Goldman Sachs pilots Devin as an AI engineer next to its 12,000 developers
Goldman Sachs started a pilot with Cognition's Devin. The tech chief said the bank could run hundreds and later thousands of Devin instances on tasks like moving code to newer languages, under human supervision.
- Organization
- Goldman Sachs
- Industry
- Banking
- When
- Jul 2025
- Evidence
- Press report
- Agents and tools
- Devin
Reported outcome
Pilot started with plans for hundreds of instances; the CIO expects 3–4x developer output on routine code updates.
- ~12,000Human developers
- 3–4x (estimate)Expected output gain
Caveats
Reported by a news outlet, usually based on company statements.
- This is a pilot announcement, not a measured result.
- The 3–4x figure is the CIO's expectation.
What the agent does
Devin takes routine engineering tickets, such as updating internal code to newer languages or frameworks. Human engineers assign the work and review the results.
Why it matters
This is a large, regulated bank treating a coding agent as a team member at scale. Watch for follow-up data on merge rates and review cost.
What to copy
- Start agents on modernization backlogs that humans avoid.
- Decide the review rules before scale: who approves, and what tests must pass.
- Count agent "seats" like headcount, with a cost per merged change.
Starter prompt
Written by us.
Ticket: {ticket_id} — move {service} from {old_framework} to {new_framework}.
Constraints:
- Keep the public API identical.
- Keep test coverage at or above the current level.
- Do not touch files outside {service_path}.
When done, open a pull request with: summary, risk notes, and the exact
commands you ran to test it.
Sources
- Goldman Sachs is piloting its first autonomous coder — CNBC (via NBC Chicago), Jul 2025
Added Oct 16, 2026. Spotted an error? Send a correction.
Get new use cases weekly
One short email. Sources included.
Related use cases
Nubank used Devin to split an 8-year-old ETL monolith into sub-modules
Cognition reports 8–12x engineering-time efficiency and over 20x cost savings on the data-class migrations.
Stripe's Minions coding agents produce 1,000+ merged pull requests every week
More than a thousand pull requests merged each week are fully agent-produced and human-reviewed.
Spotify's Honk background agent merges fleet-wide code migrations: 1,500+ PRs and counting
1,500+ merged AI-generated PRs; 1,000 merged PRs took three months at first and about ten days later on.