Build Business Apps with AI (Vibe Coding)
A finance officer who needs a claims tracker, an operations lead who wants a booking tool, a founder who would rather test an idea this week than brief a developer for a month. AI app builders can now produce all three from a written description, and that is exactly where the trouble starts. The app runs, so it looks finished, and nobody on the team can say whether it is correct, safe or maintainable. This two-day programme teaches non-developers to build a working application with AI, then spends as long on the second half: checking it, securing it, deploying it and owning it, so that a weekend project does not turn into a data protection incident.
Programme Agenda
What Vibe Coding Is and Where It Stops
Describing software in plain language and letting an AI tool write, run and revise the code. What is really happening in the loop between your prompt, the generated code and the live preview. What an application is made of, in words a non-programmer can hold: screens, data, rules, sign-in and hosting. Where these tools are quick and dependable, and where they stall or quietly get things wrong, such as permissions between users, payments, integrations and rules with exceptions. Participants build a small first app from a single paragraph so the loop is familiar before the programme leans on it.
Writing a Brief the AI Can Build From
Most poor results come from a vague first prompt, not a weak tool. Turning a business problem into a one-page brief: who uses it, what they do on each screen, what data is kept, which rules apply and what done looks like. Asking the tool to propose a plan before it writes anything, then building in small steps rather than one enormous prompt. Keeping a running notes file so the tool does not forget a decision made an hour earlier. Participants write a brief for a real tool from their own work and use it for the rest of the programme.
Building Screens and Iterating Without Breaking Things
Forms, tables, filters, dashboards and simple workflows, built and refined through short change requests. Reading the preview critically, giving feedback the tool can act on, and spotting the moment an AI edit has broken something that used to work. Checkpoints, version history and undo, and how to decide between fixing forward and stepping back. Hands-on: build a multi-screen version of the brief, then roll back a deliberate bad change.
Data, Sign-In and Who Can See What
Where an application's information actually lives, and why a spreadsheet, a hosted database and a file on someone's laptop behave so differently. Tables, fields and records in plain terms. Connecting a managed database service (Supabase or Firebase are typical), adding sign-in, and setting roles so a staff member sees their own records and a manager sees the team's. Access rules are the item most often left undone, so this module builds them rather than describing them. Hands-on: add a database and sign-in to the app and prove that one user cannot open another user's records.
When It Breaks: Reading Errors and Debugging with AI
Error messages are less frightening once you know the few kinds that exist. Reading a browser console at a beginner level, copying the right evidence to the assistant, and describing a fault so it is fixed rather than papered over. Why asking the tool to fix it five times in a row makes things worse, the one-change-at-a-time habit that prevents it, and recognising when the tool is going in circles. Hands-on: diagnose and fix three faults planted in a broken app.
Reviewing What the AI Wrote
You do not need to read every line to review sensibly. A non-programmer's review: what to check in the running app, what to ask the assistant to explain, and which warning signs justify stopping. Running the app against bad input, empty fields and the wrong user. Using a separate AI session as an independent reviewer. Weighing the packages and libraries the tool suggests, since an assistant cannot vouch for the reputation of what it recommends. Knowing when something needs a professional engineer before it goes near real users.
Security and Personal Data
The ways AI-built applications leak data are few and repetitive: a database with no access rules, keys and passwords pasted into code or a public repository, a spreadsheet published to the web, admin pages with no sign-in, and no record of who did what. Each is demonstrated, then fixed. What to collect in the first place, how to mask fields such as IC numbers, and what to keep out of prompts and test data. How the Personal Data Protection Act 2010, as amended in 2024, applies to what you build, including the point that data processors now carry security obligations of their own. This module is practical orientation, not legal advice. Hands-on: audit your own app against a security checklist and fix every finding.
Deploying, Owning and Maintaining It
Putting the app on a live address with a proper domain. Keeping a test copy separate from the live one so an experiment never touches real records. Storing the code in a repository you control so you are not tied to one tool, and how version history works in plain terms. Backups, watching the running cost of the tools involved, and who maintains the app when its builder changes role. A frank framework for deciding whether something should be built this way, bought as software, or handed to IT. Final project: each participant demonstrates a working, secured and deployed application built from their own brief.
Learning Outcomes:
Turn a business problem into a written brief that an AI tool can build from
Build and refine a multi-screen application through short, controlled changes
Connect a database and sign-in so each user sees only their own records
Diagnose and fix common faults without reading every line of code
Review AI-generated work with a checklist and an independent second opinion
Spot and fix the security mistakes that most often expose personal data
Deploy an application on a live address and set out who will maintain it
Duration: 2 Days (16 Hours)
Training Hours: 9:00 AM to 5:00 PM
Level: Intermediate
Training Mode: Physical, Online, or Hybrid
HRD Corp SBL-KHAS Claimable
Certificate of Completion included
Related programmes:
Build Your Own Website Using AI (for a brochure website rather than a working application)
No-Code Automation for Business (for connecting the tools you already use instead of building a new one)
PDPA Awareness and Compliance (for the data protection duties behind the security module, in depth)
AI Governance and Ethics (for the acceptable-use rules a team needs before AI-built tools go live)
Frequently Asked Questions
More in AI and Automation
- AI for Marketing and Content Creation · All levels, 2 days
- AI for Workplace Productivity · All levels, 1 day
- Practical AI for Business Operations · Intermediate, 1 day
- No-Code Automation for Business · Beginner to Intermediate, 1 day
- AI Fundamentals for the Modern Workplace · Beginner, 1 day
- AI for Business Leaders · Senior / Leadership, 1 day
- Generative AI for Work · All levels, 1 day
- AI Governance and Ethics · Intermediate, 1 day