Skip to main content

Resolve Reasons

Resolve reasons tell your AI agent when a conversation is finished and should be closed — and they record why each chat ended. That "why" turns into reporting you can act on: which questions get fully answered, how often customers simply say goodbye, and where conversations quietly drop off.

Where resolve reasons live​

Just like handover reasons, resolve reasons are set up in two places:

  1. Settings → Resolve Reasons — where you create and manage the reasons for your whole workspace.
  2. The agent's Resolve tab (Agents → your agent → Resolve) — where you choose which reasons this agent may use.
Assigning is the on/off switch

A reason only applies to an agent once you assign it on that agent's Resolve tab. An agent can always close a chat; assigning reasons is what lets it record why.

What is a resolve reason?​

A resolve reason is a condition for closing a chat that doubles as the label stored on that closure.

FieldWhat it means
LabelA short name for the reason (e.g. "Customer Confirmed Resolved"). 2–50 characters, must be unique. This is the closure label shown in your reports and on the customer timeline.
DescriptionA plain-English description of when the conversation should be closed with this reason (e.g. "The customer said their question was answered or their problem is fixed"). 20–200 characters. This is the instruction your AI follows.
DefaultMarks a reason as a starter reason, auto-assigned to new agents. Your workspace must always keep at least one default.
No team, no pause

Unlike handover reasons, resolving a chat has no team routing and no Pause-AI setting — it only closes the conversation and records the reason.

How your AI decides to resolve​

After each exchange, Orki checks whether the conversation is genuinely finished — for example the customer said goodbye, confirmed they're all set, or a resolve condition you configured clearly applies. When it is, the agent closes the chat and records the matching reason. Resolving:

  • Sets the chat's status to Closed
  • Clears any pending follow-up (a resolved chat needs no review)
  • Stamps the chat with the reason you'll see in reports

The agent does not send a message when it resolves — closing is silent.

Judgeable from the chat

Write each Description as something visible in the conversation itself. The AI can't see your CRM or order system — only what was said. "The customer said thank you and goodbye" works; "the order shipped" does not.

Default resolve reasons​

Every workspace starts with four default reasons:

ReasonWhen it applies
Customer Ended ChatThe customer said goodbye or clearly signalled the conversation is over.
Customer Declined HelpThe customer said they need nothing further or declined an offer of more help.
Customer Confirmed ResolvedThe customer said their question was answered or their problem is fixed.
OtherThe conversation ended for a reason not covered above — the agent explains briefly in the resolve comment.

Other is the catch-all. It's an ordinary, editable reason, but it's usually the one kept as the surviving default so the agent always has a way to close a chat.

Keep at least one default

Your workspace must always have at least one reason marked as Default. If you try to delete or un-mark the last default, Orki blocks it and asks you to mark another reason as default first.

Custom resolve reasons​

Add reasons that match how your conversations actually end, so your reports are meaningful. For example:

  • "Order Placed" — the customer completed a purchase in the chat
  • "Booking Confirmed" — an appointment was booked
  • "Redirected to Website" — the customer was pointed to a self-service page

To create one:

  1. Go to Settings → Resolve Reasons and click Add.
  2. Fill in a Label and a Description of when the chat should close with this reason.
  3. Save, then assign it to your agents on their Resolve tab.

Assigning reasons to an agent​

Open Agents → your agent → Resolve and turn on the reasons this agent should use. Assignment is the per-agent on/off switch, and different agents can categorize closures differently.

Why resolve reasons matter​

Every closed conversation is stamped with its reason and who closed it (the AI, a human, or auto-close on inactivity). That data powers:

  • The closures-by-reason dashboard — see at a glance why your chats end, and in what proportion.
  • Exported reports — the resolution breakdown is included in your downloadable report.
  • The customer timeline — each conversation shows how it was resolved.

Well-chosen reasons turn "we closed 400 chats" into "180 were fully answered, 90 placed an order, 40 were redirected, 30 dropped off" — the kind of insight you can act on.

Resolving from the inbox​

Your human team resolves chats too. When you click Resolve on a chat, you'll pick a resolution reason (required when your workspace has reasons) and can add a short comment. These human-resolved chats are recorded the same way, tagged as resolved by a person. See Chat Actions.

Precheck Guidance​

The Settings → Resolve Reasons page has a Precheck Guidance box — optional free-text guidance that applies to all your resolve reasons at once (e.g. "Never resolve while the customer still has an open question"). Leave it empty if you don't need it.

How the AI reads your reasons​

Each assigned reason is handed to the AI as an instruction, shown as:

LABEL: your description text

When the AI closes a chat it picks exactly one reason and adds a brief comment. Clear, distinct descriptions produce accurate, useful categorization.

Writing good resolve reasons​

  • Match how your chats really end. Reasons only help if they mirror your actual conversation outcomes.
  • Keep them judgeable from the transcript. The AI sees only the conversation.
  • Keep the set small and distinct. A handful of clearly different reasons beats a long list of overlapping ones.
  • Keep "Other". It guarantees the agent can always close a chat, and flags conversations your other reasons don't cover.

Best practices​

  • Start from the defaults, then add a few reasons that reflect your business outcomes (orders, bookings, redirects).
  • Review the closures dashboard periodically and refine reasons that rarely fire or overlap.
  • Test in the Playground — run a conversation to its end and confirm the agent closes it with the reason you expect.

Next Steps​