An ask was recorded on 34 of 8,479 scored calls
Read end to end rather than sampled, the clearest revenue pattern in this corpus was not a bad pitch or a hard objection. It was a conversation that simply stopped.
On 8,395 of 8,479 scored calls, the check that looks for a next-step ask found nothing to record.
What it costs you: an absent ask is the one sales defect that never produces an artefact anybody reviews, which is how it runs for four months in plain sight.
I am an AI product engineer at Metricsense and I built the pipeline that produced that count. The corpus is 9,218 calls from one anonymised Australian life-insurance business, 16 April to 10 August 2026, queried on 11 August 2026. What you should keep from this is a way of seeing absences at all, plus the one sentence I would put in a call guide.
An ask was recorded on 34 of 8,479 scored calls
Here is the distribution, because the shape matters more than the headline.
scored calls recorded the adviser asking for a next step. On another 50, the customer volunteered one unprompted. On 8,395, neither happened.
The 34 splits further: 20 where the customer said yes, 13 where they deflected, one conditional. That is the entire measured record of anyone asking for a next step across four months of an operating call floor.
This is a measurement claim, not an accusation
I want to spend a section on this before the number gets away from me, because the honest version is considerably narrower than the interesting version.
What the check recorded is that no next-step ask was present in the transcript. That is not the same as an adviser failing to close. A customer may have booked a follow-up by email an hour earlier. The ask may have been phrased in a way the check does not recognise, which is a defect in our wording, not in the adviser's. And a large share of these calls are service calls where asking for a next step would be strange.
So read the number as a floor on a gap, not as a verdict on a team. What survives all those caveats is still surprising: on a call floor that sells, an explicit ask is close to absent from the record.
The unasked close: a call that ends with the customer informed, satisfied and un-progressed. No objection was raised, because nothing was proposed. It is the most common non-event on a sales floor and the hardest one to notice.
Absence is the one thing call review cannot see
Play a call where nobody asked for a next step. It sounds fine.
The adviser was warm, accurate, patient. The customer got their answer and thanked them. A QA reviewer scores it well, and they are right to, because every single thing that happened on that call was done properly. The problem is the thing that did not happen, and nothing on a scoring sheet has a box for it.
This is why absences need coverage rather than sampling. A present defect is visible in one call: you hear it, you flag it, you coach it. An absent one is only visible as a rate across a population, because a single instance of nothing is indistinguishable from a call where nothing was needed.
One call cannot tell you that something is missing. Eight thousand can.
A 3% sample would find one of the 34
Work the arithmetic and it becomes clear why nobody has spotted this in a QA meeting.
An ask was recorded on 34 of 8,479 scored calls, which is about one call in 249. A 3% sample of that population is 254 calls, so a QA analyst reviewing diligently for four months would expect to encounter roughly one call containing an explicit next-step ask.
One. From a single example you cannot tell whether asks are rare, whether this adviser is unusual, or whether you happened to pull an odd call. You certainly cannot establish a baseline, and without a baseline there is nothing to notice a change against.
The cheapest problem on your floor is the one nobody said
Set against everything else in a compliance and quality programme, this defect is unusually easy.
It needs no system change, no new disclosure, no legal review and no product decision. It is not a conduct risk: nothing improper happened, which is precisely why it survives. Fixing it costs one sentence at the end of a call guide and a fortnight of people remembering to use it.
Compare that to any other line on your improvement list and it is the cheapest thing there. It stays unfixed because it is invisible, not because it is hard.
Four checks to run before you believe any of this
My number is from somebody else's floor. Here is how to find out whether it describes yours, using things you already have.
- Read the last ten calls that did not convert, to the end. Not the middle, where the interesting objection is. The last ninety seconds, where you find out how the conversation was allowed to finish.
- Count how many ended with a proposal rather than a summary. A proposal has a verb and usually a date. "I'll send that through" is a summary. "Shall we speak again Thursday?" is a proposal.
- Ask three advisers what they say at the end of a good call. If you get three different answers, there is no wording. If you get three versions of "I let them go away and think about it", there is a wording and it is the wrong one.
- Check whether your CRM can even answer this. Most next-step fields are filled in after the fact, by the person whose behaviour you are trying to measure. That is not a measurement, it is a self-report.
What we still cannot tell you about those 8,395 calls
We cannot tell you what the ask would have been worth. We have the absence and no counterfactual. Anyone offering you a revenue figure attached to a number like this has produced it by multiplying assumptions together, and we are not going to do that.
We cannot separate "should have asked" from "correctly did not". The check does not know which of the 8,395 were service calls where an ask would have been inappropriate. Segmenting that properly is work we have not done yet, and it is the first thing I would want in a second pass.
One business, one window, one definition of an ask. Change the definition and the number moves. That is a real limitation, and it is also the reason the check is written in a plain-English sentence you can read and disagree with.
The fix is one sentence, and it is not a close
You do not need a sales methodology. You need the call to end with a proposal rather than a summary, and the wording can be almost aggressively mild.
Constructed, not from a real call: "Before we finish, would it help if I sent you a written quote and we spoke again on Thursday?" That is it. It carries no pressure, it makes no recommendation, and it moves the conversation from information to a decision with a date attached.
Then measure the rate, not the calls. If an ask is currently recorded on one call in 249, the first honest goal is not a target percentage. It is being able to see the number at all, on a Monday, without anyone choosing which calls to listen to.
AI Product Engineer, Metricsense
Krunal Ambaliya is an AI product engineer at Metricsense. He builds the data pipelines, cloud infrastructure, and Agentic AI systems that power the platform, turning thousands of unstructured reviews, tickets, and call transcripts into structured, actionable intelligence. His focus: making sure every insight links back to the exact customer quote so product teams can act on evidence, not assumptions.
Find the ask your calls are not recording
A pilot runs this check, and any other you can describe in a sentence, across a window of your own calls, in your own cloud and region. Worth doing if your pipeline looks thinner than your call volume suggests it should. If your team already tracks next-step rate from your CRM and trusts the number, you do not need us for this one.
Keep reading
Avesta Labs is an OpenAI Select Partner. Your calls stay put.
The company that builds Metricsense has joined the OpenAI Partner Network. Here is the short list of what that changes for the recordings you have trusted us with, and the longer list of what it does not.
A confidence score is not evidence
0.94 tells you how sure a model is. It tells you nothing about what your adviser said. Only one of those survives a regulator.
When everyone makes the same mistake, it is the script
One adviser getting it wrong is a coaching conversation. The same gap across the whole floor is a wording problem, and coaching will not touch it.
