What it is
When you create a company in ōrbita, a few things turn an account into content worth publishing: your website, some questions about the business (what you sell, to whom, what makes you different), a communication plan you have reviewed, two connected networks and the first month of content. The activation guide on the dashboard walks you through those steps.
If you already use Claude or ChatGPT, you can do those steps in a conversation with your assistant. It reads your company’s guide, confirms with you which company it is talking about, tells you where its data comes from and takes you step by step: it shows you what it has found, asks you only for what is missing and saves once you say it is right. It works with the same ōrbita connector you use for everything else and is included in every plan.
How it works
- Create the company. This is the one step a person always does: you type the company name and, if we do not know it yet, your own. A new company counts towards your plan, which is why the AI does not create it.
- Choose “Connect my AI and let it guide me step by step”. Right after creating it, ōrbita asks how you want to complete it: with your AI or filling it in yourself (about 10 minutes of questions). You can always change your mind from the dashboard guide.
- Connect your AI to ōrbita. The screen shows the connector address, the same one you see in Connections. When you authorise it you will see the optional permissions, already ticked: leave them like that so your AI can complete every step. If your AI is already connected, nothing else is needed. The steps per assistant are in Connect your AI.
- Paste the text into a new chat. The screen gives you a text ready to copy. Open a new chat in your assistant, turn ōrbita on in it and paste the text. It is in Spanish because the app is; this is what it says, with your company name already in it:
Help me set up {company} in ōrbita with the ōrbita connector, step by step. Start with get_company_guide and follow its script (chat). Before saving anything, confirm the company with me and tell me which sources you have about it (what I tell you, attachments, my website or other data within your reach); do not use data from another company and do not make anything up. Show me what you find with its source and ask me only for what is missing. At every step give me its link to ōrbita and wait until I say “done” or “skip”. I confirm the plan myself.
- Confirm the company and the sources. Before saving anything, your AI tells you which company it is going to set up and which sources it has about it: what you tell it, what you attach in the chat, your website or other data within its reach. If it has none, it says so. Confirm the company to go on.
- Follow the steps in the chat. At every step your AI explains what comes next, gives you the link to that step in ōrbita and waits for you to say “done” (or “skip”, on the steps that can be left for later). The order is: your website, the questions about the business, the catalogue and the questions your customers ask you; then the plan, and with it the key dates of the year. Meanwhile, the screen in ōrbita updates on its own and ticks what the two of you complete. You can close it: the dashboard guide ticks the same things.
- Request and confirm the plan. Once you have reviewed the catalogue and the questions your customers ask, your AI offers to ask ōrbita for the communication plan, which is written from them and takes a couple of minutes. Your AI cannot tell you when it is finished: write “check” to it after that time. When it is written it offers you two ways: you review and confirm it in ōrbita (the “Confirm plan” button in Strategy), or it shows it to you in the chat, you ask for the changes you want and it confirms it when you say “I confirm”. Your AI never confirms the plan unless you tell it to in that chat. With the plan, ōrbita proposes the key dates of the year: you go through them right after.
- Connect two networks. You do this yourself in ōrbita, in Connections, because each network asks you to sign in with your account. Your AI sees which networks are connected and how many are missing, and gives you the link. You can leave it for later.
- Request the first month of content. With the plan confirmed, your AI offers to request it, or you press the button in the guide. ōrbita writes the first month’s posts on its servers; it takes a few minutes and leaves them as drafts in the calendar, without publishing anything.
- Schedule the first post. Review the drafts in the calendar, or ask your AI to show them to you. Pick one and leave it scheduled: in ōrbita or, if your AI has the scheduling permission, from the chat, where it first shows you the exact text, the network and the date and asks for your confirmation. With that, the company is up and running.
Once the guide is complete, your AI suggests, one at a time and with no obligation, what you have not tried yet in ōrbita: inviting your team, creating a post of your own, tasks, the interactions inbox, metrics, reviews, Canva or the mobile app.
What your AI does and what you do
| Step | Who | How |
|---|---|---|
| Create the company | You | In ōrbita. Counts towards your plan. |
| Choose AI or by hand | You | Screen after creating the company; can be changed from the guide. |
| Confirm the company and the sources | You | Your AI asks you before saving anything. |
| Your website | Your AI with you | It asks you for it and saves it exactly as you give it, or states there is no website. ōrbita reads the public content of the website to get to know the business. |
| The questions about the business | Your AI with you | It gathers what it knows about the company and shows you a summary in three parts: what it has found, with its source; what it needs you to confirm; and what is missing. It saves once you say it is right and asks you for the rest little by little. It also proposes the content language and how to address your audience. |
| Catalogue and customer questions (before the plan), key dates (after) | Your AI with you | It shows you what is there and adds, changes or removes what you tell it. At least one product or service is needed. It does not make up prices, deadlines or conditions. |
| Review and confirm the plan | You, always | In ōrbita or in the chat; your AI only confirms it when you say so. |
| Connect 2 networks | You | In Connections; each network asks for your account. |
| The first month of content | Your AI (or you) | Only once the plan is confirmed. |
| Schedule the first post | You, or your AI with your confirmation | See Connect your AI. |
Everything your AI does is signed in ōrbita as “{your name}’s assistant” and you see it in the guide just as if you had done it yourself.
Which sources your AI can use
Any source with data about your company: what you tell it in the chat, the website (the address is enough), documents you attach (the catalogue or menu, a brochure, a sales deck), social media texts that already work for you, or data your assistant already has within its reach, such as a Drive document, a knowledge base or your history with it. You do not need to attach anything: if you have no documents, your AI asks you.
What matters is that the data is about that company. Your AI tells you where each piece of data comes from and saves nothing until you say yes. It does not use data from another company or guesses; that is the rule carried by the text you paste and by the guide itself.
One tip: also tell it three or four questions your customers really ask you (opening hours, shipping, how to book). ōrbita uses them for the content and needs them to write the first month.
In Claude and in ChatGPT
The rules (confirming the company, naming the sources, waiting for your “done”, never confirming the plan for you) travel in the guide the connector returns to your assistant at every step, so they work the same in Claude and in ChatGPT.
If ChatGPT does not see the setup tools, delete the connector and create it again: it stores the list of functions from the day you added it (details).
If something does not work
- Your AI says it has no permission. It tells you which permission it lacks and gives you the link. In Connections > Connected AIs, press “Give permission” on that AI: there is no need to connect it again. You can also do that step yourself in ōrbita.
- Your AI takes for granted something you did not tell it. Ask it for the source of each piece of data: nothing is saved until you say yes. If something is already saved, ask it to correct it; it first tells you what it is going to overwrite.
- The screen does not tick a step. It refreshes every few seconds with what is already saved. Ask your AI where it is; every step that is finished shows up in the dashboard guide.
- The plan is not requested. ōrbita only writes the plan once there are enough answers (a meaningful description and at least one of: what makes you different, ideal customer or goal; two if there is no website) and you have reviewed the catalogue and the questions your customers ask. Your AI sees in the guide what is missing and asks you for it.
- The first month is not requested. It needs the confirmed plan, at least one product or service and one customer question. The guide tells your AI.
- You would rather carry on yourself. Press “I’d rather do it myself” or open any step from the dashboard guide. Whatever is already saved is kept.
Privacy
What you tell your assistant and the documents you hand to it are processed by your assistant’s provider (Anthropic, OpenAI) under your contract with it; ōrbita only receives what your AI writes into your account through the connector, with your identity and your permissions. If you set the website, ōrbita reads its public content to pre-fill answers, just as in the manual onboarding. ōrbita does not see your conversation. More in the privacy policy.