
What happened
Mizuho Financial Group built Mizuho Cowork, an autonomous AI agent platform, with AWS Professional Services support, and started a proof of concept for specialist desk work.
Why it matters
The platform runs on Mizuho-managed AWS and uses Claude Agent SDK, so the bank can add workflow rules and internal knowledge as Skills without giving the agent direct access to employee devices or local data.
What to watch
Whether the proof of concept shows the agent can handle tasks like audit checks and translating specialist reports hinges on how well its guardrails reflect internal rules; Mizuho plans continuous updates.
WHO IT HITSBank and financial-institution operations teams that handle specialized desk work such as internal auditing, document preparation, and report translation stand to see parts of that work shift to autonomous agents, based on the proof of concept now under way.
Summaries like this, in your inbox every morning.
Mizuho Financial Group has been building an internal base for AI agents before this announcement. It already set up Mizuho Wiz Base, a company-wide AI agent platform, and an Agent Factory that supports the mass production of agents for individual operations. Mizuho Cowork extends that effort by giving one high-performance general-purpose agent the ability to take on a wide range of tasks, using Skills that describe procedures, expertise, and internal rules.
The platform is split into two layers. A general agent layer uses Claude Agent SDK functions such as tool execution and context management to control file operations, information gathering, analysis, and deliverable creation. A Mizuho-specific layer embeds guardrails that comply with internal rules. Because the execution environment and work area sit on AWS managed by Mizuho, a series of tasks can be completed there without direct access to user devices or local data, with network and access scope kept under Mizuho's control.
The proof of concept will show whether an autonomous agent, which unlike conversational generative AI can operate files and tools on its own and run multiple steps in sequence, can be trusted with work that demands specialized knowledge and procedures. The outcome appears to hinge on how well the agent's guardrails and Skills match actual internal rules, and on whether the agent can ask the right person for confirmation or approval at the right moment as Mizuho intends to keep updating the platform.
Pick your industry and the AI tools you use, and get news related to your work every day.
Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →
Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.
Ghost AI raised $11 million, led by Andreessen Horowitz with Abstract, Audacious Ventures, Nova and SV Angel…
Moonshot AI closed its final private round at about a $50 billion valuation and is aiming for a Hong Kong IPO…

OpenAI told Fortune it is piloting a 'mission interview' for candidates on certain teams, starting with market…

Reactor said it added Nvidia's venture capital arm and Sapphire Ventures to its cap table, bringing its total…

OpenAI will roll out invisible text watermarks to all ChatGPT and Codex plan users in the EU within weeks, and…

Reflection announced Beam, a 501B-parameter open-weight model with 23B active parameters, claiming parity with…
