How to ask a good question

Muninnbase is built to be talked to the way you'd talk to a coworker. There's no syntax to learn, no special operators, no keywords you have to chant. But the way you phrase a question still shapes the answer you get back. A clear question gets a clear answer. A vague question gets a vague one.

Plain English works

Type the question the way you'd say it out loud. "How many vacation days do I get in my first year?" works exactly like "what are the first-year PTO accrual rules?" works exactly like "first year vacation policy." The system reads natural language and matches it to your company's documents.

You don't need to capitalize anything. You don't need to end with a question mark. You don't need to phrase it as a question at all. "expense reimbursement process" is a fine question. So is "tell me how to expense a client dinner."

Specific beats vague

Vague questions return vague answers. Specific questions return specific ones.

• "PTO?" gets a generic summary. "How many vacation days do I get in my first year?" gets the actual policy with the numbers in it.
• "expenses" goes nowhere useful. "What's the process for getting reimbursed for a client lunch?" returns the SOP, step by step.
• "benefits" gets a list. "When am I eligible to enroll in the dental plan?" gets the date and the rules around it.

The pattern is the same every time: name the thing, name the situation, ask the question.

Each question stands on its own

Worth knowing up front, because most chatbots work the other way. Muninnbase does not carry context from one question to the next. If you ask "How does PTO accrual work?" and then ask "what about for managers," the second question is read on its own as "what about for managers," with no memory of PTO. The result is usually a confused or off-topic answer.

The fix is to include the topic in every question. "How does PTO accrual work for managers?" gets the answer you wanted. This is a deliberate design choice, explained more in understanding answers and source citations.

If the first answer misses

You have two moves. Either rephrase, or get more specific.

Rephrase if you suspect the system did not understand what you meant. "Try 'when do I become eligible for the dental plan' instead of 'dental.'" Same question, different words; the new phrasing often surfaces the right chunk of the document.

Get more specific if the answer was technically correct but assumed the wrong situation. "Add 'as a part-time employee' if the original answer assumed full-time." That kind of context narrows the retrieval to the section that actually applies to you.

If neither move works, the system probably does not have the answer in its documents. That's what the thumbs-down button is for, covered in the next section over.

A note on privacy

Questions are logged so your admins can see what the team is asking and improve the knowledge base over time. The log is anonymous. It stores the question, not who asked it, so no one is going to see "Sarah asked about salary increases." Your admin sees the question. They do not see your name attached.

Even so, the log is still a permanent record. Don't type confidential personal information into the chat. Your social security number, your bank account, your salary, your medical history. The chat is not the right place for any of that, and it never will be. Treat the chat like a question you'd ask in front of the room, because in a sense it is one.

What to do next

Once you've asked a good question, the next thing to understand is how to read the answer. Understanding answers and source citations explains what the source pills mean and how to verify the answer against the original document.