Best practices for writing effective prompts

Updated Friday, March 06, 2026

When using the chat panel to generate logical models, well‑structured prompts help ensure accurate and relevant outputs. Effective prompts ensure that the system produces accurate logical models, and business‑specific structures.

Follow these best practices when writing prompts to generate or refine models.

  1. Specify the business domain clearly

    Start by stating the industry or domain your model belongs to. For example, Banking, Healthcare, Logistics and so on.

    The application interprets domain context to determine terminology, regulatory needs, and data types. A clear domain prevents the system from generating generic or incorrect structures.

  2. Identify the functional area within the domain

    Most domains contain multiple specialized areas. Naming the correct area helps generate accurate entities and relationships. For example,

    Banking: Payments, lending, fraud detection, customer onboarding

    Healthcare: Patient admissions, patient visits, diagnostics and imaging, lab services, billing and insurance claim

    Providing this information ensures your prompt reflects real world processes within that domain.

  3. Define the problem you want to solve

    The system needs to understand what you are analyzing, optimizing, or modeling.

    Avoid vague goals and instead describe the business question or workflow. For example,

    • Identify delays in the clinical trial recruitment process.

    • Analyze order return patterns for high-value customers.

    • Track payment failures across digital channels.

    This helps the model prioritize the correct entities and relationships.

  4. List the key entities you need to track

    Entities form the foundation of any logical model. Include major objects involved in your process. For example, Customers, Patient admissions, Patient visits, Billing and insurance claims, Orders, Products, Invoices, Patients, Clinical trial participants and so on.

    The application can generate entities based on the domain area, but specifying a few of them helps produce more accurate fields, relationships, and model structures.

  5. Use direct, structured language

    Write prompts in short sentences and avoid unnecessary jargon. Focus on the domain, the area, the problem, and the entities.

    A structured prompt makes the output more predictable and aligned with your intent.

  6. Sample well‑structured prompt

    Prompts that combine multiple use cases such as, “Create a model for order management and product returns and customer loyalty”, can affect the quality of model generated.

    Keep the prompt focused on a single workflow or outcome.

    For example,

    Create a model for the banking payments system. The goal is to understand why card transactions fail across different channels. The key entities are Customer, Card, Transaction, and Merchant.