How to Use Guru to Answer Questions Inside Slack
Direct answer: To use Guru, connect Slack or Teams, create cards owned by a subject-matter expert, set a verification interval, and let the Slack app surface cards in the channel where the question is asked.
Every company has the same knowledge problem: the answer exists, it’s in a doc, the doc is two years old, and the person who wrote it has left. Guru’s bet is that the fix isn’t better storage — it’s putting an expiry date and an owner on every answer.
What is Guru and how is it different from a wiki?
Guru is a knowledge management tool made by Guru Technologies. The unit of content is a card: a short, focused answer to one question, with an owner, a verification date, and tags.
A wiki page can be any length and anyone can edit it. A Guru card is short by design, has one accountable owner, and goes stale loudly — the owner gets prompted when verification lapses. That’s the entire differentiation.
How do you set up Guru so people actually use it?
Three weeks, realistically, and the work isn’t technical.
- Connect Slack or Microsoft Teams from the Integrations page. Slack is the better experience; the Teams integration lags on search.
- Invite the first wave — 10-15 people, not the whole company. Pick one team with repetitive questions, usually support or sales.
- Import existing docs as drafts. Guru imports from Google Drive, Confluence, and Notion. Don’t import everything; import the top 40 documents by traffic.
- Assign an owner to each card. A card without an owner is a wiki page wearing a costume.
- Set verification intervals by content type. Pricing and policy: 30 days. Product how-tos: 90 days. Company info: 180 days.
- Turn on Slack auto-suggest so Guru responds with a card when someone asks a known question in a connected channel.
The first wave matters more than the plan. Rolling Guru out to 400 people at once produces 400 people who saw a notification and never came back.
How do Slack answers work in practice?
Two modes. A user types /guru and a query, and results appear privately. Or, in channels you enable, Guru watches for questions matching card titles and posts a suggestion.
Auto-suggest is where teams either love or hate the product. Configured well, it reduces “who knows?” pings noticeably. Configured carelessly, it becomes a bot that interjects in every conversation. Restrict it to two or three high-volume channels for the first month, watch the reactions, then expand.
What does Guru cost?
Pricing is per user per month and quoted, with public list prices that have hovered around $15-$18 per user for Standard and higher for Enterprise, billed annually. There’s no meaningful free tier beyond a trial.
The cost math is the reason most adoptions die. At 200 users, that’s roughly $36,000 per year for a knowledge base. It only pays if you can point to hours saved, so instrument that from day one: count questions answered by cards versus humans in Slack.
Which content belongs in Guru and which doesn’t?
| Content | Guru? | Why |
|---|---|---|
| “How do I issue a refund over $500?” | Yes | Repeated, bounded, verifiable |
| Pricing and discount policy | Yes | Changes often, must be current |
| Product documentation for customers | No | Belongs in a help center |
| Project plans and meeting notes | No | Not stable, not searchable by question |
| Onboarding checklists | Yes | Used repeatedly, must not drift |
The misfiled project plans are how Guru implementations decay. Once a knowledge base contains meeting notes, the AI search stops returning clean answers and trust drops.
What are Guru’s honest weaknesses?
Writing cards is tedious. A good card takes 10-15 minutes: title as a question, answer in three sentences, owner assigned, tags added. For 200 cards that’s a real project, and nobody wants to do it. Teams that skip the writing phase end up with imported Confluence pages and no verification, which is a wiki with extra cost.
Search across attached files is weaker than across cards. PDFs attached to a card are searchable in the app but not reliably in Slack suggestions.
And the AI answers, while grounded, sometimes surface an expired card when a current one exists on the same topic. Expired cards need archiving, not just flagging — nobody does this, and it’s the main source of wrong answers.
How do you keep it from rotting?
One rule: if a card fails verification twice in a row, the owner is asked to either update it or archive it. Deleting is a valid outcome. A knowledge base with 60 accurate cards beats one with 400 half-true ones.
Review the verification dashboard monthly. When more than 20% of cards are expired, the tool is lying to your team, and they will start asking a human instead — which is exactly the behavior you bought Guru to stop.
Quick Summary
- Guru’s unit is a card: short answer, one owner, verification interval.
- Roll out to 10-15 people first, usually support or sales.
- Set 30-day verification for pricing, 90 for how-tos, 180 for company info.
- Restrict Slack auto-suggest to two or three channels initially.
- Archive stale cards rather than flagging them forever — wrong answers cost more than missing ones.
