Janus Henderson deploys Claude across its investment arm

TL;DR:

• London-headquartered Janus Henderson, which manages around $500bn (about £370bn), is building AI-native research and client tools on Anthropic’s Claude.
• Two flagship systems — PRISM for client engagement and LIBROS for research management — are being built with transformation firm Percepta.
• The firm is also rolling out Claude Code and Cowork to engineering and corporate teams, signalling a firm-wide bet rather than a pilot.

Janus Henderson has set out one of the broadest AI deployments yet seen in asset management, putting Anthropic’s Claude at the centre of how it researches markets and serves clients. The London-headquartered manager is working with Percepta, a General Catalyst company, to build the tooling around its own data.

Purpose-built, not off-the-shelf

The deployment rests on two systems. PRISM, a client-intelligence platform, helps teams prioritise outreach, understand what clients hold and draft personalised communications. LIBROS, a research-management tool, pulls together the firm’s internal analysis, external research and market data so portfolio managers spend more time on judgement and less on retrieval. Both run on Claude and were built with Percepta’s embedded engineers via its Mosaic platform.

Janus Henderson’s argument is that generic chatbots rarely fit how an active manager works — the value comes from wiring frontier AI into proprietary research and rebuilding workflows around it. Beyond the two platforms, it is deploying Claude Code for engineers and Cowork across investment and corporate functions. Chief executive Ali Dibadj framed the ambition as becoming “the most technologically sophisticated asset manager in the world.”

As an announcement from the firm and its vendors, the claims warrant the usual caution — deployment breadth is easier to declare than to prove. Still, it lands amid a clear UK trend of financial institutions embedding AI in core operations, from NatWest’s rollout across trade finance to a regulatory backdrop where the FCA is leaning on existing rules rather than writing new ones.

Looking forward

The test is outcomes, not tooling. If PRISM and LIBROS measurably sharpen research or client service, expect rivals to follow; if they become expensive dashboards, the episode will join the long list of AI projects that impressed in the press release and underwhelmed in practice.