SoftBank employees built millions of AI agents — and the real story is how they taught an org to build with AI, not just use it.
Hot take: SoftBank didn’t just ship an internal tool — they shipped an org-wide muscle for building AI agents that scale. SoftBank announced employees created over 2.5 million AI agents after a company-wide push to embed generative AI into everyday work, with huge participation and internal certification programs that pushed AI from novelty into habit[4].
What happened: The company combined learning programs, tooling, and incentives (including contests and certifications) to accelerate adoption, and reported strong internal survey results showing adoption and confidence gains[4]. For developers this matters because it shows how adoption is as much about developer experience, governance and culture as it is about model quality — SoftBank invested in scaffolding, not just models[4].
Practical implications / opinion: If you’re a dev building internal AI platforms, this is a playbook: provide low-friction agent templates, make learning bite-sized, create internal incentives, and bake governance into the lifecycle. The risk is obvious — millions of agents multiply governance and security surface area — but done right, this scales domain knowledge and productivity across teams[4].
Want to pilot something similar where you work? Start by shipping an internal agent template for one common workflow and measure time saved; the rest is culture and iteration.
Source: SoftBank News