Docs

Appearance

Goals Guide ​

The Goals panel sets what the agent aims for during a call: a primary objective and a fallback. Both are optional — an agent with no goal still answers and handles calls — but a clear goal gives the model something to drive the conversation toward. Set them in the Agent Studio under Goals, or as playbook.closingCTA and playbook.fallbackCTA from the API.

Looking for win criteria and data capture? Those used to live here. They now sit in the Post-call intelligence panel — see Post-Call Intelligence. This guide covers only the in-call goal.

Goal (primary objective) ​

What it is: The single thing the agent should try to accomplish on the call. One sentence, in plain language.

Why it matters: The goal shapes how the agent steers the conversation. "Book a dental appointment and collect the caller's name, preferred date, and timezone" pulls the model toward asking for those things; an empty goal lets it drift.

Good examples ​

  • "Book a demo and capture the caller's company, team size, and timeline."
  • "Resolve the billing question, or collect enough detail to open a ticket."
  • "Qualify the lead: confirm budget, authority, need, and timeline."

Tips ​

  • One objective. If you have two, pick the one that matters most and put the other in the backup goal.
  • Concrete, not vague. "Help the caller" is weak. "Book an appointment and confirm contact details" is testable.
  • Name the data you need. If the goal implies collecting fields, say so — it nudges the agent to ask.

Backup goal (fallback objective) ​

What it is: What the agent should fall back to when the primary goal isn't reachable on this call.

Why it matters: Calls don't always go to plan. If the caller can't book today, a backup goal like "capture a callback number and the best time to reach them" keeps the call useful instead of ending empty-handed.

Good examples ​

  • Primary: "Book the appointment." → Backup: "Capture a callback number and preferred time."
  • Primary: "Close the sale." → Backup: "Get the caller's email to send a follow-up."
  • Primary: "Resolve the issue on the call." → Backup: "Open a ticket with clear next steps."

Tips ​

  • Make it genuinely easier than the primary. The backup is for when the main goal stalls, so it should be reachable in almost any call.
  • Keep it short. One fallback is enough.

API shape ​

yaml
playbook:
  closingCTA: "Book a dental appointment and collect the caller's name, preferred date, and timezone"
  fallbackCTA: "If they can't book now, capture a callback number and the best time to reach them"

The primary goal is playbook.closingCTA (nested under playbook) and the fallback the model actually sees is playbook.fallbackCTA. Both are plain strings compiled into the agent's prompt at call time.

backupGoal has been removed

Earlier versions also accepted a top-level backupGoal field. It was stored on the agent but never compiled into the prompt, so it had no runtime effect. It is gone: playbook.fallbackCTA is now the only fallback, on both POST and PUT/PATCH, and it is the field the Studio's Backup goal editor writes. Any backupGoal still being sent is ignored — move the value to playbook.fallbackCTA.

Neither goal is a persona field — setting them under persona passes validation and is silently ignored.

What moved to Post-call intelligence ​

The win criteria (what counts as a successful call, used to score calls after they end) and extraction fields (structured data pulled from the transcript) are configured in the Post-call intelligence panel, not here. Under the hood they're still the successCriteria and extractionTags agent fields — they just moved to the panel that owns everything that happens after the call. See Post-Call Intelligence for both.

Next steps ​

  • Write the primary goal first — it's the one that shapes every call.
  • Add a backup goal so stalled calls still produce something useful.
  • Configure scoring and extraction in Post-Call Intelligence once the goal is set.