Goldman Sachs is Experimenting With Viral AI Agent Devin as Its New Digital ‘Employee’
In a significant sign that AI is fast becoming a part of the financial services sector, Goldman Sachs is trying out Devin, the viral artificial intelligence software agent from Cognition AI, as a “new employee.” The experimental deployment is one of Wall Street’s earliest major efforts to officially introduce autonomous AI agents into the high-risk arena of investment banking.
Devin, unveiled in early 2024, skyrocketed as the first engineering AI scratcher able not only to:
- Code software
- Debug complicated systems
- Deploy web apps
- Think through high-level engineering problems
— all without permanent human control. Now, the AI agent is being put to work in one of the toughest corporate environments: Goldman Sachs.
A New Era of Financial AI Integration
At Goldman Sachs, a traditional early adopter of tech trends, the firm utilizes Devin to:
- Automate mundane software development tasks
- Manage infrastructure
- Facilitate internal application testing
The AI is not replacing any human jobs — not yet, anyway — but is being tested to see if it can:
- Alleviate engineering workloads
- Shorten development cycles
- Raise code quality
“Devin’s getting the treatment of a junior engineer,” a senior Goldman Sachs executive, who was not authorized to speak publicly, shared.
“The goal is to see how far it can go without human involvement, how good it is at integrating with our existing workflows, and if it’s actually able to improve our delivery pipeline.”
This testing is part of a larger push by Goldman to try out next-generation automation. A few years from now, the firm wants to determine if AI agents can become dependable digital colleagues.
Sources within the company indicate that Devin has already:
- Been tasked with internal tickets
- Put forth pull requests
- Engaged in supervised code reviews
— just like a human software developer.
From Viral Success Story to Corporate Test Case
Devin made waves earlier this year when it demonstrated that Cognition AI could complete end-to-end software engineering tasks, a first in the industry. Using just natural language queries, Devin can:
- Plan, write, and code
- Work with command-line tools
- Spin up servers
- Teach itself new programming frameworks on the fly
What sets Devin apart from older systems is not only its intelligence but also its autonomy. It functions as a full-stack engineer capable of managing complex coding environments, detecting bugs, and optimizing code while working with human teams on platforms like GitHub, Slack, and Jira.
The feature has drawn interest from:
- Tech companies
- Startups
- Large banks
While previous AI applications focused on customer service, data analytics, or fraud detection, Devin’s position at Goldman Sachs marks a shift to AI applied to core engineering and infrastructure.
The Promise and the Pitfalls
AI enthusiasts at Goldman envision agents like Devin being used not to replace, but to enhance human teams. In practice, Devin might:
- Take over repetitive or time-consuming coding
- Free up human engineers for strategic initiatives
- Generate cost savings
- Shorten development schedules
- Improve system reliability
However, experts warn that the road ahead is not without challenges.
“Safety, accountability, and trust are genuine concerns here,” said Dr. Nina Patel, an AI ethicist at NYU.
“When you are giving an A.I. system autonomy over code that can impact billions of dollars of financial assets, the stakes are extremely high. These systems need to be carefully monitored and audited.”
To mitigate risk, Goldman’s tech team has established strong guardrails:
- Devin’s code is reviewed by human engineers before merging
- The AI cannot push updates to live systems without authorization
- Internal logs record every line of code written
- Security teams frequently audit its behavior
Devin also operates within a sandboxed environment, preventing access to sensitive client data or mission-critical systems without proper clearance.
Shaping the Future of Work
The trial has rekindled conversations about the future of work in tech. Although Goldman Sachs insists that AI agents like Devin are intended to enhance, not replace engineers, the broader industry trajectory suggests that automation is already transforming software engineering.
For many developers, the shift is an opportunity rather than a threat.
“I don’t see Devin as competition — I see it as a strong deputy,” said one of the software engineers working on the project.
“It’s like having a super-able intern who never sleeps and always turns up on time.”
Cognition AI, the startup behind Devin, has also expressed excitement. In a recent blog post, they called the Goldman Sachs collaboration a:
“Key milestone in proving Devin’s usefulness in enterprise settings.”
They emphasized that real-world feedback is crucial as Devin continues to evolve.
Regulatory Implications on the Horizon
As AI agents become further integrated into corporate infrastructure, regulators are taking notice. The U.S. Securities and Exchange Commission (SEC) has yet to release a formal set of rules governing AI used in financial engineering, but experts believe that regulations are coming.
“AI governance is still behind the speed of innovation,” said Raj Malhotra, a financial compliance consultant.
“Companies like Goldman Sachs will help set the standard by which these systems are judged, secured, and held accountable.”
To stay ahead of regulatory scrutiny, Goldman Sachs has created an internal compliance and ethics subcommittee. The group includes:
- Lawyers
- Technologists
- Ethicists
Their task is to ensure that the firm’s use of Devin aligns with:
- Internal policies
- Anticipated regulatory frameworks
A Glimpse Into Tomorrow
Early returns from Goldman’s experiment with Devin have been encouraging—though not flawless. The AI agent sometimes requires:
- Multiple attempts to deliver successful results
- Further refinement in understanding high-level business logic
Still, Devin has already demonstrated effectiveness in:
- Bug fixes
- Test automation
- Performance improvements
If successful, Devin may soon be a digital full-time member of Goldman’s engineering team—an example of a hybrid workplace where humans and machine intelligence collaborate in real-time.
Conclusion
The financial services industry has often served as a bellwether for technological disruption, and Goldman Sachs’s bold move is likely to be closely watched across Wall Street and beyond.
As AI agents like Devin gain more capacity and trust, the very idea of what it means to be an employee may undergo a profound transformation.
For now, Devin is still a guinea pig—a digital intern learning the ropes in one of the world’s most competitive corporate environments. But if its trial at Goldman Sachs goes as planned, we may soon need to make room for AI in the official company roster.
