Insulin · Knowledge Bases New
An AI knowledge base that answers from your documents.
Load the playbooks, policies, and runbooks your marketplace business already runs on. Insulin agents search them before answering and return the sources they used — so you can check the answer, not just take it.
Now in early access — request a demo to get started.
Illustrative sample data — not customer results.
Definition
What is an AI knowledge base?
An AI knowledge base is a searchable repository of proprietary documents that an AI agent consults before it answers. In Fours Insulin, a knowledge base is how you make sure an agent replies from your own material rather than from a model's general training.
That distinction matters more in marketplace operations than almost anywhere else. The rules that govern a private offer, a listing submission, or a co-sell referral are specific to your company and your agreements. A model that has never seen them can only approximate. An agent with a knowledge base attached reads the actual policy and shows you which document it came from.
Hybrid search finds the passage
The agent searches your knowledge base; in the recommended hybrid mode it combines keyword and semantic matching so the right passage surfaces whether you used the internal jargon or not.
Cited sources come back with it
Responses carry the documents they were drawn from, so a reader can check the claim instead of trusting it. That is the difference between an answer and a guess.
Your documents stay yours
A knowledge base holds your proprietary material, and a custom agent reaches only the bases you attach to it — scope is the boundary, the same way integration access is. The built-in Insulin assistant searches every base you own by default, until you pick an explicit set.
What to load
The documents your business already runs on
You do not need to write anything new. Most teams start by loading material that already exists and is already authoritative — it is simply scattered across drives, wikis, and inboxes today.
- Listing playbooks How your team gets a product live on each marketplace, and what tends to go wrong.
- Pricing and packaging policy The dimensions, tiers, and discount rules a deal desk answer has to respect.
- Private offer approval rules Who signs off on what, at which threshold, and in which order.
- Co-sell program requirements What each hyperscaler expects on a referral before it will be accepted.
- Contract and legal templates The clauses your team is allowed to agree to without escalation.
- Internal runbooks Reconciliation, metering, and support procedures that currently live in one person's head.
Illustrative sample data — not customer results.
Across the workspace
One base, grounded everywhere Insulin works
A knowledge base is not a separate destination you visit. Once it is attached to an agent, its grounding follows that agent into chat, into channels, and into every scheduled job it runs.
Agents answer from it
Attach one or more knowledge bases to an agent and its answers cite your documents rather than general model training.
Explore Insulin agentsJobs run against it
An unattended scheduled run reads the same grounded material an interactive session would, so overnight output follows your policy too.
Explore scheduled AI jobsChat shows the sources
Ask a question in chat and the answer arrives with the documents behind it, alongside the tool calls the agent made.
Read how agents cite sourcesChannels share the grounding
When several teammates and agents work a thread together, everyone sees the same cited material instead of arguing from memory.
Read the Channels docsWhat changes
Institutional knowledge stops being a person
In most marketplace teams, the answer to "can we discount this?" lives with whoever has been there longest. A knowledge base moves that answer into a document every agent reads and every person can check.
- New hires get the same answer a veteran would
- Policy lives in a document, not in tribal memory
- Answers from a base cite the source document for each result
- Auto-sync connectors re-index your edits about hourly
- Scheduled runs follow the same rules as people
- Custom agents stay inside the bases you attach
How it works
From scattered documents to grounded answers
Start with one domain and one agent. A single well-grounded specialist is worth more than six that guess.
Create the base
Open Insulin in the Fours console and create a knowledge base for one domain — listings, deal desk, finance.
Add your documents
Load the proprietary material that already governs the work: playbooks, policies, templates, runbooks.
Attach it to an agent
Give the specialist that owns the domain access, so its answers are grounded from the first question.
Check the citations
Ask a question you already know the answer to, read the cited sources, and fix the document if the citation is wrong.
Knowledge Bases are part of Fours Insulin, the AI workspace in the Fours console — view pricing or contact sales for availability.
A knowledge base needs something to ground. Start with AI agents for business workflows, put them on a timer with scheduled AI jobs, or install a ready-made specialist from the AI agent marketplace.
Want the grounded answer in a shared thread? Insulin channels put teammates and agents in the same conversation.
Go deeper on knowledge bases: which documents a knowledge base can read, why a knowledge base beats a wiki for agents, and keeping one fresh over time.
Insulin is a software product from Fours — a general-purpose AI platform for business teams. It is unrelated to the hormone, to diabetes care, or to any medical or pharmaceutical product.
Frequently Asked Questions
Common questions about Insulin knowledge bases
What is an AI knowledge base? +
An AI knowledge base is a searchable repository of your own documents that an AI agent consults before answering. It grounds responses in your company's proprietary information instead of the model's general training.
How do knowledge bases work in Fours Insulin? +
You create a knowledge base, add your documents, then attach it to an agent. When that agent answers a question, it can search the base, and it cites the source document for each result it uses.
What should we put in a knowledge base? +
The material that already governs the work — approval policy, pricing rules, contract templates, support procedures, and internal runbooks. Anything a new hire would otherwise have to ask someone about.
How does hybrid search work in Insulin? +
Hybrid search combines vector similarity with keyword matching, and you choose how much weight the vector side carries. A knowledge base can also be set to vector-only or keyword-only. Hybrid catches both exact terms and questions phrased a different way.
How do agents cite their sources? +
Each result an agent retrieves carries the source document it came from, so an answer shows what it was based on. An agent can search up to three knowledge bases in a single turn, and a test search also shows a relevance score per result.
What document types can a knowledge base hold? +
PDF and DOCX files, and text formats including Markdown, CSV, TSV, JSON, HTML, XML, and YAML — up to 25 MB per file. You can also crawl a public website, or connect GitHub, ClickUp, or Google Drive.
How is this different from a general AI chatbot? +
A general chatbot answers from training data. An Insulin agent searches its knowledge bases first — the ones attached to a custom agent, or by default every base you own for the built-in assistant — and cites the documents it used, so answers reflect your records, not general knowledge.
Is my data used to train AI models? +
No. Fours does not use your documents or conversations to train AI models. Knowledge base content is searched at query time to ground an agent's answer and return cited sources.
How is this different from a wiki? +
A wiki waits for someone to search it. A knowledge base is read by the agent on every question, including inside unattended jobs, so the policy is applied rather than merely available.
Can we control which agents see which documents? +
Yes, for custom agents: a custom agent reaches only the knowledge bases you attach to it, so finance material stays with the finance agent. The built-in Insulin assistant searches every base you own by default, read-only; pick an explicit set in its settings.
Ground your AI in your own documents
See how Insulin knowledge bases turn scattered playbooks into answers your team can verify.