A sales assistant should write the sentence, not a summary
The buyer says they already use something for this. You have a couple of seconds. An assistant that flashes “incumbent objection detected” has told you what you just heard. You still need the thing to say.
I think about this constantly because of Axelize, the sales assistant I build. Its whole job is to follow a live call and write your next sentence in your own words while the buyer is still talking. The reply is the product. Everything else on the panel exists to get it in front of you faster.
Three things an assistant can put on screen
| What it shows | What you do next | Where it fits |
|---|---|---|
| A summary: “Buyer is comparing vendors” | Decide what to say, then say it | Notes after the call |
| Talking points: “Mention integrations, ask about timeline” | Pick one, turn it into a sentence, say it | Prep before the call |
| The sentence: “We plug into what you already use, so nothing changes for your team. What would you want to keep exactly as it is?” | Read it | The call itself |
The sentence in the last row is an illustration I wrote for this post, not a capture from a real call.
Look at the middle column. Two of the three formats leave you with work to do while the buyer waits. Only the third hands you something you can say.
Why design for the newer rep first
The best evidence I know of comes from NBER, in a paper called Generative AI at Work by Brynjolfsson, Li and Raymond. They studied 5,179 customer support agents who got a conversational AI assistant. Issues resolved per hour went up 14% on average. For novice and low-skilled agents it went up 34%. For experienced agents the effect was minimal.
That’s support chats, not sales calls, so I’m not promising you a number. What I take from it is where to aim. A veteran can read a summary and improvise. The person who’s newest, the one who doesn’t know what to say next, is the one a ready sentence helps most.
What the rule does to an interface
If you’re building a panel like this, four things follow.
- Give the reply the most room. Everything else is smaller or quieter.
- Show one reply at a time. A list is homework.
- Write nothing in it you couldn’t say out loud. No “you could say”, no “consider asking”.
- Write it in the first person, in the seller’s voice. The output is the reply itself, ready to read aloud, not a summary the seller has to reword mid-sentence.
That last one is Axelize’s own rule, and it’s the one that’s hardest to hold. Every product instinct says to add context around the answer. On a call, context is the thing you don’t have time for. We wrote more about what a copilot has to get right in the seconds after an objection if you want the sales side of this.
Questions people ask
Frequently asked questions
The reply is written as the thing you would say, in the first person, with no 'you could say' wrapped around it. You read it as it is.
Axelize reads Google Meet captions. It doesn't join as a bot and it doesn't record audio or video.
Axelize works with Google Meet. Zoom and Microsoft Teams are on the roadmap.