> ## Documentation Index > Fetch the complete documentation index at: https://help.aibuku.com/llms.txt > Use this file to discover all available pages before exploring further. # Prompts and runs > Describe your customers and the questions they ask, then test how AI assistants answer them. Prompts and Runs are part of [Visibility](/visibility/overview), which some plans include. ## Customers and prompts A **customer** is a type of person who buys from you, such as *a busy parent*. A **prompt** is one thing that customer might say to an AI assistant, such as *"Can you help me find a salon nearby that is open on Saturday mornings?"*. Open **Visibility** → **Prompts**. Your customers are on the left, their prompts on the right. The Prompts page: one customer, Busy parent, with five prompts, each with its level. ### Add a customer Choose **New customer** and answer three questions: * **What would you call them?** — a short name, such as *Busy parent*. * **What's going on for them?** — their situation. * **What do they want from you?** — what they are looking for. You can also name things they are after, the language they speak and where they are. Then choose **Add customer**. Not sure who your customers are? **Suggest** proposes some for you. Your plan sets how many customers you can have, and how many prompts each one gets. ### Add prompts Choose **New prompts**. Then: * **Suggest a set for me** — Aibuku writes a spread of prompts for the customers you pick, from vague to specific. Choose **Write up to 5 prompts**. * **Write one myself** — type something a customer has actually asked you. You pick who says it, how plainly they ask, and which of your services should answer. ### How plainly the customer asks Every prompt has a level. It shows how directly the customer asks for what you sell: | Level | What the customer does | | - | - | | **Just chatting** | Talks about their day. They don't ask for anything yet. | | **Hinting at it** | Has a need, but does not say what they want. | | **Says what they need** | Says what they need. | | **Asks for it** | Asks for your kind of product or service. | | **Asks for you by name** | Names your business. | A mix of levels gives the most honest picture. Being found when a customer asks for you **by name** is easy. Being found when they are only **hinting at it** is hard — and valuable. ## Runs A **run** is one test. Aibuku asks AI apps your prompts and records how far each one got. Open **Visibility** → **Runs**. The Runs page: one row per AI app and part of the run, with the verdict, the prompt level and the status. ### Start a run Choose **Start a run**. In **Start a run?**, pick the AI apps to ask this time, then choose **Start the run**. A run takes a few minutes. Your plan includes a number of runs per day, shared by all your projects. When they are used up, Runs says when they reset — for example *No evaluation runs left today … resets 2 Oct, 02:00*. ### Run every day Projects are checked once a day by themselves. The **Daily check** button (the calendar icon next to the filters) turns this on or off. When it is on, it says **Daily check: on.** ### Read the results Each row is one AI app's result for one part of a run: * **App** — which AI app, and which part: **Finding your website** or **Using your connector** (**Using your MCP server** if your project tests your own MCP server). * **Chats** — how many conversations were tested. * **Verdict** — how far the app got. For finding your website, the steps are **Found**, **Read** and **Pointed**; a result can also be **Never found**. For using your connector, **Never shown** means the app did not offer your business at all, and **All 4 steps** means it went the whole way. * **Prompt level** — how plainly customers had to ask. * **Status** — whether the run is still going or **Done**. Use the filters — **Any app**, **Any time**, **Any verdict**, **Any customer**, **Any level** — to narrow the list. Open a row to read the actual conversations. Need help on the page itself? Choose the info button for **How runs work**. If a run shows **Didn't finish**, something broke on our side part way through. It is not your fault. If it shows **Not run**, the run would have cost more than your plan allows, so we stopped before spending anything.