Introducing the Ochre AI support workspace. Start a 14-day trial

Glossary

Definitions for the terms used across Ochre โ€” conversations, autopilot, BYOK, light agents, aliases, and more.

By ChristopherUpdated 4 min read

Glossary

A reference for the words you will see across the Ochre product and docs. If you came from Zendesk, Intercom, or Help Scout, this article maps the vocabulary.

Core concepts

Conversation

A single thread between a customer and your team. Whether it started by email, chat widget, or Slack Connect, it is a conversation. Equivalent to a "ticket" in Zendesk or a "conversation" in Intercom and Help Scout.

Customer

The person on the other side of the conversation. End user. Equivalent to "user," "contact," or "end user" in other tools. See End user vs user.

User

A member of your team. Has a role (Owner, Admin, Agent, or Light agent). See Roles explained.

Workspace

Your Ochre account. One company, one workspace. Has its own subdomain, branding, billing, and team.

Channel

A way customers reach you: email, chat widget, or Slack Connect. Each channel has its own settings, routing, and CSAT controls. See Channels overview.

AI vocabulary

AI agent

The AI that drafts and (optionally) sends replies. Runs on your Anthropic or OpenAI key. See AI overview.

BYOK

Bring Your Own Key. You connect your own Anthropic or OpenAI API key, and you pay the model provider directly. Ochre never marks up tokens. See BYOK keys.

Drafting modes

Three modes for how the AI participates in replies:

  • Draft-only. The AI writes a draft, you review and send.
  • Copilot. The AI writes a draft, one click to send.
  • Autopilot. The AI sends on its own when confidence is above a threshold.

See Drafting modes and Autopilot.

Confidence

A 0 to 100 percent score the AI assigns to its own draft. Used to gate autopilot. See Confidence thresholds.

Auto-labeling

The AI assigns three labels to every conversation: topic, priority, and confidence. See Auto-labeling.

Topic

One of: bug, billing, how-to, feedback, abuse, spam, other.

Priority

One of: low, normal, high, urgent.

Guardrails

Rules that block the AI from sending certain things autonomously, like refunds or account deletions. See Guardrails.

Spend cap

A monthly dollar limit on AI usage. The AI pauses when the cap is hit. See Spend caps.

AI brain

The collection of knowledge the AI draws on: your help center articles, ingested Notion or GitBook docs, customer history, and learned QA. See AI brain.

AI receipts

The source citations attached to every AI draft showing what the answer is based on. See AI receipts.

Learned QA

The system that turns your edits to AI drafts into better future drafts. See Learned QA.

Inbox vocabulary

Inbox filters

Saved views of the inbox: All, Unassigned, Mine, Mentions, AI handled, Closed. See Inbox filters.

Assignment

Pinning a conversation to a specific user. See Assignment.

Status

The lifecycle state of a conversation: open, snoozed, closed. See Conversation status.

Snooze

Hide a conversation until a future time. See Snoozing conversations.

Internal note

A comment visible to your team but not the customer. See Internal notes.

Mention

@-tagging a teammate. They get a notification and the conversation appears in their Mentions filter. See Mentions.

Tag

A label you attach for organization or reporting. See Tags.

Conversation alias

A secondary identifier for a conversation. After migration, your old Zendesk, Intercom, or Help Scout ID is kept as an alias so old references still work. See Conversation aliases.

Roles vocabulary

Owner

Full access including billing and workspace deletion. Billable.

Admin

Full access except billing and deletion. Billable.

Agent

Handles conversations. Cannot configure settings. Billable.

Light agent

Read-only collaborator. Free. Use for engineers, PMs, and execs.

Pricing reminder: $39 per seat per month, flat. See Pricing overview.

Knowledge base vocabulary

Article

A single help center page. Has a slug, title, body, SEO metadata, and visibility settings. See Authoring articles.

Category and subcategory

Organizational structure for the help center. See Categories and subcategories.

Help center

The public-facing site that hosts your articles. Available at a subdomain by default or a custom domain. See Help center domain.

Article visibility

Public, internal-only, or unlisted. See Article visibility.

CSAT vocabulary

CSAT

Customer Satisfaction. Five-emoji scale: ๐Ÿ˜ก 1, ๐Ÿ˜ž 2, ๐Ÿ˜ 3, ๐Ÿ™‚ 4, ๐Ÿ˜€ 5. See CSAT overview.

NPS

Net Promoter Score. 0-10 scale, separate from CSAT. See NPS overview.

Survey token

The unique link in the survey email that lets a customer respond without logging in. See Survey tokens.

Routing vocabulary

Routing rule

A condition-action rule that fires on incoming conversations: tag, assign, set priority, route to channel. See Routing rules.

SLA policy

A target response or resolution time tied to conditions. See SLA policies.

Office hours

Your team's working schedule. SLA pauses outside office hours. See Office hours.

Escalation

Automatic re-routing when an SLA is at risk. See Escalation.

Integrations vocabulary

Five live integrations: Stripe, HubSpot, Linear, GitHub, Slack. Plus Notion and GitBook for KB ingest.

Next steps

Was this article helpful?

Ochre glossary โ€” terms and definitions ยท Ochre