You stop answering the same questions by writing each answer down once, in a place people check before they ask, and making that place easier to use than messaging you. Track which questions come up, turn the frequent ones into short pages, link those pages to the work they belong to, and reply to repeat questions with a link instead of a new explanation. An AI that reads those pages can then answer for you.
Why the same questions keep coming
If you lead a team, you are probably the default search engine. "How do we price this?" "Which template do we use for clients?" "Did we ever decide on the delivery terms?" Each question takes two minutes. Twenty a day take your morning.
The numbers show how much time this costs on both sides:
- In a 2018 Panopto survey of 1,000 U.S. workers, knowledge workers spent 5.3 hours a week waiting for information from colleagues or recreating knowledge that already existed. 60% said it was difficult to almost impossible to get the information they needed from coworkers (Panopto, via HR Dive, 2018).
- The same study found that 42% of institutional knowledge is unique to the person in the role and not shared with coworkers (Panopto, via HR Dive, 2018).
- In Microsoft's 2023 survey, 62% of people said they spend too much time searching for information during the workday (Microsoft Work Trend Index, 2023).
- In Atlassian's 2024 survey, 55% of knowledge workers said they find it hard to track down information, even though they know many people at work (Atlassian State of Teams, 2024).
- Atlassian's 2025 survey of 12,000 knowledge workers and 200 executives found that leaders and teams lose 25% of their time just searching for answers (Atlassian State of Teams, 2025).
People ask you because asking is the fastest route they know. Your job is to make a faster one.
Step 1: Track the questions for two weeks
Before writing anything, keep a simple list. Every time someone asks you something, add one line: the question and who asked. After two weeks, group them. Most leads find that a handful of topics cover most of the questions:
- how a recurring process works (pricing, onboarding a client, approvals),
- where something is (templates, files, logins, contacts),
- what was decided and why,
- who owns what.
Start with the topics that came up most often. Ten pages that answer real questions beat a hundred pages nobody reads.
Step 2: Write short answer pages
Each page answers one question. Keep the format the same:
| Part | What goes in it |
|---|---|
| Title | The question, in the words people actually use |
| Answer | Two to five sentences, or a short checklist |
| Owner | Who keeps this page correct |
| Last checked | A date, so readers can trust it |
| Related | Links to the process, project or customer it belongs to |
Write for the person asking, not for an archive. If you would explain it in chat in four sentences, the page should be about four sentences.
Step 3: Put answers where the work is
A knowledge base that lives in a separate tool nobody opens will not help. Link answer pages to the projects, customers and tasks they apply to. When someone works on a client project, the client's rules should be one click away. If you want to go further, How to Get Knowledge Out of Employees' Heads covers capturing what only certain people know.
Step 4: Answer with a link, every time
This is the habit that makes it stick. When a question comes in that has a page, reply with the link and nothing else, or one line plus the link. When a question comes in that has no page, answer it, then write the page and share the link. After a few weeks, people check first.
Two rules help:
- No answer only in a private message. If it's worth answering, it's worth one page or one comment on the shared task.
- Decisions get written down where they apply. "We agreed to 30-day payment terms for X" goes on X's page, not in a meeting memory.
Step 5: Let an AI answer from your pages
Once the answers exist in one place, an AI assistant can read them and answer questions directly, with a link to the source. This is the biggest step change, because people can ask in plain language at any time without waiting for you.
It only works if the AI reads your company's knowledge, not just general training data. A general chatbot will give a plausible answer that ignores your actual prices and rules. See How to Give ChatGPT or Claude Access to Company Knowledge for the options.
Keep it current
Answers go stale. A wrong page is worse than no page, because people stop trusting all of them.
- Give each page an owner.
- When a process changes, the person who changed it updates the page.
- Review the most-used pages once a quarter.
- When someone points out a wrong answer, fix it the same day.
How Orakis does it
Orakis has a Knowledge area where processes, decisions, customers and people are linked pages, with a graph of how they connect. Alfred, the built-in AI agent, reads every page, so team members can ask questions and get answers from the company's own knowledge. Because goals, tasks and team chat live in the same workspace, the answers can also cover current work, such as who owns a project. Through MCP, Claude, ChatGPT and other assistants can connect to the same workspace. Orakis is a young product in closed beta.
FAQ
How do I get people to check the documentation before asking? Reply to repeat questions with a link, consistently. Once people see that the answer is always a link, checking first becomes faster than asking.
What should I document first? The questions you were asked most in the last two weeks. Track them before you write anything.
Can AI answer my team's questions instead of me? Yes, if it can read your company's own pages and current work. Keep a person responsible for the content, and let the AI point to the page it used so people can check the source.