How to grade a sales call with AI
Give the AI the call transcript and a scorecard with points for each part of the call, and make it quote the call for every score it gives. Without a scorecard and quotes, AI tends to say the call went well and hand you general tips. The RevOS scorecard grades a closer's call out of 100 across seven parts, and it's what RevOS uses when you say "Review my call with [name]."
A lot of reps go months without real feedback on their calls, because nobody has time to listen back. AI can do the listening for you, but only if you give it a scorecard and make it quote the call.
What do you need before AI can grade a sales call?
You need a transcript of the call, or at least good notes, and a scorecard that spells out what a good call looks like. Most call recorders give you a transcript, and that's what the AI reads.
Without a scorecard, the AI grades against its own idea of a good call, which is usually generic. With one, you can compare this week's calls to last week's on the same scale.
What should a closer's call scorecard measure?
A closer's scorecard should measure how deep your discovery went and whether you found the doubt the buyer never said out loud, plus how you handled objections, how you closed and whether the next step got locked in. The RevOS closer scorecard splits 100 points across seven parts, and discovery carries the most weight with 40 of them.
| Part of the call | Points | What it checks |
|---|---|---|
| Discovery depth | 20 | How far past the first answer you got, to the real reason they're on the call. |
| Discovery coverage | 20 | Whether you learned everything you need to make an offer that fits them. |
| The unspoken doubt | 15 | Whether you found the worry they didn't say out loud and brought it up. |
| Objection handling | 15 | Whether you got to the real concern behind the pushback and answered it. |
| Pitch and close | 15 | Whether you said their situation back until they agreed, pitched briefly in their own words, named the price and asked for the decision. |
| Agenda and frame | 10 | Whether you said at the start how the call would run, got their OK, and then led calmly, like the expert in the room. |
| Next step locked | 5 | Whether the next call has a purpose and a reason for them to show up, and was booked before you hung up. |
| Total | 100 |
How do you keep an AI call grade honest?
Make the AI quote the call for every score it gives, and don't accept a score it can't back up with something you or the buyer actually said. If it can't point to the moment, it's guessing.
The second rule is that the cost of waiting is the test. If the buyer never said, in their own words, what it costs them to keep things the way they are, the discovery wasn't deep enough, even if the call felt great.
What prompt can you use to grade a call with ChatGPT or Claude?
Paste the transcript, then the scorecard with its points, and tell the AI it can't give any score without a quote from the call. Then ask for the total out of 100 and the one fix that would help most on your next call.
Here is the transcript of a sales call I ran: [paste the transcript] Grade my call out of 100 with this scorecard: - Discovery depth (20) - Discovery coverage (20) - The unspoken doubt (15) - Objection handling (15) - Pitch and close (15) - Agenda and frame (10) - Next step locked (5) Rules: - Quote the call for every score. No quote, no points. - Check whether the buyer said, in their own words, what it costs them to wait. - Give me the total, then the one fix that would help most on my next call.
Run it on a few of your recent calls before you change anything, so that you know your starting score.
How should setters and SDRs grade their calls?
Setters and SDRs run a different kind of call, so they need a different card. The RevOS setter card has five parts: the opener, the reason to talk, qualification, whether the meeting got booked, and the handoff notes for the closer.
I like that the handoff notes get graded, because the closer walks into the next call with whatever the setter wrote down.
What should you do with the score?
Pick the one fix and practice it on your next call. If you try to fix everything at once, it's hard to tell what actually helped, so grade the next call the same way and see if that part of the score moved.
The same transcript is also the best place to write your recap email from, because it has everything the buyer said while the call is still fresh.
How does RevOS grade your calls?
In RevOS, you say "Review my call with [name]" and it reads the transcript or your notes, then gives you a recap email for the buyer and a score out of 100 with one fix. Before the next call, "Help me with my next call" gives you a one-page prep with what was said last time and a good first question.
Every score follows the two rules above, so no points go in without a quote from the call. You can also bring your calls to the live coaching every Tuesday at 9 am and Thursday at 2 pm Pacific, and we'll work on them together.