By December 2022, Shogo Koda was already writing a memo arguing for change to the Japan strategy he had spent most of the year negotiating.
Only three months earlier, his Dexterity team and Sumitomo Corp. had announced an exclusive partnership. Sumitomo would bring Dexterity’s AI robotics software into Japan, integrate it with hardware, and deploy the resulting systems for a Robotics-as-a-Service (RaaS) business. Sumitomo publicly targeted 1,500 robots by 2026 and planned a demonstration center outside Tokyo.
On paper, the division of labor looked clean. Dexterity would provide the intelligence. Sumitomo would turn it into a Japanese solution.
In practice, the product refused to respect the contract.
Small changes in configuration could pull Dexterity’s engineers back into the loop. Documentation was difficult to hand off because the product was evolving quickly. The supposedly independent distributor remained dependent on the startup for technical decisions.
Within months of announcing the distributor agreement, Koda was drafting a proposal for something much heavier. A joint venture.
The man in the middle
Sumitomo, one of Japan’s largest sogo shosha, or general trading companies, had already been on Dexterity’s cap table since 2020 through its Silicon Valley investment arm, Presidio Ventures. Sumitomo was interested in doing more than simply holding the investment and had begun discussing a Japanese business with Dexterity.
From California, however, the proposition was difficult to evaluate. Japan was still largely a black box to Dexterity. Having an investor say, in effect, “leave Japan to us” required a leap of faith the company was not yet ready to make.
So Dexterity decided to put its own person on the ground.
The mandate was deliberately skeptical. Is Sumitomo really the right partner? Is there a Plan B?
The person chosen to audit Sumitomo had spent roughly nine years inside it. Koda had worked on resource investments and joint ventures before eventually building the European business of a Japanese mobility startup. In January 2022, he joined as Dexterity’s first employee in Japan.
Choosing commitment
Koda began by mapping the alternatives.
The most obvious partners on paper were robot system integrators and material-handling companies, the conveyor and warehouse-automation firms that move goods through a facility. Dexterity’s core strength was software, while someone in Japan would have to integrate and deploy it in the field. Japan has over a thousand robot system integrators, most of them small. He reviewed the list from material-handling and automation giants like Daifuku, Murata Machinery, and IHI, then conglomerates like Hitachi and Panasonic, then rival trading houses. After walking trade show booths and digesting industry data, the output was a 30-page report on markets, segments, and channels.
The exercise exposed a tradeoff. Some candidates already had more of the integration capability Dexterity needed. But in 2022, “Physical AI” still felt early to many of them. They were interested, but not necessarily prepared to build a business around it.
Sumitomo was different. It did not yet have all the robotics-integration capability Dexterity wanted, but it had the institutional credibility and executive relationships of a major Japanese enterprise, and was willing to put resources behind building the missing pieces.
Dexterity tested that commitment aggressively. Exclusive rights would require upfront financial commitment, a concrete action plan and investment in infrastructure such as a Japanese demo environment. Dexterity, in return, would provide its software and technical support.
“Sumitomo may not have had the capability yet, but they had the motivation to build it,” Koda recalled.
By April, the two sides had a term sheet. The exclusive distribution partnership was announced that September.
Sumitomo eventually had about five people involved, most of them committed to the task full-time. Dexterity Japan grew from Koda alone to a small technical team that included an application engineer and a technical program manager. Sumitomo was supposed to lead much of the customer-facing work, integration and deployment.
The theory was that Sumitomo would become progressively more independent, but that part proved harder.
What could not be delegated
Dexterity had systems operating in American logistics environments, and the original assumption was that sufficiently similar applications could be replicated in Japan. Instead, even modest changes in the customer environment required Dexterity to get involved again, and the manuals and documentation struggled to keep pace with continuously evolving products.
The issue was not simply localization. The product itself was still too tightly coupled to engineering for integration, deployment and support to be cleanly pushed across a contractual boundary.
At the same time, potential customers were exposing a different constraint in this arrangement.
Warehouse automation is mission-critical equipment. Once installed, customers expect it to operate for years. Koda remembers prospective customers probing the arrangement accordingly. “Once this is running, we need it maintained for ten years. Sumitomo is just a distributor, and you’re a startup, right? American, right? Won’t you just disappear?”
For a simple software tool, that concern might be manageable. A robot embedded in the flow of a logistics facility raised the stakes.
By December, Koda had concluded that the two companies needed to appear and operate, more like one committed local business. He proposed a joint venture.
The idea did not immediately become reality. A JV was expensive organizationally, and Dexterity still needed a business large enough to justify the effort.
That business soon appeared.
The customer that justified the JV
Dexterity had been developing automated truck-loading technology with FedEx in the United States. In Japan, the team saw a natural counterpart in Sagawa Express, one of the country’s largest parcel-delivery companies.
Senior executives at Sumitomo already had relationships at the highest levels of Sagawa, helping secure an initial meeting. That opened a door that would otherwise have been out of reach for Dexterity.
The US startup treated the meeting as important enough to bring headquarters into the room. CEO Samir Menon flew to Japan, as did Wen Hsieh, then the Kleiner Perkins partner on Dexterity’s board. Koda remembers Hsieh making the case with a level of directness few Japanese executives would use themselves.
“We invested $8 million in Amazon, more than $10 million in Google, and $70 million in Dexterity. If you pass on Dexterity, you will regret it for the rest of your life,” Koda recalls Hsieh saying.
The discussions eventually led to a four-company project between Sagawa, SG Holdings, Sumitomo, and Dexterity. Announced in December 2023, the project represented roughly ¥600 million (~$4 million), of investment in developing and validating Dexterity’s truck-loading system for Sagawa. Koda believes few, if any, Japanese companies have ever paid that kind of sum for a proof of concept.
Just as important was what Sagawa had been told before signing. The JV did not yet legally exist, but Dexterity and Sumitomo made clear that they were discussing one.
For a customer wondering who would support an embedded robotics system years into the future, the structure itself became part of the sales proposition.
When the partnership became a company
For Sumitomo, the progression was relatively intuitive. A venture investment had led to an exclusive distribution partnership. The partnership had uncovered a major customer. A major customer could justify a deeper operating commitment.
Inside Dexterity, enthusiasm was more mixed. A joint venture meant governance, capital, and additional organizational complexity.
Koda tried to make the overall package attractive to both sides. The JV proposal was paired with an additional Sumitomo investment into Dexterity itself. He also argued that Dexterity did not need majority ownership of the new company because Dexterity would continue to earn its core economics through software licensing, while the local entity took on more of the labor-intensive integration, deployment, and service work.
The resulting Dexterity-SC Japan was established in June 2024, with 60% owned by Sumitomo and 40% by Dexterity. The difference became visible in the organization. The distributor model had relied heavily on Sumitomo staff working alongside a very small Dexterity Japan team. Under the JV, the local operation began adding people directly for deployment, robotics service, administration and operations rather than relying primarily on employees from the two parents. By the time Koda left Dexterity in early 2026, Dexterity Japan and the JV together had grown to more than 20 people.
Still, incorporation did not make the two companies one.
Sumitomo had expected the deeper partnership to bring broader access to Dexterity’s tools, documentation and know-how. Dexterity still had to protect information across a legally separate entity. The line between what could and could not be shared remained a matter of negotiation.
What the structure alone did not solve
There was another boundary that the ownership structure could not fix. The way the two companies communicated and made decisions.
Koda saw it most clearly in Dexterity’s relationship with Kawasaki Heavy Industries, whose robot arms Dexterity was already using when he joined. The relationship had deteriorated badly. Dexterity felt Kawasaki was slow to respond to technical problems. Kawasaki, meanwhile, did not fully understand the extent of Dexterity’s frustration.
Koda began by writing down Kawasaki’s grievances in detail and circulating them to the US engineering team. Then he pushed Dexterity’s engineers to document their own position.
It sounds mundane, but American teams are accustomed to debating issues live and reaching conclusions in the room, while Japanese counterparts are more likely to absorb a proposal, circulate it internally, and return after considering the implications. Silence in a meeting could easily be mistaken for acceptance.
Over time, Dexterity adapted and institutionalized the adjustment by structuring important Japan discussions across two days: meet in the morning, let them take it back, reconvene the next morning.
The Kawasaki relationship eventually moved from friction to product development. In 2025, the companies announced that Kawasaki had co-developed the custom eight-axis arms used in Dexterity’s Mech robot.
Lessons in local operating design
1. Match the entry structure to the product
A traditional distributor model assumes that something can be readily distributed. The product is sufficiently stable, documentation remains useful, integrations are repeatable, and support can progressively be delegated.
Fast-evolving physical AI may violate such assumptions. If every customer deployment repeatedly pulls product and engineering back into the conversation, the local model should acknowledge that dependency rather than pretending it will disappear. That might mean embedded engineers, a deeper local subsidiary, a JV, or simply retaining more direct responsibility.
2. Trade exclusivity for measurable commitment
Sumitomo was not selected simply because it was a famous Japanese company. Other candidates had more obvious technical capabilities.
What mattered was its willingness to commit before the opportunity was fully proven. Dexterity exchanged exclusivity for financial commitment, dedicated people, a demo environment, and a concrete action plan. If a startup is going to give one partner that much control over its market, it should be equally explicit about what it receives in return.
3. Do not mistake a JV for integration
Equity creates incentives, while locally staffed JV can create credibility and operating capacity. That being said, neither automatically gives the partner access to the parent company’s technology, people or internal information and those interfaces need to be designed separately.
For a fast-growing tech startup, that means being candid with a partner about what is actually ready, what remains on the roadmap, and where the product still depends on headquarters. A partner working from the public marketing story alone might sell something the product cannot yet deliver.
4. Use headquarters deliberately
Dexterity did not win Sagawa by handing the account to Sumitomo and disappearing. Menon and other senior leaders repeatedly came to Japan. Their presence signaled importance, and their more direct American negotiating style could sometimes say things the local team could not.
The local team’s job is to know when headquarters must show up, what only headquarters can credibly commit to, and how to translate its speed and directness into a decision process a Japanese organization can actually absorb.
・・・
For Dexterity and Sumitomo, the structure of the partnership kept changing because operating in Japan kept exposing where the previous boundary no longer matched the business.
By 2025, Dexterity had progressed to onsite operations at Sagawa’s X Frontier, Sagawa’s large logistics hub in Tokyo, while the repaired Kawasaki relationship had become part of Dexterity’s next generation of hardware.
Who we spoke to
Shogo Koda joined Dexterity as its first overseas employee and led the company’s Japan business through January 2026. He previously spent nine years at Sumitomo Corporation and launched WHILL’s European subsidiary in Amsterdam. He is now Head of Japan Business Development at RLWRLD, a robotics foundation-model startup.
