Improving Knowledge Base Accuracy for Effective AI Agents

👋 Hey everyone — curious if others have thought about or run into this challenge. TL;DR: I’m trying to solve the problem of keeping Knowledge Bases accurate and up to date so agents can consistently operate with the right context. Right now, context makes agents powerful — but that context decays fast as projects evolve. I want to build an agent whose sole job is to maintain and update KBs automatically based on new information (like meeting decisions or project changes). Curious if others have tried something similar or if Lindy has thought about this problem space. 1️⃣ Agent outputs are 10x better when the agent has context. For example, if a meeting note-taker agent already knows who I am, who Ethan and Flo are, what our roles are, and what “Project X” means — its notes about that meeting become dramatically more useful. 2️⃣ Knowledge Bases (KBs) are a great way to give that context. A few well-written KBs about the company, team, and active projects can make an agent actually understand what it’s talking about. 3️⃣ Those same KBs can empower every agent, not just the note-taker. Once you have living KBs documenting your org and projects, all your agents — whether they’re drafting specs, summarizing updates, or managing tasks — become way more effective “AI teammates.” 4️⃣ The problem: orgs evolve fast, and KBs quickly go stale. That’s why I want to build an agent whose only job is to keep KBs up to date. Example: if Project X’s KB lists features A, B, and C, and in today’s meeting we decide to descope C and change A — that agent would detect which KB needs updating and handle it automatically. 📢 My asks: