Use case
Customer support
A customer on the “Business” plan asks how quickly they must get a response. The agent remembers “something like 8 hours” from the internal FAQ, the policy is in one folder, the contract in another, and the manager will answer in the evening. Every such question ends in either a mistake or a wait.
← scroll the frame →
Situation
Three documents, three different answers
What gets in the way
- Customer terms are written in contracts and their annexes, procedures in the policy, reply templates in a third file.
- The internal FAQ lags behind the policy, and nobody knows which document is right.
- The answer a customer gets depends on who is on shift today and what they remember.
- Nobody counts repeat requests about the same bug until a customer leaves.
- Fix dates must not be promised — and what may be promised is written in a policy nobody scrolled down to.
Solution 01
An answer to the customer — by the policy, with a ready template
An agent asks: “A customer on the ‘Business’ plan wants a refund 20 days after paying. What do the policy and the FAQ say, and what should we tell the customer?” The AI agent reads the support policy, the internal FAQ and the reply templates, gives 14 days under the policy against 30 in the outdated FAQ, explains why the policy is right, and provides a ready refusal text offering to carry over the rest of the period.
How to ask a good question about documents← scroll the frame →
Solution 02
Repeat requests and what we answer
The support manager asks whether there were repeat requests about the same bug in September, from whom, and what we tell customers. The AI agent reads the request register, the support report and the reply templates and answers with a table: three requests, two from the same customer — and provides a ready template with a workaround; under the policy, fix dates are not promised to customers.
About “Ask”← scroll the frame →
Solution 03
Ask where people already work
An agent asks right from the ticket they are working on — and the answer arrives as a comment in the same ticket. A comment on a Notion or Confluence page, in a Google Doc, on a Figma design or in a tracker ticket that mentions the AI agent is a question, and the answer arrives in the same discussion. Research is ordered the same way: the finished document appears next to the original. There is no need to open the app.
In the app
The “Ask” section of the space: question, answer and follow-ups, like a conversation; the price of the question is shown right after the answer.
In a comment on a document
Notion, Confluence, a Google Doc or a Figma design: a mention of the AI agent in a comment is a question, and the answer arrives in the same thread.
In a tracker ticket
Jira, Redmine and others: a question as a comment, an answer as a comment; research arrives as an attached document.
Solution 04
Support rules in the knowledge base, not in the shift’s heads
The rule “quote deadlines to customers only from the contract” and the pitfall “the FAQ lags behind the documents” are written in the space’s knowledge base once. The AI agent sees them in every answer — which is why it did not trust the FAQ in the examples above. A person writes the rules; the AI agent adds facts and pitfalls itself, marked so they can be checked.
What to put in the knowledge base← scroll the frame →
Solution 05
What your IT specialist does — once
Installs the connector in the company network, picks a connection from the catalogue — Notion, Confluence, Google Drive, folders on a server, a tracker — names the department’s space and marks which sections and folders to read. Systems outside the catalogue — helpdesk, messengers, telephony — are connected with a custom tool: separate IT work following our documentation, after which the system is available to the AI agent just like the catalogue. From then on the department works on its own; a file that is not in the sources is simply attached to the question.
The connector in your network
A small program, outbound connections only by default; installed once for the whole organisation.
Connection to the documents
From the catalogue or a custom tool; an account that can read and comment, with the marked sections and folders.
The department’s space
Access is granted per space: the support space has no access to the HR documents until the owner grants it.
Solution 06
What the AI agent does not do
The limits are stated up front, so the department does not wait for what will not come.
Does not write to the customer
the agent gets the answer and the template; a person sends it.
Does not change tickets
does not close, reassign or reprioritise them.
Helpdesk and messengers
not in the catalogue yet; they are connected with a custom tool.
Outcome
Agents answer customers from the contract and the policy in minutes, the manager sees repeat requests from the documents, and discrepancies between the policy and the FAQ are found before a customer finds them.
The first step is a name
Create a space and ask your first question
Sign up by email and the space is ready at once. Until your IT specialist connects the documents, the AI agent answers from the knowledge base and from files you attach to the question. A starting balance is provided to try it out.
Until documents are connected, attach a file to the question