Playbooks & battlecards

Battlecards are hand-authored cards that fire on trigger phrases. Playbooks group them by stage or motion and set the defaults for a call.

Battlecards

Open Knowledge → a battlecard (or New battlecard). The editor has four parts:

• Trigger phrases: the things a customer says that should fire the card. Matching is semantic, so "we're looking at Gong" and "Gong quoted us less" both hit a Gong card. Add 3–8 phrases.
• Headline: ≤ 12 words. This is the glance cue; reps read it in a saccade.
• Body: ≤ 40 words, first person, the actual line to say. The meter turns amber when you're over.
• Provenance: which documents back this card. Cards with provenance show a why this footer like any grounded suggestion.

The right column renders the card at true HUD size, dark glass, with its keyboard shortcut, so you see exactly what the rep sees.

Publishing and versions

Publish creates a new version; every previous version is kept and can be restored from Version history. Unpublishing takes a card out of rotation without deleting it.

Test in roleplay

Launches a roleplay session in the desktop app where the AI customer opens with one of the card's trigger phrases. Useful for checking the card fires and reads well before reps meet it on a live call.

Performance

Once a card has fired, the panel shows suggestions driven, used-by-reps %, helpful/unhelpful votes and deals advanced. A card that fires often and is rarely used is a writing problem, not a retrieval problem.

Playbooks

A playbook is a named set of battlecards plus defaults:

• Stage: discovery, demo, negotiation, renewal.
• Assertiveness: how eagerly the copilot offers cards (shadow → normal → assertive). New reps often start in shadow mode: transcript and passive cards only.
• Discovery checklist: which of budget, authority, need, timeline, competition and next step the live checklist tracks for this motion.
• Language: default transcription language for calls under this playbook.

A deal's stage picks the playbook automatically when a CRM is connected; otherwise the rep picks one when arming.

Writing guidance

• Say the thing, don't describe the thing. "Our list is $69 a seat; at 40 seats I can do $59 on an annual" beats "discuss volume pricing".
• One idea per card. Split a long objection into two cards with different triggers.
• End on a question or a next step where it's natural; it keeps the rep in control of the conversation.
• Never put a commitment in a card the rep can't make. The guardrail layer blocks discounts below the configured floor, but a clean card never gets near it.