Process.

At Being Guided we work in four steps for digital innovation: Think; Simplify; Value; and Build.

Think is Design Thinking: based on the Stanford d.school¹ method. It is five iterative stages to create a solid foundation for digital innovation. Simplify means stripping away of needless steps for user experience with Web apps. Applying Fierce Reduction until only the shortest path from intent to outcome remains - following The Laws of Simplicity by John Maeda².

When we apply Design Thinking to maximise human receptivity, we must also Value. This means applying the financial rigour of Value Engineering. Originally conceived by Lawrence D. Miles³, Value Engineering eliminates waste and determines value over price. With digital innovation this means generating a Return On Investment (ROI) Model for creating a new Web app.

The fourth step is Build: Vibe Coding your Web app. The term Vibe Coding was popularised by Andrej Karpathy⁴ to describe a conversational, high-level way of programming with AI models, where you specify the intent, style, or 'vibe' of what you want and let the model draft most of the implementation details of a Web app.

Platforms.

At Being Guided our Salesforce innovations are powered by an AI Agent: Abacus.ai AI Agent⁵.for Salesforce and Lovable⁶ for full stack Web apps The AI Agent underpinned by multi-Large Language Models (LLMs), such as: Anthropic Claude⁷: Google Gemini 3⁸: and, OpenAI GPT-4o⁹.

The AI Agent serves as Vibe Coding platform for Salesforce or Lovable Supabase¹⁰ infrastructure by enabling designers to build, automate, and deploy complex Web apps and workflows using natural language intent, rather than manual software coding.

Vibe Coding transforms the Web app development lifecycle from a traditional 'code-first' approach to an 'intent-first' generative process. With Lovable we are talking about the world's fastest-growing software company in history (Faria et al., 2026)¹¹,

In the jargon, traditional software experts use two words to highlight the potential risk associated with AI-powered Vibe Coding versus human-crafted software: probabilistic v. deterministic¹². This is simply referring to the inherent probabilistic nature of LLMs that underpin Lovable Vibe Coding.

A traditional software expert will rightly ask: "How do I know tthat he AI-generated code will not hallucinate my data?" The answer is this: the probablistic Vibe Coded User Interface (Lovable) has an underlying Core Logic that is rigidly validated, and therefore, is deterministic (Supabase).

There are a number of key things that we apply at Being Guided to ensure Lovable creates scalable, secure Web apps fit for the real world, as set-out below.

Pydantic/Zod Schema Validation¹³.
Every AI agent output must pass through a strict schema validation layer before hitting the Supabase database. If the LLM returns a string where a number belongs, the system will reject it.

Zod transforms a probabilistic AI experiment into an enterprise-grade System of Record. In a non-tech-centric business, 'data drift' or 'garbage-in' is the fastest way to lose executive trust. This is accomplished by implementing Zod (for the TypeScript frontend/backend) or Pydantic (for Python-based AI Agents).

Before AI was mainstream, Zod was created by developer Colin McDonnell¹³ as a tool for TypeScript (the language Lovable uses). At the time, its only job was to stop human users from breaking Websites:

The Problem:
A user might type "Free" into a box meant for "Price." This would "break" the database.

The Solution:
Developers wrote a 'Schema' (a digital blueprint). If the input didn't match the blueprint, Zod blocked it instantly.

We use Zod for more than just validation; we use it for Governance. When we prompt Lovable to "build a deal tracker," Lovable writes a Zod Schema in the background. This schema ensures that even if an AI agent is making autonomous decisions, the result of those decisions is forced into a deterministic structure. It creates a Zero-Exception Environment where your new Web app data remains clean, despite being built with 'vibes'.

Human-in-the-Loop (HITL).
Triggers: For high-value outcomes (e.g., moving a £1million deal to 'Closed'), the AI does not act; it proposes. Your sales professional reviews the AI Agent logic for final approval.

Unit Test Autogeneration.
We ensure that Lovable generates a Test Suite alongside the code. Every feature is backed by a deterministic test (Jest/Vitest) that must pass before deployment.

Task Automation.
The Lovable AI Agent offers a powerful toolset for automating everyday tasks: say, finding potential Opportunities and running Outreach Campaigns on LinkedIn, or running Email Campaigns on either Google Workspace or Microsoft Office 365. .

References.

1. The Hasso Plattner Institute of Design. (2004) Stanford d.school. https://dschool.stanford.edu/about
2. Maeda, J. (2006). The Laws of Simplicity. Design, Technology, Business, Life. Cambridge, Great Britain: MIT Press.
   https://mitpress.mit.edu/9780262539470/the-laws-of-simplicity/
3. Miles, L.D. (1947). The Lawrence D. Miles Value Engineering Reference Center Collection.
   https://minds.wiscon.edu/handle/1793/301
4. Karpathy, A. [@karpathy]. (2025, February 2). There’s a new kind of coding I call “vibe coding”, where you fully give in to vibes, embrace exponentials, and forget that code even exists. X. https://x.com/karpathy/status/1753472166197080428
5. Abacus.ai. (2025). Abacus AI Deep Agent: The world's first super assistant. https://deepagent.abacus.ai/
6. Lovable. (2026). CRM Systems: AI-Powered Customer Relationship Management. https://lovable.dev/solutions/use-case/crm-systems
7. Anthropic. (2024). The Claude 3 model family: Technical report.
   https://www-files.anthropic.com/production/images/Model-Card-Claude-3.pdf
8. Google. (2024). Gemini 1.5: Unlocking multimodal understanding across millions of tokens of context. Google DeepMind Technical Report.
   https://storage.googleapis.com/deepmind-media/gemini/gemini_v1_5_report.pdf
9. OpenAI. (2024). Hello GPT-4o: OpenAI’s new flagship model. OpenAI System Card. https://openai.com/index/hello-gpt-4o/
10. Lovable.dev. (2026). Supabase Integration & Architecture for Enterprise. https://lovable.dev/docs/supabase
11. Faria, C., & Wang, J. (2026). Is vibe coding the future of software? IEEE Xplore, 7. https://doi.org/10.1109/MITP.2026.3533851
12. Warden, P. (2025). Why Deterministic Wrappers are Essential for LLM Production Systems. AI Engineering Journal.
    https://petewarden.com/2025/deterministic-llms
13. McDonnell, C. (2024). Zod (Version 3.23.8) [Software].
    https://github.com/colinhacks/zod