> ## Documentation Index > Fetch the complete documentation index at: https://docs.glood.ai/llms.txt > Use this file to discover all available pages before exploring further. # Using it > What to ask your assistant for, and how to read the answers it gives back ## The first thing to ask > "What Glood organizations can you see?" A good answer names each organization it can reach, the role you hold there, and the stores available to you. It also lists any organization you belong to that you have **not** shared with this connection, with a link to change that. Ask this first in a new conversation. It is also how you find out whether a problem is "the assistant cannot see that store" or "that store has no data". **When a phrase could mean two organizations, the assistant is instructed to ask rather than guess.** If it ever picks one silently, that is a bug worth reporting. ## What you can ask for | Ask | What happens | | ----------------------------------------------------------- | -------------------------------------------------------------------------------------------------------------------- | | "What Glood organizations can you see?" | Your organizations, your role in each, the stores you can reach, and anything not shared with this connection | | "Which reports can you run for acme-in.myshopify.com?" | The reports **that store** can answer, which depends on which Glood apps it has connected | | "How did conversion do last week on acme-in.myshopify.com?" | Runs the report and answers with the figures, the window actually covered, and any caveat the data carries | | "Revenue for the last 30 days, compared with the 30 before" | The same, with the comparison window computed for you | | "Which of my products convert best?" | Per-product and per-collection metrics from Glood's own indexes | | "What did this shopper do before they bought?" | One session's timeline, event by event | | "Which of these are out of stock?" | Read-only queries against your own Shopify admin: inventory, fulfilment, metafields | | "What are my stores, and which Glood apps are on them?" | Every store this connection reaches, with its apps and settings | | "What is the returns agent watching?" | Its instructions, its schedules and where it is deployed | | "What did it find last night?" | Its runs, the findings it raised with how often each recurred, and what it read to get there | | "Make me an agent that watches for stockouts" | Creates it with the instructions you describe. It runs nowhere until you deploy it to a store and give it a schedule | | "Deploy it to acme-in and run it daily at 9" | Puts it on that store and schedules it, in the store's own timezone | | "Pause it on acme-in" / "dismiss that finding" | Changes it, and reads back a sentence naming exactly what changed | | "What procedures have we written down?" | Your organization's skills, and the platform library beneath them | | "Who is in this organization?" | Members and their roles, and what is connected to the organization | That is the whole dashboard, minus the short list under [What it will not do](#what-it-will-not-do). **Roughly fifty reports are reachable this way**: revenue, conversion, AOV, attribution, funnels, top products, collection performance, returns, orders by channel or gateway or province, customer cohorts, data integrity. Ask what exists rather than guessing at a name. **The numbers are the same ones the dashboard shows.** The assistant runs the same code the screen does, so a figure in chat and a figure on the page cannot disagree. ## How to read the answers Every number comes back with **the report it came from and the window it covers**, and the assistant is instructed to quote both. A number without its window is a vibe with a decimal point. Some results carry a note, and those notes matter: | Note | What it means | | ------------------------------------ | ----------------------------------------------------------------------------------------- | | "the window was shortened" | Your data does not go back as far as you asked. The answer covers less than the question. | | "the top few, not everything" | Rows were capped. There is a tail you are not seeing. | | "estimated, not measured" | Exactly that. | | "unavailable rather than zero" | A share could not be computed. That is not the same as it being zero. | | "only one Glood app's data answered" | The store runs both apps but one index answered, so Glood's reach is understated. | The assistant is told to pass these on rather than quietly dropping them. If a number ever arrives bare, ask what window it covers. ## One organization at a time Asking about two organizations in one breath fails, deliberately. Currencies, timezones and roles differ between them, and a single table spanning two is a number that means nothing. Ask one organization at a time. ## Changes read themselves back Anything that changes an agent names what it changed — the agent, the store and the organization: > Paused Returns desk on acme-in.myshopify.com / Acme India That sentence is the only confirmation the right target was hit. Read it. Archiving an agent, stopping a deployment and removing a schedule go further: they ask you to confirm first and change nothing until you do. ## What it will not do A connection carries **your own role and nothing more**, in the organizations you ticked. On top of that, these have no tool at all: * Sign you in, or set up two-factor. * Create or delete an organization. * Create or change roles. * Write to Shopify. Reads only, always. * Change **which organizations it can reach**. That last one is deliberate and structural. With no tool for it, no instruction reaching your assistant — from a web page, a document, or anyone else — can widen its own access. Only you can, and only from the dashboard. Anything on that list, you do in the dashboard. For the connection's own reach, that is [app.glood.ai/account#connections](https://app.glood.ai/account#connections).