# Engine: Train. Knowledge always current. Procedures always ready.

![](/assets/einstein-hats.svg)
The loop, phase 01 of 04

# Train. Always current. Always ready.

Living Knowledge keeps your business knowledge fresh. Living Context captures the procedures that turn knowledge into action.
  [See it in action→](/en/contact-us/)
Why Train

## Content rots. Train fixes that.

Most AI projects fail because the content goes stale. Articles drift, runbooks rot, the agent answers from yesterday. Train solves that. Knowledge stays current on autopilot. Procedures stay ready without anyone writing a script.

How Train works

## Two senses. One pipeline.

Living Knowledge. Your knowledge, automated.

A three-step pipeline. Each step does one job.

01 - Sources

### It listens.

Pull raw, unstructured signal from your business systems. Tickets, internal docs, Slack, Confluence, merge requests, recordings, websites.
  →
02 - Content Library

### It maintains.

Organizes and deduplicates the result. This is the grounding data the AI uses to answer.
  →
03 - Surfaces

### It serves.

Publishes the human-readable version across every surface that needs it - embedded inside the agent, served back to the team, available to every other system on autopilot.

Feedback loop

The support team is your highest-quality signal. Team Assistant captures every accepted draft, every override, every edit, and feeds it back into the Library. The loop learns from the people who already know the answer.

Living Context. Meaningful actions, captured.

Runbooks, but agentic. Defined once, run with permission.

#### Capture procedures.

Reset a password. Check an order. Trigger a renewal. Fire an in-app notification at the right moment. Same way you would write a runbook.
 [See it in Deploy →](/en/engine/deploy/)
#### Connect to your stack.

CRM, billing, ERP, support tools, custom APIs. The agent does not invent actions. It runs the ones you defined.
 [See all connectors →](/en/integrations/)
#### Permission every step.

What the AI can do alone, what needs human approval, what is off limits.
 [See the trust architecture →](/en/trust/)
Inside Train

## Sources in. Knowledge out.

Plug in any source. The Library maintains itself on autopilot, and the procedures defined here run across every customer moment in Deploy.
              unless.com/en/dashboard/train/knowledge       [![](/assets/svg/icon-light.svg)](/)
Dashboard

## Knowledge base
        6 sources · knowledge base
 - Engineering wiki ·  1.2k docs      Connected
- Sales playbook ·  348 docs    Team only   Connected
- Team Assistant ·  5.8k tickets      Connected
- Public docs site ·  89 pages      Syncing
- Compliance library ·  0 docs      Error
- Pricing & contracts · Not connected     Disconnected

## Your AI gets smarter every day. Without anyone editing a single article. Or writing a single script.

The Library

## One Library. One source of truth.

Every signal flows into one structured Library - merged, deduplicated, labeled. Your team reviews suggested edits, accepts or rejects them in one click, and audits the full history of every change. When the agent gives an answer, this is where it came from. Non-ambiguous, always current, fully traceable.
              unless.com/en/dashboard/train/knowledge/content-library       [![](/assets/svg/icon-light.svg)](/)
Dashboard

## Knowledge base
      2,102 articles generated
Next in the loop
  [Phase 02
Test

Every role, every skill, before a customer sees it.](/en/engine/test/)    See it applied [See Train in action on Acquisition →](/en/agent/acquisition/)                  unless.com/en/dashboard/train/knowledge       [![](/assets/svg/icon-light.svg)](/)
Dashboard

## Knowledge base
        6 sources · knowledge base
 - Engineering wiki ·  1.2k docs      Connected
- Sales playbook ·  348 docs    Team only   Connected
- Team Assistant ·  5.8k tickets      Connected
- Public docs site ·  89 pages      Syncing
- Compliance library ·  0 docs      Error
- Pricing & contracts · Not connected     Disconnected
                  ![](/assets/einstein-hats-paper.svg)
## See Train against your knowledge.

Plug in your sources. Watch the Library and the Help Center build themselves. No edits, no scripts.
[Book a pilot→](/en/contact-us/)[Read the Test phase](/en/engine/test/)