Introduction to Vibe Coding
Vibe Coding is a new paradigm in coding, where developers and non-developers use natural language prompts to generate functional code. The concept was coined by Andrej Karpathy. He wrote in a post in X:
“There’s a new kind of coding I call “vibe coding,” where you fully give in to the vibes, embrace exponentials, and forget that the code even exists. It’s possible because the LLMs (e.g. Cursor Composer w Sonnet) are getting too good. Also I just talk to Composer with SuperWhisper so I barely even touch the keyboard. I ask for the dumbest things like ”decrease the padding on the sidebar by half” because I’m too lazy to find it. I ”Accept All” always, I don’t read the diffs anymore. When I get error messages I just copy paste them in with no comment, usually that fixes it. The code grows beyond my usual comprehension, I’d have to really read through it for a while. Sometimes the LLMs can’t fix a bug so I just work around it or ask for random changes until it goes away. It’s not too bad for throwaway weekend projects, but still quite amusing. I’m building a project or webapp, but it’s not really coding - I just see stuff, say stuff, run stuff, and copy paste stuff, and it mostly works.”
One aspect of this development paradigm is that the “developer” accepts the code with little to no review. The process basically uses natural language—either prompts or voice— to instruct the AI tool to generate code. If any error arises, then the “developer” asks the AI to fix it. “Developers” articulate their ideas, and the AI generates the code. Opposed to AI-assisted coding, Vibe Coding relies solely on the AI to code. The “ developer” does not review or need to understand the code. Instead, they “fully give in to the vibes” (Andrej Karpathy).
Whether this move is a true paradigm shift, a hip, or both, remains to be seen in the test of time. However, let’s acknowledge historical patterns. C++ did not make assembly absolute, and legacy COBOL programs are still around. Generative AI has proven to produce code as humans do; however, its quality does not surpass code produced by humans. This is primarily due to its training on human-generated code. The code it can produce can only match a human’s code quality.
The future of software development is unpredictable, even with a magic ball. There are significant implications for this development. Generative AI may become a partner in the coding process, where boilerplate code and trivial tasks can be sourced to the AI, and developers’ cognitive load would be freed for complex tasks and decisions. However, ethical challenges would persist. For example, in such a partnership, where the code is partly produced by the machine and the human, who is accountable for its quality, security, and maintainability? We may not currently possess all the answers. But, as a business analysis community, we need to reflect and assess how to embrace this evolution and integrate it into our practices.
How Business Analysts Can Leverage Vibe Coding?
Vibe Coding can be leveraged in business analysis by accelerating and enhancing Rapid Prototyping, Automating Technical Tasks, and Enhancing Requirement Validation.
Rapid Prototyping
Traditionally, BAs rely on wireframing tools (e.g., Balsamiq, Figma) or UX design- ers/developers to create UI prototypes. With Vibe Coding, BAs can describe UI elements using natural language, and the AI can instantly generate a working prototype, probably after a few iterations depending on the complexity of the design. Opposed to traditional tools, Vibe Coding may accelerate the iterative feedback cycle with stakeholders. Instead of going back to the static mockups for edits, in a live session with the end users, the BAs can modify UI elements on the fly during the meeting (e.g., prompt the AI: “Reduce padding on the sidebar”) to fine-tune based on feedback. This rapid feedback loop would enhance the engagement with the end users and prompt them to share more feedback as they see enhancements taking place immediately.
This adoption of Vibe Coding could reduce dependency on UX designers and developers for prototyping. Early-stage UI/UX prototypes can be created without UX designers and developers intervention, freeing up engineering resources for core system development and finalizing usability with the help of professional UX experts. Vibe Coding can assist BAs in experimenting with different UI structures without requiring coding expertise and using resources not dedicated to the requirements engineering stages of a product’s development.
Automating Technical Tasks
Business analysts often perform technical tasks such as querying databases, transforming data, and generating reports—tasks that typically require scripting or SQL knowledge. With Vibe Coding, these tasks can be automated using natural language, reducing the need for manual intervention from developers and other resources, such as data analysts, which will speed up analysis workflows.
Even when the BA is a skilled data analyst, Vibe Coding could assist in automating and accelerating related tasks. Instead of writing SQL queries manually, BAs can describe their data needs using plain language, and the AI can generate the corresponding queries. For example, instead of crafting a SQL statement from scratch, a BA could simply prompt the AI: “Retrieve the top 10 customers who made the highest purchases in the last quarter.” The AI would generate the necessary SQL query, which the BA can then run or refine as needed. This solution eliminates the barrier for BAs who may not be proficient in SQL but still need to extract and analyze business-critical data.
Similarly, data transformation tasks—such as converting CSV files to structured JSON, cleaning messy datasets, or formatting reports—can be streamlined through Vibe Coding. A BA could instruct the AI: “Convert this customer transaction CSV into a structured JSON format grouped by customer ID.” Instead of relying on data engineers, the BA can directly obtain a usable output.
Additionally, report automation is another area where Vibe Coding can be a game-changer. Instead of manually creating Excel summaries or Power BI dashboards, BAs can use AI-generated scripts to automate these processes. By leveraging Vibe Coding for technical tasks, BAs can significantly reduce reliance on developers and data engineers for day-to-day analysis work, allowing them to focus on higher-value strategic tasks. However, careful validation remains a challenge. AI-generated queries and transformations should always be reviewed before being used for decision-making to ensure accuracy and quality.
Enhancing Requirement Validation
Traditionally, business analysts rely on discussions, workshops, and documentation to refine requirements, ensuring that they are clear and aligned with business needs. However, some requirements remain ambiguous or complex. With Vibe Coding, BAs can take a more hands-on approach by rapidly generating functional outputs that allow for tangible validation of requirements before development begins.
One of the primary ways Vibe Coding can enhance requirement validation is by enabling real-time functionality generation based on natural language input. Instead of debating over the meaning of a requirement, BAs can instruct the AI to generate a quick proof-of-concept (POC) or basic implementation. This rapid transformation of rough requirements into tangible system behaviors will align understandings and bring the BA closer to their end user’s vision.
By incorporating Vibe Coding into requirement validation, BAs can reduce ambiguity and accelerate requirement refinement. Instead of relying solely on verbal and written communication, they can demonstrate working examples, allowing stakeholders to interact with AI-generated prototypes and adjust their expectations before handing requirements over to development teams.
Conclusion
Although the tooling market and the capability of AI are very close to the use cases we proposed in this article, we should remain realistic with such integration. Many challenges persist. For example, integrating such tools into an existing workflow raises security and privacy concerns for many organizations, as these tools remain literally black boxes. In addition, the scenarios we proposed may not be universal. Some of us are working on complex business cases; without prior and contextual training, the AI may not be as useful as we claimed in the discussed use cases. However, this could potentially become a reality in the near future! Lastly, this development may have other potential integrations into business analysis tasks and processes beyond the one discussed in this article. This leaves us with many challenging questions, e.g., is this development a disruption, an augmentation, or both?
Author: Adam Alami, Professor of Software Engineering, University of Southern Denmark
Adam Alami is an Associate Professor of Software Engineering at the University of Southern Denmark and a published author with a deep-rooted career in the information technology sector. With over two decades of experience, Adam began as a software developer before expanding his expertise into business analysis, project management, and organizational change. His professional background includes spearheading major IT transformation projects and guiding Agile implementations across diverse industries.
Adam’s research is at the intersection of software engineering, social sciences and human-AI collaboration, with a focus on how teams can leverage both behavioral insights and technical methodologies to improve software development outcomes. He delves into the socio-technical aspects of engineering processes, exploring the impact of ceremonies, team norms, and cultural values on quality. Adam holds a Ph.D. in Computer Science from the IT University of Copenhagen and frequently collaborates with industry partners to bridge academic research with real-world applications. LinkedIn: https://www.linkedin.com/in/adamalami/ X: @AdamAlamiDK