Skip to content
c12s.ai logo App & Bots Documentation

Core Concepts

This page defines the terms you’ll see across the product so teams use the same language when configuring, operating, and measuring your AI agent.

Skimming through? Start with Organization, Community, Agent, Knowledge Base, then jump to Overrides and Moderation.

Glossary index

Section titled “Glossary index”

People & containers:
Organization · Organization Members · Company

Community & platforms:
Community · Community Rules · Community Platform

AI & knowledge:
Agent · Knowledge Base · Knowledge Entry · Knowledge Source · Shared Knowledge Base · Auto-Learn · Semantic Similarity · Web Search · Allowlist vs Denylist

Control & safety:
Overrides · Content Moderation · AI Policies · Logic Rules · Spam vs Scams · Sentiment

Visibility & outcomes:
Dashboard · Analytics · Reports · Hotspots · Unanswered Questions

Organization

Section titled “Organization”

What it is (general): A workspace/brand that holds people, settings, and connected assets.
In C12s: Your top-level container for communities (under the same brand), knowledge, moderation policies, analytics, and billing. You can belong to multiple organizations and switch via the org switcher (top-right).

You’ll use it to:

  • Configure brand details (name, description, logo) and global AI defaults.
  • Invite teammates and assign roles.
  • Connect communities (Discord/Telegram/Slack).
  • Manage plan, usage, and invoices.

Organization Members

Section titled “Organization Members”

What it is (general): People with access to an organization, each with a role.
In C12s: Members inherit permissions based on role. Typical roles:

  • Administrators — full control: billing, invites, platform connections, global AI defaults, policies, and data exports.
  • Community Managers — day-to-day configuration: knowledge base, channel overrides, moderation rules, analytics, and reports.
  • Moderators — operational tools: review support requests and unanswered questions, review flagged items, submit knowledge improvements, view relevant dashboards.

For now, all organization members are considered ‘Administrators’. We will introduce RBAC (role-based access controls) in the future.

Company

Section titled “Company”

What it is (general): Legal entity used for invoicing and tax details.
In C12s: The billing profile (company name, VAT/tax ID, address, plan) associated with an organization’s subscription. This doesn’t affect community behavior or permissions. One company can have multiple Organizations (e.g. marketing agencies).

Community

Section titled “Community”

What it is (general): A real-time space where your users chat about your product/brand/project.
In C12s: A connected Discord community, Telegram group, or Slack workspace your Agent can read from and (optionally) reply in.

Community Rules

Section titled “Community Rules”

Your server/group rules (e.g., “no soliciting or commercial messages, no harrassment, etc.”) that users must follow within the community’s chats. If “Self-moderation” is enabled, users can report other user’s messages to the Agent. It’ll then assess if the contents of the message break any of the rules and take the appropriate action.

Community Platform

Section titled “Community Platform”

The provider hosting the community: Discord, Telegram, or Slack. Platform choice determines available permissions, commands (e.g., /verify), and some moderation capabilities.

Agent

Section titled “Agent”

What it is (general): The AI assistant that reads and analyses messages, answers questions, and moderates content.
In C12s: Your org’s configured AI agent with a Personality and different modes:

  • Personality — writing style and tone (e.g., Short & Friendly).
  • Read / Write / Learn — global defaults:
    • Read: allowed to read messages (needed to answer questions, moderate content and generate analytics).
    • Write:
      • Disabled: the agent never replies to users.
      • Passive (default on free): answers only when explicitly prompted (e.g., mention or command).
      • Active (recommended on paid): answers users automatically when confident.
      • Thread (recommended on Discord): always creates a thread (or replies in a thread created by the user (forums)) for every message sent in a dedicated “#ai-support” channel.
      • NOTE: Staff messages are never auto-answered; staff must prompt explicitly (either using the /ask command or by mentioning the bot).
    • Auto-Learn: can learn from staff replies to unanswered questions.

Where: Settings → AI Configuration (global) and Overrides (per channel/category/group).

Knowledge Base

Section titled “Knowledge Base”

What it is (general): Canonical information used to answer questions (such as FAQs, docs, etc.).
In C12s: A curated set of Entries and Sources the Agent consults to produce accurate replies.

  • Keep it concise and authoritative/factual.
  • Start with 10–20 topics covering ~80% of questions.
  • Update weekly based on list of Unanswered Questions and analytics.

Knowledge Entry

Section titled “Knowledge Entry”

A curated topic you create manually with these components:

  • Title — the title of the entry.
  • Source (optional) — The source of the knowledge, this is used for citations and should be a URL.
  • Questions — example phrasings users might ask; used for semantic matching.
  • Answer — the content which the agent will learn from to answer users’ questions.

Good entries are relatively short (1-3 paragraphs), source-linked, and neutrally written.

Tip After writing or pasting an Answer, click on the “Generate Questions” button to generate example questions automatically.

Note When an Entry is set to ‘FAQ mode’, the bot will literally paste the ‘Answer’ with no AI-generation. Individual Entries can also be turned off (e.g. for maintenance, or when needing an update).

Knowledge Source

Section titled “Knowledge Source”

An imported artifact (URL, docs page, PDF, etc.). Sources can be Live (re-crawled/index every period of time) or Imported (one-time import). While Imported sources can be changed manually after import (but won’t be updated over time), Live sources cannot be changed as they are re-crawled/indexed automatically. The content captured from Live sources can still be reviewed though.

Shared Knowledge Base 1

Section titled “Shared Knowledge Base 1”

A KB that can be reused across multiple Organizations (e.g., ecosystems, marketing agencies, venture studios, etc.). This avoids duplication, keeps answers consistent, and allows for a single source of truth.

Auto-Learn

Section titled “Auto-Learn”

When enabled, the Agent treats qualified staff replies to previously detected Unanswered Questions as teachable examples to improve future answers. New knowledge appears in your KB for review and refinement. Turn Learn off in sensitive/private channels (staff rooms, support tickets, etc.) via Overrides.

Section titled “Web Search”

What it is (general): Let the Agent search the web at answer time.
In C12s: (Paid) Let the Agent search allowed domains (or the entire web) so you don’t have to ingest every page. Useful for large docs/blogs that change often.

Allowlists (whitelists) vs. Denylists (blacklists)

Section titled “Allowlists (whitelists) vs. Denylists (blacklists)”
  • Allowlist — domains the Agent is allowed to search within (e.g., docs.example.com, blog.example.com). Everything else is blocked.
  • Denylist — domains the Agent will ignore for search, or subpaths or pages within allowed domains to exclude (e.g., /changelog/older/*).
    Start with a small allowlist and add denylist entries only when needed; or if you are bold, only use a denylist and exclude the domains you don’t want the Agent to search for information (such as your competitors, or untrustworthy news outlets (e.g. The Onion)).

Overrides

Section titled “Overrides”

What it is (general): Local exceptions to a global setting.
In C12s: Per-channel / category / group settings that override global AI defaults. You can change Personality, Read/Write/Learn (see Agent), and other behaviors for specific spaces.

Examples:

  • Disable Write in #announcements.
  • Use a Formal personality in support channels and Playful in general chat.
  • Turn Learn off in private staff rooms.

Where: Settings → AI Configuration → Overrides (select platform to target).

Content Moderation

Section titled “Content Moderation”

What it is (general): Detecting and acting on unwanted or harmful content.
In C12s: A combination of AI Policies (language understanding) and Logic Rules (deterministic checks) that trigger Actions: warn, delete, mute, ban, or escalate to humans, etc.

AI Policies

Section titled “AI Policies”

Model-driven classifiers that detect intents and patterns (e.g., phishing attempts, impersonation, wallet-drainer lures, NSFW, hate/harassment). Tunable sensitivity and actions.

Logic Rules

Section titled “Logic Rules”

Deterministic checks like RegEx/keywords, link filters, rate & burst limits, or role-gated posting. Ideal for crystal-clear policies (“no links in #general” or “no soliciting in #support”).

Use Overrides to restrict where moderation runs.

Community Self-Reporting

Section titled “Community Self-Reporting”

What it is: Communities that self-moderate allow for moderation scalability, all social platforms allow for content reporting, but this often falls under a human for review which most often adds reputation-impacting delays.
In C12s: A /report command any user can use to alert the staff of an ill-intended message or user. Once a report is submitted by a user, the content of the reported message is analysed against the Community Guidelines, and if an appropriate action is warranted, our AI will automatically execute it. This feature is available in Beta, and is off by default. Make sure to submit clear directives for the users to follow and the AI to analyse against (when reported).

Sentiment

Section titled “Sentiment”

What it is (general): An estimate of message tone (positive/neutral/negative).
In C12s: Aggregated sentiment over time by channel, topic, or tag to spot trends and correlate with releases or incidents.

Semantic Similarity

Section titled “Semantic Similarity”

What it is (general): Vector-based matching to find semantically similar text (for RAG and other workflows).
In C12s: Used to match user questions with KB Entries/Sources. You’ll see similarity scores in semantic search; a min. of ~70%+ is required to match or it is ignored. Adjust questions to reduce false positives/negatives. Typically longer questions reduce false positives (AI triggering on a question it doesn’t have an answer to in the KB), while shorter/broader questions will increase the hit rate (reducing false negatives). Ideally, a user question should never match an existing question (within an entry) if said Entry doesn’t respond to the question (otherwise we’re feeding irrelevant information to the AI for processing).

Spam vs Scams

Section titled “Spam vs Scams”
  • Spam — noisy or low-value promotional/repetitive content. Actions often include: warn, delete, temp-mute.
  • Scams — malicious content aiming to deceive/steal (phishing sites, fake support, “airdrop” wallet drainers). Actions often include: delete, ban, escalate, auto-report internally.

AI Policies help distinguish the two; configure actions conservatively first and monitor log channels and analytics dashboards.

Dashboard

Section titled “Dashboard”

What it is (general): An at-a-glance operational view.
In C12s: We have several dashboards, including the main one (which only lists Unanswered Questions for now, ordered by most recent), Analytics (graphs & activity hotspots under the ‘Stats’ section), and Billing (plan usage and management, etc.).

Analytics

Section titled “Analytics”

What it is (general): Metrics and trends over time.
In C12s: Volume and coverage of AI answers, moderation actions, sentiment trends, and engagement. Use it to prioritize KB updates and policy changes.

Reports

Section titled “Reports”

What it is (general): Periodic summaries sent to stakeholders.
In C12s: Scheduled digests or exports (e.g., weekly) covering answers, gaps, sentiment, and incidents. Great for leadership updates and vendor reviews.

Hotspots

Section titled “Hotspots”

What it is (general): Areas requiring attention, or displaying activity patterns over time.
In C12s: Channels/topics/time windows with unusual spikes in unanswered questions, negative sentiment, or moderation actions. Use hotspots to allocate staff time and expand the KB.

Unanswered Questions

Section titled “Unanswered Questions”

What it is (general): Questions the Agent didn’t answer due to lack of knowledge.
In C12s: A list in the main dashboard that shows missed queries with context. Use it to add or refine Knowledge Entries and improve semantic coverage.

Related:

Footnotes

Footnotes

Section titled “Footnotes”

  1. This feature is still in planning and not yet available. ↩