothership — an operating system that coordinates robots, carts, trucks, and human workers inside logistics warehouses. The founder, Manthan Pawar, earned a master's degree in robotics at NYU Tandon and spent roughly eight years inside the U.S. supply chain before starting Destro. He watched robotics companies build impressive machines that sat idle because nobody solved the surrounding chaos.

"One of the biggest reasons we are winning against robotics companies is because we are not a robotics company," Pawar told TechCrunch. He said plenty of competitors begin as engineers asking what cool things they could do. Destro asks what customers actually need done.

A Practical Example of Robotics in Artificial Intelligence

Destro's cross-dock deployment with Yusen Logistics is one of the clearest working examples of robotics in artificial intelligence that isn't a demo video. Here's why.

Cross-docking sounds straightforward. You unload goods from one truck and sort them into mixed loads for other trucks that make final delivery to customers. In practice, it's a paper-chasing, radio-call scramble across dozens of workers and vehicles arriving and departing in unpredictable sequences. Yusen Logistics — the Japanese shipping giant — runs roughly 30 U.S. facilities overseen by Richard Brunelle, its director of automation for the American logistics group. Most of those sites still rely on clipboards and walkie-talkies for the orchestration layer even though they've deployed fixed automation like conveyors and sorters.

Brunelle was already piloting new equipment — trailer-unloading robots, autonomous floor scrubbers, but the coordination between those machines and the humans around them was still manual. His U.K. subsidiary had built out a fully autonomous facility, and he wanted to replicate that approach in the U.S. When he met Pawar, Destro was focused on picking and pack (sorting goods into single packages). Brunelle recognized that his cross-dock problem was "fundamentally the same problem" and asked if they'd adapt. They said yes.

The pilot launched at a Yusen facility in the Pacific Northwest with three cart-moving robots built by Miva Robotics, controlled by Destro's Vision operating system, which is based on open-weight vision-language-action models that interpret camera images and instructions. The mothership layer above Vision coordinates everything: where each robot goes, when a cart needs to be staged, what the human worker should do next, and which truck dock gets priority.

"We make that operation less labor intensive, and we get away from the paper," Brunelle said. "Everything is now systematic."

The Competition That Couldn't Fit the Workflow

Yusen evaluated two other major robot startups before settling on Destro. One company offered robots that could move carts from point A to point B but had no solution for the loading and unloading steps that bookend every shift. The other brought fleet management software but still required a human to orchestrate the entire sequence, essentially automating the walking while leaving the thinking untouched.

Destro won because Mothership treats the human worker as a resource to schedule alongside the robots, not as a supervisor standing on the sidelines. The AI decides what the person does next as clearly as it decides where the cart rolls. That's the orchestration gap most hardware-first companies ignore, and it's the reason their pilots stall before they reach full deployment.

Destro is now expanding the initial pilot to a full rollout of 26 robots at that facility and has launched a second pilot with 17 robots at Yusen's Southern California site. Pawar's plan is to take this cross-dock workflow and "copy-paste" it across the thousands of warehouses that run similar operations. He's not overselling. He knows the customer problem cold enough to say the company is "on path to be cash-flow positive at the end of this year."

The Harness Thesis and Its Limits

Pawar describes Destro's position in the stack using a metaphor he's clearly thought about. "Robots are a platform. Every layer model is a platform. But we build a harness around it," he told TechCrunch. "That harness is just so complex and value-added that without that harness, these workflows are completely [impossible, right?]"

The harness, the orchestration logic, the human-machine coordination, the sequencing, is where Destro claims the durable value lives. Foundation models will get better. Hardware will get cheaper. But someone still has to make all of it cooperate inside a warehouse that's running at 3 a.m. with a skeleton crew and a truck schedule that changed twice.

That said, the thesis has an open question. Workflows that demand real dexterity and fine manipulation, things beyond pushing a cart from one dock to another, haven't been solved by generic robot bodies or open-source models yet. And there's the strategic risk that the companies building foundation models or ultra-dexterous hands decide to build their own orchestration layer rather than sell to Destro. Pawar's counter is that he's positioned at the layer where value accrues, and he'll ride the wave of research dollars pouring into better models underneath him.

Brunelle, for his part, remains a pragmatist. He hasn't found a good use case for a bipedal humanoid robot and says flatly: "We're not putting automation into a building just because the technology is interesting. It must solve a real operational problem."

What This Means for the Robotics Category

Destro's approach, a software-first, orchestration-layer play that explicitly rejects building hardware, is a useful signal for where the robotics category is heading. The companies that win at deployment aren't necessarily the ones with the best robot. They're the ones who make the whole messy system tick. Base10 Partners, which describes its thesis as "investing in automation for the real economy," seems to think this insight scales. The proof point won't come from a demo booth at TechCrunch Disrupt in October 2026, where Destro will show up among the Startup Battlefield competitors. It'll come from whether Pawar can actually copy-paste his workflow from one warehouse to the next without each site becoming a bespoke project.

The company's website still says "Launching Soon," which is either a modest understatement or a reminder that the product marketing hasn't caught up with the sales motion. Either way, $8M and a path to profitability by year's end speak louder than a landing page.

a contrarian bet on software over metal

More blogs