Effective Prompting Guide for ERP•AI

Effectively communicating with ERP AI ensures that you get precise, actionable results. This guide will help you craft clear and structured prompts, making it easier for the AI to understand your requests. ERP AI consists of specialized agents designed for tasks like app creation, data management, reporting, and role-based access control. Follow these best practices and examples to get the most out of your ERP AI assistant.
Introduction
This guide will help you effectively communicate with your ERP AI assistant to accomplish various tasks. The AI automatically understands your intent and routes your request to the appropriate specialized agent. While you don't need to specify which agent to use, understanding their capabilities can help you frame your requests better.
Understanding ERP AI Agents
The system consists of several specialized agents, each handling specific types of tasks:
Auto Agent
- Acts as the main interface, understanding your requests and routing them to appropriate specialized agents.
- You don’t need to specify which agent you want to use—just express what you want to achieve.
- Maintains context of your conversation and ensures consistent handling.
App Creation Agent
- Creates new applications from scratch.
- Sets up database schemas and tables.
- Configures initial roles and permissions.
- Handles basic app setup and structure.
Dashboard Agent
- Creates individual charts and visualizations.
- Supports four types of widgets:
- Bar charts (for comparisons)
- Pie charts (for proportions)
- Line charts (for trends)
- Stat cards (for single metrics)
- Creates one visualization at a time for best results.
Record Management Agent
- Handles all data operations (CRUD):
- Creating new records
- Reading/querying existing records
- Updating record values
- Deleting specific records
- Works with any table in your app.
Reporting Agent
- Generates comprehensive data reports.
- Performs data analysis and aggregations.
- Creates SQL queries from natural language.
- Provides insights and summaries.
App Editor Agent
- Creates new tables and views.
- Adds new fields to existing tables.
- Modifies table and view configurations.
- Cannot delete tables, views, or fields.
Roles Management Agent
- Creates and updates role definitions.
- Manages permissions and access levels.
- Handles role-based access control.
- Sets up user role mappings.
User Invitation Agent
- Manages user invitations to your application.
- Assigns appropriate roles to new users.
- Handles email-based invitations.
- Ensures proper access control during onboarding.
General Prompting Guidelines
Be Specific and Clear
❌ DON’T
Show me sales data
✅ DO
Show me the total sales by product category for the last month, including the sales representative who made each sale
Provide Context
❌ DON’T
Update the status
✅ DO
Update the status to 'Completed' for all orders from customer ABC Corp that are currently 'In Progress'
Use Incremental Steps
❌ DON’T
Create a complete HR system with everything
✅ DO
First
Create an HR app with employee management and leave tracking
Then
Add a performance review table with fields for goals, achievements, and ratings
Prompting by Intent
Creating a New App
✅ Best Practices
- Describe the purpose of the app.
- List main features needed.
- Specify key data entities.
- Mention user roles if known.
Example
✅ Good
Create a project management app where:1. We need to track projects, tasks, and team members2. Each project has a name, start date, end date, and status3. Tasks should have assignees, due dates, and priority levels4. Team members need different access levels (admin, manager, member)
❌ Avoid
Make a project app
Adding/Editing Data Records
✅ Best Practices
- Specify the table/entity.
- Provide all required information.
- Include any conditions for updates/deletes.
- Mention relationships if relevant.
Examples
Creating Records
Add a new employee with:- Name: John Smith- Department: Engineering- Position: Senior Developer- Start Date: Today- Manager: Sarah Johnson
Updating Records
Update the status to 'Approved' for all expense claims submitted by John Smith in the last week
Creating Dashboard Widgets
✅ Best Practices
- Specify the type of visualization (bar, pie, line, or stat).
- Describe the data you want to see.
- Mention any time periods.
- Include any grouping or filtering.
Examples
✅ Good
Create a bar chart showing monthly sales by product category for this year
✅ Also Good
Show me a stat card with the total number of active projects
❌ Avoid:
Create multiple charts showing sales, inventory, and customer data
Generating Reports
✅ Best Practices
- Specify what data you need.
- Include any calculations or aggregations required.
- Mention any filtering criteria (e.g., date range, category, region).
- Describe how you want the results presented (e.g., summary, detailed breakdown, comparison).
Examples
✅ Good
Generate a sales performance report showing:1. Total revenue by region2. Top 5 products by sales volume3. Compare this month's numbers with last month
❌ Avoid
Give me a sales report
Creating New Tables or Views
✅ Best Practices
- Describe the purpose of the table or view.
- List required fields and their data types.
- Mention any relationships with existing tables.
- Specify access permissions if needed.
Examples
✅ New Table
Create a Vendors table with:- Company Name (text)- Contact Person (text)- Email (text)- Phone (text)- Category (select: Hardware, Software, Services)- Status (select: Active, Inactive)
✅ New View
Create a 'My Team's Tasks' view that:- Shows tasks from the Tasks table- Only includes tasks assigned to my team members- Shows task name, due date, status, and assignee- Allows editing of status and comments
Managing Roles and Permissions
✅ Best Practices
- Specify the role name.
- List required permissions.
- Mention any specific restrictions.
- Include access levels for different features.
Example
✅ Good
Create a Sales Manager role with:1. Full access to customer and order records2. Can view team's performance reports3. Can approve orders up to $50,0004. Can manage team members
❌ Avoid:
Make a sales role
Common Pitfalls to Avoid
- Overly Complex Requests
- ❌ "Create a dashboard with 10 different charts"
- ✅ "Let's start with a sales overview chart, then we can add more"
- Vague Instructions
- ❌ "Make it better"
- ✅ "Add a status column to track project progress: Not Started, In Progress, Completed"
- Missing Context
- ❌ "Update the price"
- ✅ "Update the price to $299 for all products in the Electronics category"
- Requesting Unsupported Operations
- ❌ "Delete the Customers table"
- ✅ "Add an 'Active' status field to the Customers table"
Important Limitations
- Dashboard Creation
- Can only create one chart/widget at a time.
- Must specify exact data and visualization type.
- Cannot delete or modify existing charts.
- Table/View Management
- Cannot delete tables or views once created.
- Cannot delete fields from tables.
- Can only add new fields or create new tables/views.
- Data Operations
- All data operations must specify exact tables and conditions.
- Updates and deletes require clear filtering criteria.
- Complex joins should be broken down into simpler queries.
Conclusion
- Be specific and clear in your requests.
- Provide necessary context.
- Break down complex tasks.
- Focus on what you want to achieve.
- The AI will guide you if it needs more information.
- Respect system limitations.
The more precise and structured your prompts are, the better results you'll achieve with your ERP AI assistant.