HUB International rolled out Claude to 20,000+ staff and reports 2.5 hours saved a week each

Insurance broker HUB International deployed Claude Enterprise, Claude Code and custom agents to 20,000+ employees. It reports 2.5 hours saved per person per week and 90%+ satisfaction in early use cases.

Organization: HUB International
Industry: Insurance brokerage
When: Dec 2025
Evidence: Company reported
Agents and tools: Claude Enterprise Claude Code Claude API

Reported outcome

HUB reports an 85% productivity increase in targeted use cases, 2.5 hours saved per employee per week on average, and over 90% user satisfaction in early use cases.

20,000+ (company)
Employees covered
2.5 hours per employee per week, average (company)
Time saved
+85% (company)
Productivity in targeted use cases
over 90% in early use cases (company)
User satisfaction

Caveats

Stated by the company that uses the agent (blog, earnings call, interview). Not independently audited.

This is a joint press release by HUB and Anthropic. Both gain from the partnership.
The page gives no method, baseline, sample size or time period for any figure.
The gain applies only to 'targeted use cases' and 'early' ones, not all staff.
The page carries a May 2025 dateline but says the rollout began in late Q4 2025, so the timeline is unclear.
Much of the rollout is chat and coding help, not fully autonomous agents. The custom agent work is via the API platform.

What the agents do

HUB rolled out three Anthropic products: Claude Enterprise for knowledge workers, Claude Code for technology teams and the API platform for custom agent-based tools. The rollout was phased, with defined use cases for account managers, producers and customer support.

What happened

The release reports early results from targeted use cases. The page shows no baselines or method, so treat the numbers as the company's own claims.

What to copy

Roll out in phases, with named use cases per role.
Measure time saved and satisfaction from the start, and publish the method.
Keep custom agent work separate from general chat tools.

Starter prompt

Written by us, not by HUB.

Role: {role}
Our weekly tasks: {task_list}

For each task, estimate whether an AI assistant could help: yes, partly, or no.
For "yes" and "partly", say what a person still has to check.
List 3 tasks to pilot first and how we would measure time saved.