How do you create a knowledge base with AI?

Most companies already have a knowledge base. It is just scattered across sent emails, a few outdated docs, and the heads of the two people who answer everything. We built Cassie, Sintra's  AI customer service agent, to pull it into one place: hand her the answers you already give and she turns them into structured FAQs and help articles, then stores them in Brain AI so your whole team, human and AI, answers from the same source. Write it once, and every answer after that has somewhere to come from.

Time to task completion: minutes to a knowledge base structure and your first articles.

What Cassie covers: FAQ and help articles · Knowledge base structure · Brain AI storage · Grounded support replies · Gap analysis

Pricing: Cassie is included with Sintra X at $15.60/mo billed yearly, with a 14-day money-back guarantee.

Start building your knowledge base →

What can Cassie do for building a knowledge base?

Cassie is one of Sintra's AI employees, and we built her for the part of support that never ends: turning what you already know into documentation people can actually find. Here is what she does.

Draft FAQs and help articles. Cassie generates FAQs, manuals, and helpful documentation content from the material you give her, turning scattered answers into articles customers can find and follow.

Structure the knowledge base. Give Cassie your list of common questions and topics, and she organizes them into categories and a prioritised article plan, so building a knowledge base starts with a map instead of a blank page.

Store everything in Brain AI. Brain AI is Sintra's central knowledge base: it holds your brand facts, files, documents, and uploaded web pages, and every helper uses it automatically. Load your articles once and the whole  AI team works from them.

Answer support questions from your knowledge. With the Brain AI knowledge base expanded, Cassie crafts accurate, on-policy responses to customer questions. You review and send them, and every reply stays consistent with what your knowledge base says.

Find the gaps. Cassie is trained on customer support best practices and can review your setup, flag missing topics, and, if you give her customer feedback data, generate a report showing where customers struggle and what to document next.

Setting up Cassie to create your knowledge base

The raw material matters more than the tooling, so gather it first.

1. Sign up and open Cassie. Create your Sintra account and select Cassie, the AI customer service agent, from the Helpers list.
2. Collect what you already have. Support emails, the questions customers ask over and over, product descriptions, policies, troubleshooting notes. Paste or upload it all directly in the Cassie chat. Messy is fine; that is what she is for.
3. Set up Brain AI. Add your brand basics, tone of voice, and product or service details to  Brain AI. This is the ai knowledge base everything will live in, so the articles you create come back here and every helper answers from them.
4. Decide who the knowledge base is for. Customers on your help center, your internal team, or both. It changes the tone, the depth, and where the finished articles get published.

Start your 14-day free trial →

How to create a knowledge base with AI in 7 steps

Run these seven steps in order, from scattered answers to a knowledge base your team and your AI both answer from. Each one builds on the last, and in our experience the reason most knowledge bases fail is skipping step 2, the inventory: they get written from what the company assumes instead of what customers actually ask.

Step 1: Open Cassie

Sign in to Sintra, select Cassie from the Helpers list, and confirm Brain AI holds your business basics. A knowledge base is only as good as the source material, so have your collected questions and documents ready to paste in.

Step 2: Take inventory of what customers actually ask

Start from reality, not assumptions. Feed Cassie your raw material and let her extract the questions worth documenting.

Prompt: Here's a batch of our recent support emails and the questions we get asked most: [paste]. Extract every distinct question, merge duplicates, and rank them by how often they come up.

Output: A deduplicated, ranked list of real customer questions, the honest table of contents for your knowledge base, built from what people actually need.

Step 3: Structure the knowledge base

Turn the question list into an organized plan before writing anything. Structure is what separates a knowledge base from a pile of articles.

Prompt: Organize these questions into a knowledge base structure: logical categories, an article for each question or group of questions, and a priority order for writing them. Flag anything that should be one combined guide instead of separate articles.

Output: A category map and prioritised article plan: what to write, in what order, and how it all fits together.

Step 4: Write the articles with Cassie

Work through the plan top-down. Cassie drafts FAQs, how-to guides, and troubleshooting articles in a clear, customer-friendly tone.

Prompt: Write the knowledge base article for '[question/topic]'. Audience: [customers/team]. Include a short direct answer first, then step-by-step instructions, and a 'still stuck?' closing line pointing to support. Tone: [tone].

Output: A publish-ready help article: direct answer up top, steps below, consistent with your brand voice. Repeat down the priority list, batch by batch.

Step 5: Publish on your platform

Here is the honest split: Sintra doesn't host your public help center. Copy the finished articles into whatever knowledge base software or website platform you already use. Cassie writes the content, you own where it lives.

Prompt: Format these three articles for publishing: consistent headings, short paragraphs, and a one-line summary at the top of each that I can use as the article description.

Output: Clean, consistently formatted articles ready to paste into your help center, with the same structure and voice across every one, wherever you publish them.

Step 6: Load the knowledge base into Brain AI

Now make it an AI knowledge base. Store the finished articles, policies, and guides in Brain AI so the knowledge stops living only on a website and starts powering answers.

Prompt: I've added our knowledge base articles to Brain AI. Confirm what you now know about [topic] and answer a test question the way you'd answer a customer: [test question].

Output: A test answer drawn from your own documentation, proof the knowledge base is loaded and Cassie answers from it rather than from generic assumptions.

Step 7: Answer from it and keep it growing

Put the knowledge base to work. Cassie crafts customer replies grounded in Brain AI, you review, copy, and send them, and every question she can't answer well is your next article.

Prompt: A customer asks: [paste question]. Draft a reply based on our knowledge base. If our documentation doesn't cover it properly, tell me, and draft the article that would.

Output: A consistent, on-policy reply ready to send from your own channels. And when the knowledge base has a gap, the gap comes back as a drafted article instead of a shrug.

Two alternative workflows

The seven steps above cover building a customer-facing knowledge base from scratch. Two common situations need a different starting point.

▸ Turning a support inbox backlog into a knowledge base

If support answers live only in sent emails, the knowledge base already exists. It is just trapped in the outbox.

Prompt: Here are 30 of our past support replies: [paste]. Identify the answers we keep rewriting, group them by topic, and turn the top five into proper knowledge base articles that would have made these emails unnecessary.

Output: The most-repeated answers converted into articles, the fastest possible start, because every article is guaranteed to address a question customers really send.

▸ Building an internal knowledge base for your team

Not every knowledge base faces customers. For processes, policies, and how-we-do-things documentation, the same workflow points inward.

Prompt: Help me build an internal knowledge base for our team. Here's what keeps getting asked internally and our current process notes: [paste]. Structure it, draft the first articles in a plain internal tone, and tell me what's missing for a new hire to get up to speed.

Output: An internal documentation plan and first articles, stored in Brain AI so the whole AI team applies your actual processes, and new teammates stop learning by interruption.

Creating a knowledge base manually vs. with Cassie

A note on output quality: Cassie produces strong first drafts, but a knowledge base speaks for your business, so review every article for accuracy against your current product, pricing, and policies before publishing. She can't catch what changed last week unless Brain AI knows about it.

Tips for building a knowledge base with AI

These are the habits we lean on to build a knowledge base people actually use.

Write from questions, not from features. Customers search "why won't my order go through", not "payment module overview". Building a knowledge base around the inventory from step 2 keeps every article findable by the person who needs it.

Answer first, explain second. Have Cassie put the direct answer in the first two lines of every article, with steps and context below. People come to a knowledge base mid-problem, so make the answer the headline.

Batch the writing, not the structure. Approve the full category map once, then write articles in batches of three to five. Structure decided piecemeal is how knowledge bases turn into junk drawers.

Keep Brain AI and your published articles in sync. Whenever an article changes on your help center, update it in Brain AI in the same sitting, or Cassie's replies and your public documentation slowly start disagreeing.

Treat unanswerable questions as the roadmap. Every time Cassie says the documentation doesn't cover something, that is the next article, already prioritised by a real customer. A knowledge base built this way never goes stale.

Build a knowledge base your whole team answers from

Cassie turns the answers you already give into structured documentation, Brain AI stores it as the single source your AI team pulls from, and every customer reply after that starts grounded instead of improvised. Through Sintra's  AI integrations, that same knowledge feeds the tools you already work in.

Try Cassie today →

14-day money-back guarantee · Cancel anytime, no questions asked

The complete knowledge base checklist

From inventory through ongoing maintenance, this checklist covers every step worth ticking off. It works with or without Cassie, and each category expands into a standalone list you can save, print, or come back to.