US-based Citi has unveiled Arc, a proprietary platform designed to enable the development and deployment of AI agents across the firm. The platform is intended to support the firm’s broader AI strategy, allowing developers to build agents for specific, well-defined use cases across its businesses and functions. Citi has stated that all agents operating on Arc will be monitored, auditable, and governed, with the firm retaining visibility into what each agent does, how it operates, and the value it generates.
From tools to agents: the next phase of AI at Citi
Arc represents a shift from AI tools that assist individual tasks to AI agents capable of taking on multi-step workflows autonomously. According to Citi, these agents are designed to handle tasks such as research, synthesis, preparation, and execution, to reduce manual effort and accelerate how teams operate. The firm positions this development as a natural extension of the AI foundation built over the past two years, during which it modernised its infrastructure and strengthened its data capabilities.
Citi reports that more than 80% of the 180,000 colleagues with access to its AI tools use them regularly, and that most have completed prompt training. Arc is designed to build on this base. Initially, access to the platform will be limited to the firm’s developers, who will use it to construct agents for targeted applications. Over time, Citi intends to expand access so that agents become integrated into everyday workflows across a wider range of roles and teams.
The firm has offered a specific illustration of how agentic AI could change the nature of client-facing work. In wealth management, a banker currently involved in preparing for client meetings, gathering portfolio data, analysing market trends, and modelling scenarios, could see these preparatory steps handled by a team of AI agents. The intended result is that the banker’s role shifts towards relationship management and advisory work rather than coordination and data retrieval.
Governance and responsible deployment
Citi has emphasised that the rollout of Arc is structured around the same risk and governance standards it applies to its broader technology capabilities. The platform is described as operating within the firm’s established risk framework, with each agent subject to ongoing monitoring and auditability. This approach reflects a wider industry trend in which large financial institutions are moving towards agentic AI deployments while simultaneously seeking to demonstrate oversight and control to regulators and stakeholders.
The launch positions Citi alongside other major financial institutions that are exploring or deploying agentic AI systems, as the sector moves beyond conventional generative AI tools towards more autonomous, workflow-integrated capabilities.
