> ## Documentation Index > Fetch the complete documentation index at: https://docs.onsomble.ai/llms.txt > Use this file to discover all available pages before exploring further. # Choose prompts by customer intent > Use Understand, Evaluate, and Act to choose prompts that represent meaningfully different customer needs. **Customer intent** is the help someone wants from an AI assistant. It explains the goal behind a question, not only the topic being discussed. Customers can ask about the same product while seeking different kinds of help: * **Understand**: "What does business interruption insurance cover?" * **Evaluate**: "Is business interruption insurance suitable for a small retailer?" * **Act**: "How do I add business interruption cover to my policy?" Use customer intent to decide whether each prompt represents a distinct need. A new prompt adds useful coverage when it asks for different help or includes context that could materially change the answer. Understand, Evaluate, and Act are practical review tools. They are not fields in the Prompt Library, and one question can involve more than one customer job. ## Use three practical customer jobs ### Understand The customer wants facts, an explanation, a diagnosis, or a clearer view of requirements. Common prompt shapes include: * What is ...? * Why does ... happen? * What should I know about ...? * What features or requirements matter for ...? A useful answer gives the customer clearer knowledge. Prompts with an **Understand** job often appear during problem discovery, solution exploration, and requirements building. They can also appear later when a customer needs to clarify a detail. ### Evaluate The customer wants advice that helps them compare, choose, validate, or reduce risk. Common prompt shapes include: * What is the best ... for ...? * Which option is better for ...? * How do these providers compare? * Is this product reliable or worth it? A useful answer gives the customer greater confidence in a choice. Prompts with an **Evaluate** job show how AI assistants respond when customers compare, validate, or choose between options, including whether they mention your business, a competitor, or neither. ### Act The customer wants a concrete next step, instruction, service, or outcome. Common prompt shapes include: * How do I get a quote for ...? * Where can I book or buy ...? * How do I set up or use ...? * What should I do next if ...? A useful answer gives the customer a concrete next step. Prompts with an **Act** job often appear during decision and post-purchase support. ## Classify by the primary help sought A prompt can involve more than one customer job. Classify it by what would make the answer most useful: * clearer knowledge points to **Understand**; * greater confidence in a choice points to **Evaluate**; * a concrete next step points to **Act**. For example, "What should I look for in a commercial insurance broker?" may teach the customer about important criteria, but its primary job is **Evaluate** when those criteria will help them choose a provider. ### Follow one customer situation Hold the coverage context steady, then change the help being sought: | Coverage decision | Selection | | - | - | | **Product** | Commercial insurance | | **Persona** | Two-store retailer | | **Journey stage** | Decision | | **Region** | Auckland | The same situation can support three distinct prompts: | Customer job | Prompt | What a useful answer provides | | - | - | - | | **Understand** | "What information does a two-store retailer need before requesting a commercial insurance quote in Auckland?" | A clear list of requirements | | **Evaluate** | "Which commercial insurance broker is best suited to a two-store retailer in Auckland?" | Guidance for choosing a provider | | **Act** | "How can a two-store retailer get a commercial insurance quote in Auckland?" | A practical next step | These prompts share a Product, Persona, Journey stage, and Region, but each adds a different customer job. See [Build a prompt strategy](/prompts/prompt-strategy) for how those coverage dimensions fit together. ## Intent and journey stage are different Journey stage describes the decision context. Intent describes the immediate job the customer wants the AI assistant to perform. The relationships below are common patterns, not fixed rules. A customer can seek understanding late in a decision or ask for action-oriented help early. | Journey stage | Common customer jobs | | - | - | | Problem discovery | Understand | | Solution exploration | Understand or Evaluate | | Requirements building | Understand or Evaluate | | Product comparison | Evaluate | | Validation | Evaluate | | Decision | Evaluate or Act | | Post-purchase support | Understand, Evaluate, or Act | ## Intent and category are different Onsomble assigns each finished prompt a **category**: Comparison, Recommendation, Specific, or General. Category describes the kind of question that was written. Intent describes why the customer is asking it. For example, two Specific prompts can have different customer jobs: * "What does the low-stock alert feature in \[product name] do?" has an **Understand** job. * "How do I configure low-stock alerts in \[product name]?" has an **Act** job. Use category to organise and filter prompts. Use intent to judge whether the set represents the customer decisions you care about. ## Use intent to test distinct coverage Read prompts by customer job and context, not wording alone. These two prompts both ask an AI assistant to help the same customer evaluate the same kind of product: * "What is the best accounting platform for a small agency?" * "Which accounting software is best for small agencies?" Keeping both adds wording volume but little new coverage. A prompt such as "How do I move client invoices to a new accounting platform?" adds an **Act** job and a different customer situation. Your intent coverage is ready for matrix review when: * you can state the primary customer job behind each prompt; * prompts with similar wording represent materially different jobs or contexts; * minor paraphrases have been removed; and * the remaining prompts reflect the customer decisions in your measurement objective. ## What's next Apply the framework through Products, Personas, journey stages, and regions. Write a distinct customer question and attach its coverage details. This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.