Write better prompts, get better answers

Learn how to ask questions that get accurate, actionable answers from your AI assistant. Transform vague queries into precise insights about your Odoo data.

Understanding prompt structure

Build effective prompts with context, action, and constraints

A well-crafted prompt contains three elements:

Context: What situation or background matters
Action: What you want the AI to do
Constraints: Any limitations, formats, or preferences

Context: We're preparing for Q4 planning

Action: Show me inventory levels for electronics products

Constraints: Only items with stock below 50 units

Prompt categories

Compare weak vs. strong prompts across different use cases

Data queries

Vague approach

Show products

Specific approach

Show the top 20 products by sales revenue in the last quarter, including product name, category, units sold, and total revenue

Analysis and insights

Basic approach

What's our revenue?

Insightful approach

Compare our revenue for the last 3 months. Break it down by product category and highlight any significant trends

Common mistakes to avoid

Learn from typical pitfalls that reduce AI accuracy

Being too vague

"Check sales"

"Show me total sales for October 2024, broken down by week"

Mixing unrelated questions

Ask one thing at a time or clearly prioritize. The AI performs better with focused requests.

Not providing IDs

"Update the laptop price"

"Update the price of product ID 42 (Gaming Laptop Pro) to $1,299"

Omitting time ranges

"How many customers?"

"How many new customers did we acquire in the last 30 days?"

Reusable prompt templates

Fill-in-the-blank templates for common query patterns

Template 1: Filtered list query

Show me [ITEMS] that meet these criteria:

- [CRITERION 1]

- [CRITERION 2]

- [CRITERION 3]

Sort by [SORT FIELD] and limit to [NUMBER] results

Include fields: [FIELD 1], [FIELD 2], [FIELD 3]

Template 2: Comparison analysis

Compare [METRIC] between [TIME_PERIOD_1] and [TIME_PERIOD_2].

Break down by [DIMENSION] and highlight [INSIGHT_TYPE]

Domain-specific examples

Real-world prompts for sales, inventory, and operations

Sales and CRM

Show me all opportunities in the "Proposal Sent" stage that haven't had activity in 7+ days. These might need follow-up.

Inventory management

Alert me to any products where current stock is below the minimum reorder quantity. Include supplier information.

Customer support

Analyze our support ticket resolution times. What's the average? Are we meeting our 24-hour SLA? Show breakdown by ticket type.

Accounting and finance

Show me all invoices over 30 days past due, sorted by amount. Include customer contact details and outstanding balance.

Purchasing

List all purchase orders in "To Approve" status that have been waiting more than 3 days. Sort by order value, highest first.

Pro tips

Advanced techniques to improve response quality

Use discovery context: Reference discovery by saying "based on what you know about our business" to leverage your knowledge base

Request specific formats: Ask for tables, bullet lists, or specific field ordering to get results in your preferred format

Iterate and refine: If the first answer isn't perfect, refine your question with more specifics based on what you received

Set limits: Always specify limits (top 10, top 20) to avoid overwhelming results and improve response time

Chain of thought experiments

Use step-by-step reasoning for complex Odoo analysis and problem-solving

What is Chain of Thought (CoT)? Instead of asking for a direct answer, you prompt the AI to "think step by step" through the problem. This dramatically improves accuracy for complex multi-step tasks like debugging Odoo workflows, analyzing financial discrepancies, or investigating data anomalies.

When to use chain of thought

Complex analysis: Multi-step calculations, financial reconciliation, data investigation
Debugging workflows: Sales order stuck, invoice not generated, inventory discrepancy
Root cause analysis: "Why did this happen?" questions that need investigation
Decision making: Evaluating options, comparing approaches, planning implementations

Example 1: Debugging a stuck sales order

Direct question

Why isn't sales order SO-2024-001 generating an invoice?

Chain of thought

Let's investigate why sales order SO-2024-001 isn't generating an invoice. Think step by step:
1. Check the order status and stage
2. Verify if there are any delivery requirements
3. Look for invoicing policy settings
4. Check for any blocking issues or error logs
5. Explain what you found and what's preventing the invoice

Example 2: Financial reconciliation

Simple query

Why doesn't our revenue match between the sales report and accounting?

Chain of thought

Investigate the revenue discrepancy between sales report and accounting. Work through this systematically:
1. Get total revenue from sales orders (confirmed orders in Q4 2024)
2. Get total revenue from posted invoices (same period)
3. Compare the two numbers - what's the difference?
4. Check for uninvoiced orders or draft invoices
5. Look for refunds or credit notes that might explain gaps
6. Summarize findings and explain the discrepancy

Example 3: Root cause analysis

We have 15 products showing as "in stock" but customers are getting "out of stock" errors. Let's debug this step by step:

1. First, list the 15 products with their inventory levels

2. Check if there are any reserved quantities or pending deliveries

3. Verify the website inventory configuration for these products

4. Look for any inventory location restrictions

5. Check if virtual vs. physical stock settings are causing the issue

6. Explain what's causing the discrepancy and how to fix it

Example 4: Module implementation planning

I need to implement a new approval workflow for purchase orders over $10,000. Think through this step by step:

1. What Odoo modules are needed for this workflow?

2. What configuration changes are required?

3. Do I need custom code, or can this be done with standard features?

4. What user permissions need to be set up?

5. Create a step-by-step implementation checklist

Pro tip: Add "think step by step", "work through this systematically", or "let's investigate" to any complex query. The AI will break down the problem, show its reasoning, and give you a more thorough, accurate answer.