Conversation design¶

Conversation design is how an agent runs the call. callcat.ai offers three modes, from simplest to most powerful. You pick one per agent on the agent page.

Mode	Best for	Who can use it
Questions & forwarding	Straight intake plus simple "if X, transfer" rules.	Existing agents already using it (legacy).
Questions & conditions	Intake plus instructions that fire only when a topic comes up.	Everyone — the default for new agents.
Decision tree	Fully scripted, branching conversations.	Operators a super admin has enabled it for.

Defaults and availability

New agents start in Questions & conditions. Questions & forwarding is a legacy mode kept only for agents that already use it, so their setup is preserved. Decision tree is off by default; a super admin enables it per operator (Super Admin dashboard → Operators → Decision tree toggle), and it then appears in the mode switcher for every agent under that operator.

Questions & forwarding (simple)¶

The classic setup: the agent works through your intake questions and, separately, applies a flat list of call-forwarding rules — "if the caller says X, transfer to this number."

Use it when intake and forwarding are independent and you don't need the agent to change its behavior based on what the caller says.

Questions & conditions (scenario)¶

The recommended mode. You keep your standard intake questions and add conditional instructions — called triggers — that only activate when a condition is met.

A trigger has:

a condition in plain language (for example "the caller wants to book a viewing"), and
one or more actions: say something, ask extra questions, and/or forward the call.

When a caller raises a topic that matches a trigger, the agent acts on it immediately — even mid-intake — instead of finishing every standard question first.

Write specific conditions

Vague conditions ("the caller needs help") make the agent fire triggers at the wrong time, or miss them. Be concrete about what the caller says or wants.

Turning a trigger off¶

Each trigger has an on/off switch. A trigger that is off is ignored on calls and its details are folded away, but it stays in your list. Use it to pause a rule (for example a seasonal offer) without losing the wording.

Forwarding is just an action on a trigger here — see Call forwarding & transfers for how that behaves.

Decision tree (advanced)¶

A decision tree scripts the whole call as a branching flow: each step is either a yes/no condition or an action (say something, collect a field, transfer, answer from knowledge, or end). It replaces standard intake entirely and gives you precise control over every branch.

Decision tree suits complex, highly-regulated, or heavily-scripted call flows. A super admin must enable it for your operator before it appears in the mode switcher below.

Required steps¶

A collect step has a Required box. When it is ticked, the agent must get that answer before it moves on. If the caller hesitates or declines, the agent explains why it is needed and asks again, patiently. You can write that explanation in the insist box; if you leave it empty the agent makes a short one up. Only if the caller firmly refuses after several tries does the agent give up on that step. When the box is not ticked, the agent tries about twice, then moves on if the caller can't or won't answer.

Switching modes¶

The mode switcher on the agent page only shows the modes available to you. Switching to Questions & conditions brings your existing forwarding rules across as forward-only triggers, so nothing is lost.

Call forwarding & transfers — Set the conditions under which an agent hands a call to a human, and understand exactly how the transfer behaves.
Configuring an agent — Set an agent's persona, voice, language, greeting, knowledge base, and intake questions.
Advanced agent switches (super-admin) — Per-agent super-admin switches that change how the agent collects and speaks back spelled details and the next question.
Catalog & order taking — Let an agent answer questions about your products and take orders, and what each catalog switch does.
Agents — What an agent is and the three things you configure — who it is, how it converses, and when it transfers.