JPMorganChase is taking Claude Code beyond its engineers one stage at a time
For most of the history of corporate software, the line between the people who specify work and the people who build it was clear. Product managers wrote requirements, risk teams reviewed controls and engineers turned all of it into code. Coding agents blur that line, and one of the world's largest financial institutions is now testing how far it can safely move.
JPMorganChase is rolling Claude out across the firm to more employees beyond its engineers. Product teams are turning requirements directly into engineering backlogs. Other teams use Claude Code, the coding agent, for risk and controls assessments and for program planning. Infrastructure and operations teams use it for monitoring, alerting and root cause analysis, and the security team uses frontier models to find and fix vulnerabilities faster.
Small teams in asset and wealth management
The most interesting organisational change is in Asset & Wealth Management. There, the firm is putting the people with the deepest business insight alongside technology and product experts in small, empowered teams, with the aim of building and delivering customised client investment solutions at scale. Claude Code sits inside their engineering workflow, and every change goes through more rigorous testing before it reaches clients. In the Commercial and Investment Bank, work moves into production faster and the firm redeploys that capacity against business priorities. Engineers there are modernising legacy systems and building and sharing their own agents, including pull request review and self-healing agents.
Access that widens behind controls
What makes this a governance story is the set of constraints around it. Claude Code runs in controlled environments, including Devspace, the firm's containerised development platform, on its own infrastructure. Agents are kept away from employee credentials and internal systems. Access widens in stages, and every new Claude model goes through the firm's own approval process before anyone there can use it. Thousands of automated tests and approval gates check each change, and an automated gate returns a hard pass or halt that the pipeline enforces, so Claude cannot sign off on its own work. People stay accountable for every production change.
That design answers a question many organisations have not asked yet. When a risk analyst or a product manager can produce a change that ends up in a production system, who owns it? The firm's answer is that the person does, and that the system around them is built so the agent never gets the final word.
No usage figures or headcounts came with the announcement, so it is too early to say how many people outside engineering are working this way. The direction is clear enough.
What this means for L&D
Widening agent access to non-engineers changes the curriculum for those people. A product manager whose requirements become a backlog has to specify work precisely. A risk analyst using a coding agent has to read its output critically and know when to stop. Everyone has to understand that accountability stays with them, whatever the agent produced. Those are judgement and governance skills before they are technical ones, which is the ground our AI Governance programs cover. Staged access only works if people's capability keeps pace with each stage.
Sources:
News and insights for innovation, digital transformation, future of work and L&D leaders.
Stay ahead of learning and development, corporate innovation and digital transformation news. Plus the future of work. For leaders in AU, NZ, HK, SG, the US, the UK and Canada.
