We built dodoAI because enterprise AI had a trust problem
Founded in Osaka in 2024. Three use cases. One architecture principle: your data never leaves your perimeter.
How dodoAI started
Hitoshi Murakami spent the better part of a decade working inside enterprise operations in Osaka — running procurement, managing vendor relationships, and dealing with the compliance overhead that comes with handling sensitive contract data at mid-size Japanese manufacturers. When AI automation tools started becoming viable in 2023, the question was obvious: could we use agents to handle the routine approval decisions that consumed disproportionate human time?
The answer was yes — but with a condition that most cloud AI vendors were not prepared to accept. The procurement data, vendor pricing, contract terms, and compliance thresholds these agents needed to read were exactly the data that Japanese enterprises are most protective of. Sending that material through shared cloud model infrastructure was a non-starter for every company in the conversations Murakami was having.
The answer was not to abandon AI adoption. The answer was to build an agent runtime that worked entirely inside the enterprise perimeter — reading your policies, using your systems, writing to your audit storage, and never establishing a connection outside your network boundary. dodoAI was started in Osaka in early 2024 to build exactly that. Three use cases — procurement approvals, compliance checks, vendor onboarding — chosen because they are the highest-volume, highest-sensitivity back-office workflows where data sovereignty is not a preference but a requirement.
Three people. One very specific problem.
Hitoshi Murakami
10 years in enterprise operations across manufacturing and logistics in Osaka. Built dodoAI after experiencing firsthand how cloud AI tools fail the data sovereignty requirements of Japanese enterprises.
Yuki Tanigawa
Former infrastructure engineer at a Tokyo-based financial software company. Designed dodoAI's air-gapped deployment architecture and the on-premises model inference pipeline.
Kenji Ohmori
Previously built enterprise procurement software at a Nagoya-based ERP vendor. Leads the rule mapping methodology and customer onboarding process at dodoAI.
Three principles we don't negotiate on
Sovereignty by default
Data sovereignty is not a premium feature or an add-on tier. It is the baseline. Every deployment we do guarantees that your data stays in your environment — this is not configurable away.
Explainability over magic
Enterprise AI that can't explain its decisions isn't deployable in regulated environments. Every agent decision references a specific rule from your policy documents. Auditability is not an afterthought.
Precision over breadth
Three use cases, done right. We did not build a general-purpose automation platform. We built three specific agents for three specific processes where we know the problem deeply enough to deliver reliably.
Talk to the team directly
We review every inquiry personally. Enterprise deployments start with a discovery call — no pitch deck, just a real conversation about your environment.