Is now the time to design the next generation Customer Relationship Management (CRM) app that creates a better user experience for sales professionals? Is now the right time to embrace an AI Full Stack Engineer: an AI Agent that creates a CRM app from natural language prompts?

The integration of AI into software development may revolutionise the field, enabling rapid code generation and prototyping. AI Full Stack Engineers - AI systems that can handle all frontend, backend and database tasks - are becoming increasingly viable for quickly iterating on ideas and building functional prototypes. This is also referred to as Vibe Coding¹.

However, when it comes to developing production-class CRM apps, these AI-driven tools have, to date, revealed limitations. Of course, IT production environments demand robustness, security, scalability, and compliance: qualities that require the subtle, nuanced judgment and oversight of experienced Software Developers. However, I think now is the time to take the first step.

Let's explore this further.

The first step is engaging in a Mutual Value Discovery to determine where Vibe Coding can generate your next generation app: a CRM solution created at a fraction of the cost of existing technologies, yet inherently built around a more effective user experience.

This post argues that AI should serve as an augmentation tool for software developers rather than a replacement. At Being Guided we are now helping enterprises adopt Vibe Coding for next generation CRM apps, where non-programmers, or Citizen Developers, collaborate with Software Developers under the guidance of an AI Agent.

Timeline? This calendar year is when we expect to see Vibe Coding moving from prototypes to creating trusted production CRM apps. But, of course, this is a journey - not a destination.

At Being Guided, our approach with Vibe Coding leverages the strengths of both AI and human expertise while building apps through a No-Code First approach with a new AI Agent: Base44. Unlike other early stage 'AI Full Stack Engineers', Base44 is an 'All-In-One' Platform, including:

• Database Setup
• AI Large Language Model (Anthropic Claude 4)
• Email System
• Authentication
• Analytics
• Storage

Base44 is capable of generating code across the entire software development stack. This starts with prototyping with an inherent ability to quickly translate natural language prompts into fully functional code (Chen et al., 2021³). See screenshot below.

Base44 prototypes generated by AI may not account for all edge cases, security vulnerabilities, or performance bottlenecks - all key issues that are critical in production-class applications (Dahlin, 2021³). So our Discovery Engagements are truly minful of these realities. It's Phase One: Proof-of-Concept - putting everything to a rigorous business and technical analysis.

Creating a CRM App Using Base44

As illustrated above, at Being Guided we are creating an App Framework called 'CRM Pro'. Here we started with a simple Data Model of typical Entities (Objects) you would find in a CRM app: Leads, Contacts, Accounts and Opportunities. In this screenshot, you see the natural language prompts entered on the lefthand side, with the Lead Entity showing a List View of Fields.

Production-class CRM apps must meet stringent requirements for security, scalability, and maintainability. These apps often handle sensitive data, integrate with complex systems, and serve large user bases - all of which demand a level of precision and reliability, including:

Security Risks: AI-generated code must avoid introducing vulnerabilities, such as SQL injection or cross-site scripting, due to a lack of context about broader system (Perry et al., 2022⁷).

Scalability Challenges: AI tools may not optimize for performance or anticipate the need for horizontal scaling, leading to bottlenecks as user demand grows (Bass et al., 2019²).

Business Logic Complexity: AI struggles to fully grasp intricate business rules or compliance requirements, which are often context-specific, require human judgment (Holstein et al., 2019⁵).

Maintenance and Debugging: AI-generated code can be difficult to maintain, as it may lack the modularity and documentation that human developers prioritize (Lwakatare et al., 2020⁶).

These limitations underscore the need - at least for now - human oversight in the development of production-class SaaS apps with our Base44 AI Agent innovations.

Rather than replacing Software Developers, AI should augment their capabilities, enhancing productivity and creativity. In this model, AI tools assist Software Developers by generating boilerplate code, suggesting optimisations, or identifying bugs, allowing Software Developers to focus on higher-level tasks such as architecture design and business logic (Dahlin, 2021⁴).

Our approach to Vibe Coding is Pair Programming: where a non-programmer (known as Citizen Developer) collaborates with one or more Software Developers to co-create a CRM app. Guided by Base44 as our AI Agent, the Citizen Developer can contribute domain expertise and high-level requirements, whilst the Software Developers ensure technical rigour.

This pairing leverages the strengths of both parties: the Citizen Developer’s understanding of the problem space and the Software Developer’s ability to implement scalable, secure solutions (Holstein et al., 2019⁵).

Conclusion

Consider a scenario where a knowledge-intensive services company needs a custom CRM app to manage client portfolios. Using Vibe Coding, a Citizen Developer (an analyst) describes the app’s requirements in natural language, and Base44 generates the code.

A paired Software Developer then refines this code, ensuring it meets best practices, security standards and integrates with existing business logic.

At Being Guided our approach is to use Base44 to augment Software Developers, leveraging collaborative methods like Vibe Coding, where Citizen Developers and Software Developers work together, guided by a Large Language Model (LLM) powering the AI Agent: Base44.

This is also when at Being Guided we apply Fierce Reduction: the practice of simplifying all business processes, tasks and information systems by removing redundant or non-essential elements before considering Vibe Coding for next generation app development.

Next?

Let's Meet to explore this further.

References

1. Karpathy, A. [@karpathy]. (2025, February 2). 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. [Post]. X.
   https://x.com/karpathy/status/1753472166197080428
2. Bass, L., Clements, P., & Kazman, R. (2019). Software architecture in practice (4th ed.). Addison-Wesley.
   https://www.oreilly.com/library/view/software-architecture-in/9780136885979/
3. Chen, M., Tworek, J., Jun, H., Yuan, Q., Pinto, H. P. D. O., Kaplan, J., ... & Zaremba, W. (2021). Evaluating large language models trained on code. arXiv preprint arXiv:2107.03374.
   https://arxiv.org/abs/2107.03374
4. Dahlin, M. (2021). The role of AI in software engineering: A review. Journal of Systems and Software, 174, 110887.
   https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3633525
5. Holstein, K., Wortman Vaughan, J., Daumé III, H., Dudík, M., & Wallach, H. (2019). Improving fairness in machine learning systems: What do industry practitioners need? Proceedings of the 2019 CHI Conference on Human Factors in Computing Systems, 1-16.
   https://arxiv.org/abs/1812.05239
6. Lwakatare, L. E., Raj, A., Bosch, J., Olsson, H. H., & Crnkovic, I. (2020). A taxonomy of software engineering challenges for machine learning systems: An empirical investigation. Journal of Systems and Software, 162, 110496.
   https://colab.ws/articles/10.1007%2F978-3-030-19034-7_14
7. Perry, D. E., Porter, A. A., & Votta, L. G. (2022). Empirical studies of software engineering: A roadmap. ACM Computing Surveys, 54(3), 1-36.
   https://dl.acm.org/doi/10.1007/s10664-009-9121-0