How big should my knowledge base be for AI agents to work well?

Quality matters more than size. 30-50 well-written articles covering your top-volume questions outperform 500 inconsistent articles. Most production deployments start with the top 20 ticket categories—if you have good docs for those, deflection rates of 30-50% are achievable from day one. Expand coverage as the AI's escalation patterns reveal gaps.

Should AI agents only answer from the knowledge base?

For high-stakes domains (legal, healthcare, finance), yes—restrict the agent to cited sources to prevent hallucination. For lower-stakes tasks (general product questions, how-tos), allow the model to use its training knowledge as a fallback, with clear citation when it does. The right setting depends on your risk tolerance and how comprehensive your KB is.

Knowledge Base

Written by Max Zeshut

Founder at Agentmelt · Last updated Sep 9, 2026

A collection of articles, FAQs, or documentation that an AI agent can search and cite. Support and legal agents use knowledge bases to answer questions accurately and deflect tickets. Often synced from your help desk or wiki.

Example

A B2B SaaS company connects its 240-article Zendesk Guide to an AI support agent. When a customer asks how to set up SSO, the agent retrieves the SSO setup article, summarizes the steps, links to the full doc, and offers a follow-up if the steps don't work—answering in 4 seconds what would take a human rep 8 minutes.

Frequently asked questions

How big should my knowledge base be for AI agents to work well?: Quality matters more than size. 30-50 well-written articles covering your top-volume questions outperform 500 inconsistent articles. Most production deployments start with the top 20 ticket categories—if you have good docs for those, deflection rates of 30-50% are achievable from day one. Expand coverage as the AI's escalation patterns reveal gaps.
Should AI agents only answer from the knowledge base?: For high-stakes domains (legal, healthcare, finance), yes—restrict the agent to cited sources to prevent hallucination. For lower-stakes tasks (general product questions, how-tos), allow the model to use its training knowledge as a fallback, with clear citation when it does. The right setting depends on your risk tolerance and how comprehensive your KB is.

See it as a workflow

Support Ticket Deflection Workflow

Trigger, steps, n8n nodes, guardrails and an importable template — plus what it costs to have it built.

Or skip the build

Workflows from $197/month, custom agents from $2,000.

Related glossary terms

Related niches

Back to glossary

Loading…