Self-learning FAQ

Self-learning lets your AI agent automatically build and update a knowledge base from real support conversations. Instead of writing every FAQ by hand, you turn on self-learning and the agent improves over time from the kinds of questions your customers actually ask and the answers your team (and the agent) give. This article explains what self-learning does, how to turn it on, and what support teams and managers get from it.

What self-learning does

When you enable self-learning for an agent in an inbox, Chatlane periodically reviews recent conversations in that inbox. It extracts useful questions and answers from those threads—from your team's replies and the agent's replies—and updates an internal FAQ. That FAQ is then added to the agent's knowledge base so the agent can use it in future replies. Over time, the agent gets better at answering the kinds of questions your customers actually ask, using the language and style of your team.

You don't have to write or edit the FAQ yourself. The system runs on a schedule (e.g. every few hours) and updates the agent's knowledge automatically. You can still upload your own files (policies, product guides); self-learning adds another layer of knowledge from live conversations.

Which conversations are used

The system looks at conversations in the same inbox where the agent is enabled. It only uses threads that have at least one real reply—from a team member or the agent—not just notes or drafts. That way, it learns from actual answers, not from unanswered or internal-only threads. It processes conversations in batches and picks up where it left off so it doesn't repeat work or reprocess old threads unnecessarily.

How to turn it on

In Inbox Settings, when you attach an agent to an inbox, you can enable self-learning for that agent in that inbox. One FAQ is maintained per agent–inbox combination. So if you have one agent attached to two inboxes, you can turn on self-learning for one or both; each will have its own learned FAQ.

No technical setup is required. Once it's on, the system runs on a schedule (e.g. every few hours) and updates the agent's knowledge automatically. You can leave it on and let it run; your team and customers will see the agent's answers improve over time.

What support teams and managers get

• Support teams get an agent that improves from real usage instead of static docs only. The agent reflects how your team actually answers—tone, level of detail, and common phrasing—so replies feel consistent with your brand.
• Managers can rely on the agent to reflect how the team answers and to pick up new topics as they appear in conversations, without constant manual FAQ updates.
• Founders get better return on the agent without hiring someone to maintain a huge FAQ. Self-learning keeps the agent relevant as your product and support evolve.

You can still upload your own files (policies, product guides); self-learning adds another layer of knowledge from live conversations. The two work together: your core content stays in place, and the agent gets smarter from real threads.

Optional: running or checking manually

If your organisation runs Chatlane on your own servers, your admin may have the option to run a manual "self-learning" run for a specific agent or inbox—for example, for testing or to catch up after a pause. For most users, leaving it on and letting the schedule run is enough. You don't need to do anything else once it's enabled.

For giving your agent a knowledge base with your own files (FAQs, policies, product guides), see Give your AI agent a knowledge base. For step-by-step setup of your first inbox and agent, see Get started with Chatlane.