Joget AI Designer – Best Practices for Prompting and App Creation

Overview

Joget AI Designer is optimised for generating workflow-driven applications for Joget, focusing on assembling processes, forms, and components from high-level descriptions. Unlike general-purpose vibe-coding tools that generate code or detailed configuration structures, Joget AI Designer performs best when guided through progressive, process-oriented prompting.

This document provides recommended practices for creating applications with Joget AI Designer, supported by real-world examples and practical techniques to accelerate prototyping.

1. Guiding Principle: Workflow-First Prompting

The AI Designer is built around Joget’s workflow engine. Therefore, the most effective initial prompt is one that focuses on:

• What is the process?
• How many steps or approval levels?
• What is the high-level use case?

The ideal starting prompt is extremely concise, far simpler than prompts used with ChatGPT or Claude for code generation.

Recommended Prompt Pattern

“Generate an <app name> app with <N> levels of approval.”

Examples

• “Generate a 3-level purchase requisition approval app.”
• “Generate a car insurance claim app where the customer uploads photos and a claims officer reviews them.”
• “Generate a simple legal case management workflow.”

These short prompts allow the AI Designer to generate:

• Process workflow
• Core forms (auto-generated)
• CRUD structures
• Navigation menus

From this foundation, you can iteratively enhance forms, lists, and relationships.

2. Why Not Use Highly Detailed Prompts Initially? (Understanding Sequential Generation in Joget AI Designer)

Joget AI Designer generates applications through a sequential, stage-by-stage process, not all at once. The generation pipeline always starts with the workflow (process) and only afterwards proceeds to forms, lists, and other components. Because the workflow is the foundation for how the entire app is composed, the AI Designer prioritises building:

1. The process flow
2. Assignments and routing
3. Basic forms needed to support that flow

This has two important implications.

2.1 The Designer Starts With Workflow, Not Data Models

Even if the initial prompt includes detailed data structures, table definitions, or entity relationships, the AI Designer does not process those structures first. Instead, it:

• Creates the workflow
• Derives minimal supporting forms
• Generates only the fields necessary for that process

Since data modelling is not the initial step in generation, any attempt to specify full database schemas or relational models in the first prompt will often be ignored, partially captured, or interpreted incorrectly.

2.2 Why Detailed Prompts Are Ineffective at the Start

When the first prompt contains detailed data models, the AI must attempt to interpret them before the workflow exists. This imposes conflicting requirements on the engine:

• It tries to build the workflow and complex data models simultaneously
• It must guess how each entity aligns with process tasks
• It attempts to map fields without having a workflow structure to anchor them to

This often results in:

• Incomplete or broken workflows – routes, states, or tasks missing
• Misaligned or overgenerated forms – incorrect fields, redundant sections
• Lost or partial data relationships – foreign keys not generated or mismatched
• Overly complex app structures – unnecessary forms/lists created prematurely

The system is optimised to assemble apps incrementally, not to interpret fully specified architectures in one step.

2.3 When to Describe Data Models and Relationships

The correct stage to introduce detailed data structures is after the initial app and workflow are created, specifically when designing or refining forms.

At the form-design stage, the AI Designer:

• Has a stable workflow context
• Understands the entities required for each task
• Can correctly map fields to process states
• Accurately captures relationships, lookups, and form sections

This is the stage where you should provide:

• detailed entity definitions
• field lists
• relationships between objects

The AI Designer can then produce precise forms that match the target data model.

3. Recommended Iterative Workflow

Step 1: Generate Initial App

Use a concise, process-centric prompt (see above).
This creates the core workflow and working skeleton.

Step 2: Modify or Add Forms

Once the base app exists:

• Use “Add Form” to generate new forms via AI
• Use “Replace Form” to regenerate or expand existing ones
• Modify forms manually using the visual builder

Joget AI Form Designer is excellent for quickly generating 10+ field forms via one-line prompts.

Step 3: Enhance Workflow Logic

Enhance or refine:

• route conditions
• assignments
• status updates
• notifications

Step 4: Add Business Rules and Validations

(Where AI Designer is still limited)
Manual adjustments are recommended for complex logic, such as:

• conflict checks
• conditional validations
• cross-entity rules

4. What Enhanced Query Does (and Why You Should Always Use It)

Enhanced Query improves the quality, structure, and clarity of your instructions before they are processed by the Joget AI Designer. It does not generate Joget components itself; instead, it prepares your prompt so that the subsequent generation stage can interpret it correctly.

Different large language models (ChatGPT, Claude, Gemini, etc.) follow different prompting styles and patterns. Enhanced Query acts as a normalisation layer that adapts your input to these differences.

4.1 How Enhanced Query Works

Enhanced Query performs two key functions:

1. Natural Language Refinement

It rewrites your prompt into clearer, smoother, and better-structured language.
This reduces ambiguity and ensures the intent is easy for the LLM to understand.

Enhanced Query transforms your input into:

• logically sequenced instructions
• less fragmented descriptions
• easier-to-parse requirements

This alone dramatically improves LLM reliability.

2. Alignment With LLM-Specific Prompting Practices

Every LLM has different strengths and different patterns it responds best to.
Enhanced Query automatically adapts your instructions so that they align with the prompting style preferred by the active LLM.

This provides:

• More consistent responses
• Reduced misinterpretation
• A clearer understanding of workflow requirements
• Better overall generation accuracy

Enhanced Query ensures your prompt is optimised for whichever model is being used behind the scenes.

4.2 Why Enhanced Query Should Always Be Used

Enhanced Query ensures that your prompts enter the Joget generation pipeline already optimised and easy to interpret. Without it, prompts may be:

• Interpreted too literally
• Parsed inconsistently across different LLMs
• Misunderstood when describing multi-step processes

With Enhanced Query, the downstream generation stage (workflow creation, form creation, etc.) receives a cleaner, clearer request — leading to stronger, more predictable results.

Never skip Enhance Query.

5. Example: Legal Management System Prompt Evaluation

A long, detailed prompt describing:

• Entities
• Fields
• Relationships
• Statuses
• Multi-step flows

…is better suited for a general-purpose LLM, not AI Designer’s generation engine.

Joget AI Designer performs better when starting with:

“Generate a legal case management workflow with intake → conflict check → retainer → matter management.”

Then forms and modules are created iteratively. The image below displays the complete prompts for designing a complete Legal Management System Prompt Evaluation app with AI Designer.

6. Summary of Best Practices

Do's

✔ Start with a simple workflow prompt.
✔ Use Enhanced Query.
✔ Add or refine forms later.
✔ Use Form AI Designer for field-heavy forms.
✔ Iterate component-by-component.

Don’t

✘ Don’t specify every field in the first prompt.
✘ Don’t attempt to generate full data models upfront.
✘ Don’t describe complex validation logic initially.
✘ Don’t over-constrain the app structure in the first prompt.

7. Conclusion

Joget AI Designer is optimised for rapid workflow-centric app creation. Simplifying your initial prompt allows the AI to generate a clean, functional base application. From there, progressively enhance forms and logic using both the visual editors and the AI assistance built into the platform.

This process:

• Speeds up prototyping.
• Reduces design errors.
• Improves maintainability.
• Aligns with how Joget AI Designer is architected.

By following the guidelines in this document, users can achieve significantly more consistent and robust outcomes when building applications with Joget AI Designer.