> ## Documentation Index > Fetch the complete documentation index at: https://docs.erdo.ai/llms.txt > Use this file to discover all available pages before exploring further. # Core concepts > How work starts, progresses, and turns into durable outputs in Erdo. Erdo exposes the work and its result. You do not need to choose the internal AI specialist or tool that performs each step. ## Conversations A **conversation** is the flexible way to work with Erdo. Ask for something in plain language and results such as tables, charts, pages, and files appear inline. Conversations keep their history, so you can return, add context, or ask for a change. Conversations are best for open-ended work. If the task has a fixed sequence and needs durable progress outside the transcript, start from a template instead. An agent's identity is separate from its **deployments**. The agent owns its instructions, knowledge, and skills; deployments let that same worker operate on a website (chat, voice, or video) or over the phone. Channel-specific settings such as voice, greeting, appearance, phone number, and usage limits stay on the deployment. Manage both from **Agents**. ## Activity **Activity** is the operational return loop. It answers four questions: * Is everything okay? * What needs me? * What changed? * What is running? Approvals, attention items, progress, and meaningful results meet here. Activity shows a quiet state when nothing needs action; it is not an automation catalogue or a raw run log. ## Pages A **page** is a durable app Erdo builds for you: a dashboard, report, landing page, or tool connected to live data. Pages run in a managed sandbox under your permissions and can be shared privately or publicly. Pages created in a project appear in its recent outputs. Pin the important ones to keep them on the project home and sidebar. See [Pages](/pages). ## Agents An **agent** is a persistent worker you configure for a repeatable job. It keeps its own instructions, knowledge, skills, data, versions, and activity, and can be tested and improved over time. Build one when the worker itself is something your organisation needs to own and reuse. You do not choose an agent before every conversation. Erdo coordinates its internal specialists automatically for ordinary work; the Agents workspace is for building and operating the durable agents you deliberately create. See [Agents](/agents). ## Templates and progress A **template** is the starting point for a defined, repeatable outcome. It collects the right brief and establishes the stages the work must pass through. Erdo tracks the durable state of that outcome, including: * current stage and overall progress; * anything waiting for your input; * outputs and connected resources; * tests and decisions; * the activity log and source conversation. This keeps a campaign, lead engine, recurring report, or similar body of work understandable without reading every message. Its changes and decisions return through [Activity](#activity). ## Projects A **project** is optional shared context for sustained work. It can collect conversations, progress, learnings, and outputs without forcing every question into a hierarchy. Pinned pages act as its durable output shelf. Quick questions and one-off deliverables do not need a project. Use the project switcher at the top left to change scope, create a project, or open project management. ## Data and connections There are two ways to give Erdo business data: * **Datasets** are files or tables that live in Erdo and can be queried directly. * **Connections** are authorised links to external databases and tools. They let Erdo read data and, with your approval, take action in the connected system. Manage both from **Setup → Data & connections**. Each connection is encrypted. See [Data & connections](/data). Some capabilities depend on a specific connection. For example, Erdo can only book a meeting on a call after Google Calendar is connected. See [Voice](/voice). ## Knowledge **Knowledge** is supporting context Erdo should reuse: business definitions, the shape of your data, approved learnings, and reusable instructions. It keeps future work consistent instead of making every conversation start cold. Manage it from **Setup → Knowledge**; you do not choose an agent or attach skills before starting work. See [Knowledge](/knowledge). ## Review and approvals Erdo keeps a human in the loop. Proposed knowledge can be reviewed, and consequential actions such as sending an email, changing a connected system, or placing a call require approval according to your autonomy settings. Decisions surface in the relevant conversation and in Activity. ## Automations and voice **Automations** run work on a schedule or in response to an event. Their results, failures, and decisions return through Activity. See [Automations](/automations). Erdo can also place outbound **phone calls** to qualify a lead, confirm a detail, or book a meeting. See [Voice](/voice). *** Connect a data source and ask Erdo for a real outcome.