Skip to content
Agents tracked: 501 Releases (7d): 777 Agent status: no reported issuesSince last limit reset: Claude 10d · Codex 8d For AI agents: Markdown Updated Nov 21, 2026
Case study Norges Bank Investment Management (NBIM)Results as reported by Anthropic

How NBIM rolled out Claude to investment and compliance staff

NBIM, which manages Norway's sovereign wealth fund, deployed Claude across investment research, ESG analysis, compliance and data operations. Anthropic reports weekly time savings per employee and fast adoption.

The problem

Staff had to process large daily volumes of research reports, market data, regulatory filings and multilingual news. ESG reporting covered about 9,000 portfolio companies. NBIM needed a tool rigorous enough for this work that non-technical staff could still use.

How they did it

  1. Evaluated several LLMs with human-in-the-loop, finance-specific tests and chose Claude Sonnet 4.5 with extended thinking.
  2. Ran a two-week pilot, then expanded across departments.
  3. Created role-specific training paths and an AI Ambassador Network of 50 trained employees.
  4. Connected Claude Code to Snowflake and internal systems over MCP for data-quality checks and workflow automation.

Results as reported by Anthropic

  • According to Anthropic, employees save 20% of their week on the analytical and operational work they now do with Claude.
  • 600+ active Claude users across all departments within the first two months.

These figures are as reported by Anthropic in Claude customer story ↗. Anthropic sells the product used here. AgentGid did not measure or independently verify them.

Takeaway. Pair a broad rollout with role-specific training and a network of internal ambassadors so adoption extends beyond early enthusiasts.
AgentGid's take

This suits a larger organization with engineers who can wire Claude Code to a warehouse like Snowflake over MCP, plus budget for training and an ambassador network. Note that the 20% weekly time saving is Anthropic's figure, and Claude Code's downloads are slowing (-34% in 30 days). A small team could test the terminal side with Pi (Free, OSS), though you pay your model provider.

The agent used here

Similar use cases

Guides