An AI internal knowledge base agent gathers in one place what your store knows: the return policy, courier rules, supplier lead times, product specifications, decisions made in the past. Anyone on the team can ask it and get the answer with the source in plain view. A person approves what becomes an official rule.


In a small store, the answers live in two people’s heads and in dozens of scattered places. A supplier’s lead time is in an email from three months ago. The return exception for opened products is in a WhatsApp message. A product’s real specifications are in a PDF from the importer. When the person who knows is on vacation, the team guesses.
Gathers internal documents from Google Drive, Notion, email and team conversations and makes them searchable by question, not by file name
Keeps the store’s policies current: returns, warranty, delivery, exchanges, with the version in force and the date of the last change
Collects supplier data: lead times, minimum orders, contact people, terms for returning goods to them
Links the technical specifications and data sheets received from manufacturers to the products in the catalog
Answers “what do we do if” questions, with an exact pointer to the document or message the answer comes from
The difference from a well-organized folder is that you do not need to know where to look. You ask how to handle a parcel refused on delivery and you get the procedure, plus the last similar case and how it was resolved.
The AI agent also keeps a log of decisions. When the team decides something, for example that electronics are only taken back sealed or that a supplier is no longer used for urgent orders, the decision goes into the knowledge base with the date and the name of whoever made it. A few months later, when someone asks why things are done this way, the answer exists and does not depend on anyone’s memory.
Tell us what slows you down. We talk, then decide together whether and how we work.

On Monday, your support colleague gets a question about the warranty on an imported product. Instead of digging through emails, she asks the AI agent. She gets the warranty period from the manufacturer’s data sheet, the steps to follow and the name of the supplier who handles claims. She writes the reply to the customer herself.
On Wednesday, the courier changes its rules for bulky parcels. Someone uploads the notice. The AI agent notices that the internal shipping procedure says something else and flags the conflict, with the old document and the new one side by side. Until someone confirms the new rule, both stay visible, with a marker.
On Friday you get the list of questions the AI agent could not answer that week. These are the gaps in the store’s knowledge, exactly the things only one person knows or nobody knows. You fill them in once and they stop coming up.
A new hire in the warehouse or in support no longer learns by asking colleagues something every ten minutes. They ask the AI agent, and for anything not documented, the AI agent points them to the right person and logs the question so it gets filled in.

Google Drive, Google Docs and Notion, where procedures and policies usually live
Gmail or Outlook, for agreements with suppliers and couriers, only from the mailboxes you choose
Slack or work group chats, wherever the team makes decisions in chat
The store platform, whether Shopify, WooCommerce, PrestaShop or Gomag, for the catalog, attributes and published policies
easySales, for multichannel order processing rules, and SmartBill or Oblio for suppliers and documents
The public pages of the couriers you work with, from Fan Courier and Sameday to Cargus, for their current terms
The AI agent reads these sources with the permissions you give it. If a folder with contracts or payroll should not be visible to the whole team, it stays out of the knowledge base the whole team can ask.

It does not invent procedures where none has been written, it says one is missing and who might know
It does not change the policies published on your site
It does not answer customers directly from the internal knowledge base, unless you also have support capacity set up separately
It does not show sensitive information to anyone without access to it
Nor does it treat an old message as a rule in force. When the only source for an answer is a conversation from a year ago, the AI agent says so and shows you the date, so you can judge whether it still holds.

You start with the area that generates the most repeated questions, usually returns and delivery. You show us where the documents are and who owns them. The AI agent first takes an inventory: what exists, what contradicts what, what is out of date.
The owners confirm the correct versions, and then the team starts asking. In the first weeks the list of gaps builds up and gets filled in. Once returns and delivery work well, you add suppliers, then product specifications. Each new area comes in with its owner, so the knowledge base does not turn into yet another folder nobody maintains.
A person on the team owns each area of the knowledge base: someone for return and warranty policies, someone else for suppliers and logistics. The AI agent proposes updates when it finds new information or a contradiction, but a rule becomes official only after its owner confirms it.
A person also decides who sees what. Margins, purchase costs and the terms negotiated with suppliers stay accessible only to those who need them. And any text that goes from the knowledge base to customers, such as a new return policy page, is reviewed before it is published, because consumer protection law matters there too.
The most important link on the page. From here you go up to the agent we sell.
Sideways, on the same level, so you can move easily between pages.
It stays in your tools, and the AI agent reads it with the permissions you give it. We agree together which sources go in and which do not. For sensitive information, such as supplier contracts or purchase costs, access is limited to the people who need it. You see the cost of the tools used for storage and search before signing.
The AI agent does not pick a version on its own. It shows you both documents, with their dates, and notifies the owner of that area. Until it is confirmed, any answer on that topic comes with a marker saying there is a contradiction. That way nobody on the team applies, without knowing it, a rule that has since changed.
No. The AI agent works with what exists, including emails, messages and documents written in a hurry. It takes the inventory, shows what contradicts what and what is missing, and you confirm the correct version. New procedures only appear where nothing existed, and the AI agent can prepare a first draft for the person who knows how it is done to correct.
Yes, if you also have customer support capacity with us, but with a clear separation. Customers only get information marked as public, for example the return policy in force. Internal notes, margins and discussions with suppliers never end up in a reply to a customer. We set up the separation during configuration and you check it before launch.
The first area, usually returns and delivery, can be used once the AI agent has taken the inventory and the owners have confirmed the correct versions. How long that takes depends more on how quickly you confirm than on the AI agent. The other areas are added one at a time, and the knowledge base is useful from the first area, not only once it is complete.
Tell us what slows you down. We talk, then decide together whether and how we work.