Live objection-response prompts

The job is a single approved reply, shown only to the AE, when a buyer names a competitor or pushes on price. The AE can use it, paraphrase it, or mute it. If the buyer can hear a script, you built the wrong thing.

This is quality control on language, not a prediction of who will win. Do not publish a win-rate lift for live prompts. Yours would depend on which objections you wired, whether the AE already knew the card, and whether they were still listening. Vendor decks will imply a number. It is not yours.

Live-assist tools in this class include Cresta and Balto. Recording and post-call review often sit in Gong or Chorus. This page does not rank those products. None of them writes the rebuttal if you never approved a sentence, never named the trigger, and never told the AE they may ignore the card.

This is not real-time call coaching. That page is discovery: missed questions, unexplored pain, talk tracks while the buyer is still explaining the problem. This page is negotiate: a named competitor or a price pushback, one sentence, then silence. If you prompt both jobs in the same overlay, the AE reads instead of listening.

Write the triggers a stranger could follow. Two families, not a catalogue of every hard moment on a call.

• Named competitor. The buyer says a specific vendor you have a reviewed battle card for. "We're also talking to [Competitor]" is a trigger. "We're looking at a few options" is not.
• Price pushback. The buyer says you are expensive, asks for a discount, or compares a number you quoted to a number they already have. "We need to think about budget" is not, by itself, a price objection. It is a stall. Do not treat it as a request to cut price.

Everything else stays off the overlay. Discovery pain ("our last tool was a mess") is not an objection. It is context. Prompting a rebuttal there trains the AE to argue instead of asking what happened. That is how you get a card that fires on empathy.

The sentence on the card comes from approved battle-card language, the same words legal and product already signed. Do not generate a new claim while the buyer is still talking. Unreviewed competitive comparisons are published material with a shorter fuse. Unreviewed discount language is a commercial commitment. If the right move is a number inside policy, that number belongs in pricing recommendation engine after the call, not as a live offer the AE reads aloud.

One sentence. Not a paragraph. Not a case study. If the AE needs a logo-matched story, that is case-study retrieval RAG for the follow-up, not a wall of text over the buyer's next remark.

The prompt is mute-able. Default on is fine for new hires on two triggers. Default off, or a one-tap hide, is required for anyone who already knows the card. If they cannot kill it mid-call, they will read it to make it go away.

The walk-through is illustrative, not a measured result.

An AE is on a commercial call. The buyer names the incumbent and says the quote is a lot higher. The overlay fires on named competitor plus price pushback. The card shows one approved sentence, for example: "The gap is usually implementation and the seats you are actually using, not list. I can walk the two stacks side by side if that would help."

The AE still listens. The buyer may already be explaining that the incumbent is bundled with a parent contract. That fact is more useful than the card. The AE mutes the prompt, asks who owns the bundle, and does not quote a discount. After the call they pull the approved comparison and, if a commercial exception is even in play, they take it to the pricing path rather than inventing one on the phone.

The failure version does three things at once.

It reads the card over the buyer. The AE's voice goes flat, the sentence does not match what the buyer just said, and the buyer hears a script. You would have been better with no overlay.

It invents discount language. The card, or the AE under pressure, offers to match the incumbent or to cut price "today." That sentence was not approved. It now sits in a recording.

It had already been prompting. Twenty minutes earlier the buyer described a painful rollout. The system treated that pain as an objection and suggested a rebuttal. The AE argued with a story instead of asking who ran the last implementation. By the time price came up, the buyer had stopped volunteering detail.

Same call, different buyer: they mention a competitor you have no card for. Do not generate a comparison. Show nothing, or show "no approved language for this name." Silence is the honest output.

The card is for the AE. It is not in the buyer's ear, not in chat, not on a shared screen. If you screenshare the desktop that holds the overlay, you have put the script in the room. Use a layout the buyer never sees.

The AE still has to listen. A prompt that arrives after the moment has passed is a second conversation competing with the first. Latency is a design constraint, not a slogan. If the sentence cannot land before the AE needs to speak, turn the trigger off.

Live prompts need audio, or a transcript of audio. That means recording or streaming. Disclose it. Use the same notice you already use for Gong-class or Chorus-class recording: verbal at the start, and whatever your jurisdiction and your counsel already require. Do not hide the live-assist layer behind a generic "this call may be recorded" if counsel wants the processing named. Do not skip notice because the prompt is "only for coaching." The buyer is still on a captured line.

If a region or a customer forbids recording, do not run live prompts on that call. There is no silent mode that still hears the competitor name.

Review the tape after, not during. Listen for whether the prompt fired on the right trigger, whether the AE used the sentence or ignored it, and whether they talked over the buyer. That review is enablement. It is not a score that should set quota.

If the first time an AE sees the approved reply is live, they will read it. Put the same sentence in negotiation role-play simulator until they can say it in their own words and then drop it when the buyer is doing something else.

After the call, the next action on a late-stage deal is last-mile deal coaching: missing signature, stalled legal, missing economic buyer. The overlay does not own that. It owned one moment of language.

Before you add a third trigger, sit with enablement and replay recent negotiate calls. Check four things: did it fire only on a named competitor or price pushback, was the sentence approved, could the AE mute it, and did they still listen. If a tape shows someone reading the card, offering an unapproved discount, or rebutting discovery pain, delete that trigger before you scale.