Is vibe coding reaching its limits in the enterprise?

AI-powered development is accelerating application creation. Explore how enterprises can balance the speed of vibe coding with security, IT governance, and long-term application management. Copy
Vibe Coding en entreprise

Artificial intelligence is transforming how business applications are designed and developed. While AI-powered code generation now makes it possible to move from an idea to a working prototype in a matter of hours, deploying these applications at scale raises important questions about reliability, architecture, governance, and long-term maintenance.

For enterprises, the challenge is to balance business autonomy, faster digital transformation, and control over the entire application lifecycle.

Vibe Coding Meets Enterprise Requirements

In October 2026, Goldman Sachs revised its internal policies governing the use of advanced AI-powered coding tools.

According to information reported by Bloomberg in early November and subsequently covered by Silicon.fr, the bank restricted access to certain tools, including Anthropic’s Claude Code and advanced GitHub capabilities, for many employees outside its technical teams.

These tools remain available to software engineers and hybrid professionals known as “strats,” who combine financial expertise, quantitative research, and software development skills.

Just a few months earlier, in July, Goldman Sachs CIO Marco Argenti had described vibe coding as an opportunity to bring business teams and IT closer together.

The idea was to allow employees to describe their needs in natural language and rapidly create their own prototypes before handing them over to engineers for production-ready development.

Potential use cases were practical: monitoring client transactions, modeling market scenarios, performing financial calculations, and automating operational tasks.

After several months of experimentation, the bank ultimately chose to limit access to professionals best positioned to use these tools effectively over time, taking productivity gains and risk management into consideration.

This decision does not signal a retreat from AI. Goldman Sachs continues to deploy artificial intelligence across different areas of its business.

Instead, it highlights a growing challenge for large organizations: how can enterprises democratize application development without losing control over how applications operate and evolve?

When Rapid Prototyping Creates a New Form of Shadow IT

The main appeal of vibe coding lies in its simplicity.

An employee can describe a business requirement, generate an interface, and quickly obtain a tool that appears to meet their needs.

This approach encourages experimentation and significantly reduces the time required to turn an idea into a working prototype.

However, its limitations become apparent when these prototypes begin to be used and shared across an organization.

Reports from software developers in the technology press have highlighted the proliferation of locally developed applications, often created without centralized oversight and sometimes reused by other employees.

This trend introduces a new form of Shadow IT, where AI-generated applications gradually become part of the enterprise technology landscape without necessarily meeting established technical standards.

Common concerns include authentication, access management, data protection, testing, and deployment procedures.

A prototype may work perfectly during a demonstration while still requiring substantial additional work before it can operate reliably within an enterprise environment.

For IT teams, the risk is that time saved during prototyping will be offset by the effort required to rework, secure, deploy, and maintain these applications.

AI does not eliminate technical debt. Without a structured development framework, it may simply shift that debt to the teams responsible for bringing applications into production.

Faster Development Also Requires Stronger Controls

The debate surrounding vibe coding extends beyond Goldman Sachs.

At JPMorgan Chase, artificial intelligence is also playing an increasingly important role in software development.

The bank is deploying AI assistants across technical teams and various business functions while maintaining a strong focus on application quality.

Gill Haus, CIO of Chase, has emphasized the growing importance of software engineering fundamentals as AI accelerates code production.

As the volume of generated code increases, traditional verification processes must evolve alongside it.

Automated testing, human review, security controls, and structured deployment procedures become essential to maintaining application reliability.

This shift also changes how enterprises should measure the productivity gains delivered by AI.

Productivity can no longer be assessed solely by the number of lines of code produced or the time required to generate an initial application.

It must also account for the work involved in validation, production deployment, ongoing maintenance, and future changes.

For enterprises, the question now goes beyond selecting AI models or coding assistants.

It concerns their ability to bring AI-powered development into production at scale, integrate new applications into their existing architecture, and manage their evolution over time.

Use AI to Define Requirements, Build Within a Controlled Framework

One alternative is to establish a clear distinction between using AI to formalize business requirements and the process used to build the application itself.

Artificial intelligence offers significant potential to help teams articulate their needs, structure business rules, and produce actionable specifications.

This stage, which often requires considerable time in traditional IT projects, can be accelerated without necessarily allowing a generative model to make every technical decision involved in application development.

This distinction is central to kShuttle AI, the artificial intelligence capabilities integrated into the ExRP® platform.

AI supports the preparation of application specifications based on requirements expressed by business teams.

The application is then built through a deterministic process using proven application components.

This approach makes it possible to reproduce the same result from the same specification while keeping development within a structured application framework.

For enterprises, the advantage is a more direct connection between business requirements and operational applications, without multiplying isolated development projects or relying exclusively on freely generated AI code.

This approach brings business, IT, and transformation teams together within a shared application framework.

Business requirements can be translated into operational applications more quickly while maintaining architectural standards, governance requirements, and technological control.

From Application Development to IT Modernization

Application transformation is not limited to new development projects.

In large organizations, many critical business processes still rely on legacy applications, custom-built systems, and Excel spreadsheets that have become essential to day-to-day operations.

These systems often support important business needs, but evolving them can become costly and complex as business rules or organizational priorities change.

New regulations, organizational restructuring, changes in reporting requirements, or business process transformation may require updates to interfaces, modifications across multiple applications, or significant maintenance resources.

As AI accelerates business transformation, the ability to adapt existing applications is becoming just as important as the ability to create new ones.

With kShuttle AI, this approach supports an application lifecycle built around three complementary capabilities:

  • Build: Develop new performance management applications based on business requirements.
  • Rebuild: Modernize existing applications to meet evolving business and organizational needs.
  • Run: Continuously adapt applications as business strategies evolve, within a governed environment and under the control of enterprise teams.

Within the ExRP® platform, these capabilities support a progressive approach to IT modernization, integrating new applications with the systems organizations choose to retain.

The goal is no longer simply to develop applications faster, but to establish an application environment that can evolve sustainably with the organization.

Regulatory Expertise Supporting Broader Application Transformation

For large enterprises, developing reliable applications requires control over data, business rules, processes, and the associated validation mechanisms.

These requirements are particularly important in financial and regulatory environments, where calculation traceability and data reliability are essential.

Historically recognized for its expertise in financial and sustainability regulations, including IFRS 16, Pillar Two, and CSRD, kShuttle is now applying this experience to a much broader range of business performance management processes.

These requirements span multiple departments and business functions:

Finance, including reporting, forecasting, and cash management; Tax; Sustainability and ESG; Human Resources; Risk and Environment, Health and Safety (EHS); as well as other enterprise performance management (EPM) processes.

While each department has its own applications, business rules, and objectives, all contribute to the organization’s overall strategy.

The ExRP® platform brings applications, data, business rules, and workflows together on a shared foundation while integrating with existing systems.

This enables different departments to address their specific requirements while improving alignment across the organization.

With kShuttle AI, this expertise extends to the creation, modernization, and continuous evolution of business applications, combining the capabilities of artificial intelligence with a governed application environment.

The Future of Enterprise Applications Goes Beyond Development Speed

The experiences of Goldman Sachs and JPMorgan Chase highlight two dimensions of the same challenge.

AI is fundamentally changing software development practices, but it does not eliminate the need for quality, reliability, or technological governance.

For large enterprises, the challenge is to take advantage of faster development without creating an expanding portfolio of applications that are difficult to maintain, secure, or adapt.

This requires considering the entire application lifecycle, from the initial expression of business requirements through deployment, operations, and modernization.

By combining AI-assisted specification, deterministic application development, and application governance, kShuttle AI and the ExRP® platform offer a structured approach to enterprise application transformation.

Because an enterprise application should do more than address an immediate business need. It should also be able to support the organization’s strategy as it evolves.

What if your next transformation project started with a simple description of your business requirements?

Discover how kShuttle AI can help turn those requirements into a performance management application.

Ask for a demo


FAQ – Vibe Coding and Enterprise Application Governance

What is vibe coding in an enterprise environment?

Vibe coding is an AI-assisted software development approach that allows users to generate code through natural language instructions.
In enterprise environments, this method simplifies prototyping and enables faster experimentation with new business applications. However, moving these applications into production requires additional controls to ensure security, reliability, and maintainability.

What are the main risks of vibe coding for enterprises?

The main risks include data security, access management, code quality, maintenance, and the proliferation of applications developed without centralized oversight.
Without appropriate governance, vibe coding can contribute to Shadow IT and increase technical debt, particularly when AI-generated applications become part of critical business processes.

Why isn’t an AI-generated application necessarily production-ready?

A working prototype does not guarantee that an application meets enterprise IT requirements.
Before deployment, applications must satisfy architectural, cybersecurity, testing, traceability, integration, and maintenance standards.
These requirements are essential for ensuring reliable operations and compatibility with the organization’s existing IT environment.

How can enterprises balance business autonomy with IT governance?

Organizations can enable business teams to express their requirements and participate in application design while maintaining a shared technical framework.
Structured specifications, validation procedures, access controls, and proven application components help balance business autonomy with IT governance.
This approach enables faster application development without compromising enterprise technology standards.

What is the difference between AI-generated code and deterministic application development?

AI code generation uses generative models to produce software code from user instructions. Results may vary depending on the model, prompt, and context.
Deterministic application development relies on defined specifications and established components to produce predictable and reproducible results.
AI can therefore accelerate the specification process while application development follows a controlled and consistent framework.

Can artificial intelligence help modernize legacy applications?

Yes. AI can support the analysis of existing applications, documentation of business rules, and preparation of modernization projects.
This approach can help organizations progressively adapt legacy applications to new business requirements, provided that technical dependencies, data structures, and integrations are properly assessed and validated.

How does kShuttle AI support the development and evolution of business applications?

kShuttle AI, integrated into the ExRP® platform, supports the structuring of business requirements and the preparation of application specifications.
Combined with the App Builder and the platform’s application components, it contributes to the development, modernization, and continuous evolution of business applications within a governed, traceable environment connected to the enterprise IT landscape.

Sources and References

kShuttle on LinkedIn

Don't miss our upcoming publications

Follow kShuttle on LinkedIn to receive notifications when we release articles, analyses and regulatory news.

Follow us

Latest posts

revised ESRS 2026 reporting

Revised ESRS Standards: the delegated act is published, what options for 2026 reporting?

The European Commission has adopted the delegated act on the revised ESRS. Companies now need to understand what changes, which options apply for 2026 reporting and how to prepare for the revised standards from 2027.
EFRAG State of Play 2026 and ESRS reporting maturity

ESRS: what EFRAG’s 2026 report really says about corporate maturity 

EFRAG’s State of Play 2026 report provides one of the clearest views yet of how companies are applying ESRS in practice. Beyond compliance, the report raises a key question: are companies only learning to publish sustainability information, or are they starting to manage what they declare as material?
Enterprise regulatory platform: reducing the cost of complexity

Ména Dogan: “Companies do not need more tools. They need to reduce the cost of complexity.”

In this interview, Ména Dogan looks back on the evolution of kShuttle, the development of ExRP and why companies need a platform to reduce the cost of regulatory complexity.