On May 22, 2025, five Toyota Group companies jointly announced a coordinated investment in artificial intelligence and software capability: Toyota Motor Corporation, DENSO Corporation, AISIN Corporation, Toyota Tsusho Corporation, and Woven by Toyota launched two linked initiatives. The first is known as the Toyota Software Academy, a training organization offering roughly 100 technical courses in AI, data science, cybersecurity, vehicle regulations, and related fields. The second is the Global AI Accelerator, or GAIA, which expands group investment in AI research, development, and implementation across 11 initial categories, including manufacturing, knowledge retention and transfer, robotics, and office productivity. Toyota framed the effort in its own terms, describing GAIA as rooted in the longstanding practice of jidoka (automation with a human touch).1
What I think is worth noticing about the announcement is that the two programs are not a technology purchase. There is no flagship system, no extensive vendor partnership at the center, and no single platform being rolled out. The chief unit of investment is the person. This aligns heavily with the Toyota concept that “making things is about making people.” However, even before this announcement, Toyota and Toyota Group companies were already making significant inroads with artificial intelligence.
The D Room: Toyota’s Bottom-Up Community Driving AI
Toyota has been growing AI capability from the shop floor up for years behind the scenes. Over a decade ago a retired colleague shared with me some advanced measurement and quality-control practices. This technology involved both machine learning and precision measuring devices. He explained in simple terms that in some cases precision machined components tolerances were 5X tighter and yet also experienced 5X fewer defects compared to when I previously worked in the company. One main ingredient was the sensory technology. Unfortunately much of that early work remained behind closed doors. We do have, however, a more recent well documented example published in the Toyota Times about a community called the D-ROOM.2
The story begins in 2018 at the Miyoshi Plant, in the Quality Control Division. Two members, Masaya Furutani and Manabu Okuyama, received a deliberately vague assignment from a supervisor: “There’s this thing called AI that looks promising. Why don’t you try something with it?” The two studied machine learning on their own and attended outside training. The main obstacle they hit was not the technology. It was the environment. Company PCs blocked programming setups, and software downloads required application and approval. Okuyama later recalled sending more than 100 emails just to obtain approvals.
In 2019 they went directly to the Miyoshi Plant general manager and secured part of a meeting room, with their own network line separate from company infrastructure. The arrangement gave them a sandbox that could neither endanger corporate IT nor be blocked by it. Their pilot involved three AI models deployed on the factory floor. An outsourcing quote for the work came in at over 100 million yen. Instead they built it in-house, using Raspberry Pi single-board computers, USB cameras, and inexpensive lenses. According to the article the project component cost was ultimately completed for under 10,000 yen.
There is an important but easy to overlook detail in the story. Furutani had spent more than 10 years in Production Engineering before moving to Quality Control, and that background is what made the project work. Okuyama put it plainly: “Anyone can build an AI inspection model on a PC if they put in the study. The real challenges come after that — figuring out how to integrate it into equipment that is already running on the factory floor.” Building the model was the learnable half. Integrating it into running production equipment took a decade of process knowledge that no course teaches and no outside hire brings in the door. This practice is in stark contrast to the emerging Western model of bringing in AI savvy Forward Deployed Engineers (FDEs) from technology companies and having them instruct you in how to implement AI on your work practices.
The experiment in one corner of one manufacturing plant spread. An open chat grew to roughly 500 members. When Toyota moved company-wide to Microsoft Teams during 2020, the community launched its own channels and widened its scope from AI to digital problems generally. Physical hubs followed, and according to the Toyota Times article, there are currently 18 on-site locations across plants including Miyoshi, Motomachi, and Tahara, each run by volunteers with no dedicated staff. Response speed became the community’s proof of value. One request for a development environment was answered in nine minutes. One post about a connection problem drew 79 replies.
One distinct part of the culture in the Teams channel is around the importance of learning by doing. For example a Teams channel exists for showing off failures, where the more interesting the failure, the more likes it receives. Furutani told one member who was afraid of breaking a borrowed PC: “You’re not a fully fledged member until you’ve broken three PCs.”
In 2022, D-ROOM was absorbed into Toyota’s Digital Transformation Promotion Division, which brought budget, equipment purchasing, and lending. The two founders did not vanish into anonymity. Okuyama is now group manager of Toyota’s AI Center of Excellence, responsible for AI implementation, optimization, and governance across the company. Furutani leads community-building work aimed at the wider Japanese auto industry. A bottom-up experiment that began with two Quality Control members and a borrowed meeting room now sits inside the formal structure of the company, and its founders run parts of that structure.
Inside One Supplier: The Denso Vision
Another useful insight about AI investment in Toyota comes from looking deeper inside Denso Corporation. In software, Denso has often led the Toyota Group. One little known example is the QR code, invented at Denso in 1994 to track parts in production, and now in daily use worldwide.3 Similarly, much of the software that guides Toyota engine systems is jointly designed with Denso in Product Development.
Denso laid out its software direction in a strategy briefing on July 12, 2024: The company plans to grow its software workforce to 18,000 people by fiscal 2030, roughly 1.5 times its 2023 level, and to expand its software business fourfold, to 800 billion yen, by fiscal 2035.4 A notable element of the plan is where those people come from. Alongside recruiting, Denso also runs explicit re-skilling programs that convert its own hardware engineers into software engineers. The company is not replacing its workforce to meet a new technical era. Instead it is converting the one it has.
The most distinctive piece of Denso’s system is a certification program called SOMRIE, introduced in 2022, three years before the larger Toyota Group academy was established. SOMRIE defines 18 capability types, among them data scientist, security specialist, and system architect, organized into four segments and graded on seven levels. Assessment is not an internal formality. External assessors, including university professors and IT-industry professionals, take part in the evaluations. Certification begins at level 4. As of late 2024, no engineer had yet been certified at level 6 or 7, which are defined as an industry-leading figure and someone who sets the direction of the industry. The top level of the ladder is deliberately above anyone currently on it.5
The certification further sits inside a larger structure Denso calls the Career Innovation Program, which combines recurrent training, a role assignment process that matches certified skills to open work, and a buddy system in which senior certified professionals coach engineers on the job. The practical effect is that a Denso engineer can see the full skill map of the company, locate themselves on it, and plot a career across software, hardware, and AI without leaving. Skills are visible, the path upward is defined, and the coaching to move along it is part of the system.
Two further facts help complete the development story. First, the Toyota Software Academy adopted SOMRIE for skill visualization across the group. The supplier’s internal system became the group standard, which says something about where software leadership in the group actually sits. Second, Denso’s ambition extends past its engineers. A separate digital talent certification, launched in fiscal 2025 with three tiers, has already put more than 10,000 employees through training, against a stated goal of a 100% digitally capable workforce by 2030.6
Toyota’s AI Development Path
There are some key similarities worth reflecting on across all three initiatives discussed in this article. A group-level program trains and funds people across five companies. A shop-floor community grew its own experts bottom-up and was eventually absorbed into the formal organization. A supplier built a certification program and career system so its existing engineers could become its software workforce. In each case the capability being built lives in the company’s own people, close to the work.
The Miyoshi experience gives some insight as to why this is important. The scarce expertise for Toyota was not adopting artificial intelligence. Two Quality Control members taught themselves the technology in their spare time. It was the integration of models into running equipment, which drew on 10 years of Production Engineering. Knowledge of that kind exists only inside the company, and the only way to apply it is to teach AI to the people who hold that knowledge.
Concluding Thoughts
For leaders working out how to build AI capability, the practical question is not whether to use outside expertise but how. Forward deployed engineers from vendors and subject matter experts from consulting firms bring real skill, and they can accelerate early projects considerably. The tradeoffs are cost, quality, and risk. As of mid-2026, leading AI laboratories post forward deployed engineering roles at base salaries of several hundred thousand dollars per year, before equity.7 When outside experts do the work while the organization watches, the capability leaves when the engagement ends. Company knowledge can also exit the first as well.
When outside experts do the work while the organization watches, the capability leaves when the engagement ends.
The Toyota Group examples in this article all point to an alternative pattern, and I think it is the leaner path in the long run. Investment goes to training and certification, and it buys time to experiment. Authority for AI work sits close to the process knowledge. Formal structure follows demonstrated success rather than preceding it. None of this is a new theory. It is how Toyota often builds capability.
Organizations pursuing AI face the same questions Toyota answered: who will hold the capability, and what is the company willing to invest in developing them. Readers interested in development-oriented fundamental workshops on this path can contact Tyson Heaton, LEI Executive Director of Lean Tech/AI.
Note: This article borrows from both Japanese as well as the English language sources cited. In certain cases, the Japanese original article provided more details than the English versions available on-line.
- “Five Toyota Group Companies to Accelerate Skill Development and Innovation in AI and Software,” Toyota Motor Corporation global newsroom, May 22, 2025.
English: https://global.toyota/en/newsroom/corporate/42805724.html
Japanese: https://global.toyota/jp/newsroom/corporate/42801307.html ↩︎ - Initiatives,” Toyota Times, April 3, 2026.
English: https://toyotatimes.jp/en/spotlights/1091.html
Japanese: https://toyotatimes.jp/spotlights/1091.html ↩︎ - “QR Code invented by DENSO,” DENSO Corporation.
English: https://www.denso.com/global/en/business/innovation/qrcode/ ↩︎ - “Quality and Quantity Will Decide Competitiveness in the SDV Era: Denso Targets 18,000 Software Personnel by Fiscal 2030,” Nikkei xTECH, July 2024.
Japanese: https://xtech.nikkei.com/atcl/nxt/news/24/01185/ ↩︎ - “SOMRIE Certification System,” DENSO Corporation recruiting site (in Japanese): https://careers.denso.com/graduate/growth/CIP_SOMRIE/. Independent coverage: “How Denso Built Its System for Developing and Evaluating Software Talent over 10 Years,” MONOist, May 2024.
English: https://www.denso.com/global/en/driven-base/career-life/kondo_2407/
Japanese: https://monoist.itmedia.co.jp/mn/articles/2405/23/news108.html ↩︎ - “Digital Technology Use and Talent Development at Denso,” Japan Institute for Labour Policy and Training (JILPT) forum report, January 2026.
Japanese: https://www.jil.go.jp/event/ro_forum/20260108/houkoku/04-corp_report1-denso.html ↩︎ - Anthropic job listing, “Manager, Forward Deployed Engineering,” posted base salary range $320,000–$400,000, accessed July 2026.
English: https://jobs.accel.com/companies/anthropic/jobs/79002358-manager-forward-deployed-engineering ↩︎
