What Your Phone System Reports Natively, and What Sits Outside It

Most of what a brokerage principal needs to know about the phone is already being recorded, automatically, by the phone system itself. Not in a report anyone commissioned. As a by-product of running the service.

A hosted PBX counts every call that arrives and every call that leaves. It knows when each one came in, how long it ran, whether it connected, which published number it arrived on and what it cost. It tracks how long callers waited in a queue and how many gave up before anyone answered. In a separate report it shows which staff member took which calls, and how long they spent on the phone. For a firm trying to answer an operational question about capacity, that is a genuinely good data set, and it is better than it is usually put to use.

The reason it is underused is rarely the data. For many firms it may simply be that the reporting lives in a portal there is little reason to open between quarters, and the questions that would send someone there tend to arrive at the busiest time of the year. Renewal season raises staffing questions, and the answers have been accumulating in the system since the day it went live.

What that record cannot do is describe a conversation. It can tell you a renewal call ran for 14 minutes on a Tuesday afternoon and connected on the second ring. It cannot tell you whether the premium movement was explained in a way the client accepted, whether the client's changed circumstances were revisited, or whether an objection was raised and left hanging. Those sit in the audio, and the audio is a different kind of thing from a row in a table.

That distinction is the whole subject of this article, and it is worth getting straight before any reporting decision is made, because the two sides fail in opposite directions. Reading the phone system's numbers as a quality signal can produce confident conclusions from data that was never measuring quality. Dismissing the numbers because they are not a quality signal throws away the best capacity data the firm has. The standard call metrics a firm tracks are worth reading closely, and they are worth reading for what they actually measure.

One practical point sits underneath all of it. What the reporting can separate depends on how the system was configured at the beginning. A brokerage with one published number and one queue will read total volume accurately and will find call type hard to isolate. A brokerage running separate renewal, claims and service numbers gets that separation in every report for free. Neither configuration changes what the phone system can see. It changes how finely the firm can slice what it already has.

MaxoTel holds call information in three separate places: two reporting surfaces, and the store that holds the audio. They are documented as separate capabilities and they behave differently, and a brokerage that knows which is which can send each question to the place that can answer it.

The call history export. One row per call, covering inbound and outbound calls on the service. The account portal's call history view takes a start and end date, displays the matching calls with type, time, origin, destination, recording, rate, status, duration and cost columns, and offers the same set as a CSV download.

The same records come out through the Call History API in CSV or JSON for anything that needs to repeat. This is the capacity and reachability layer, and it is the one usually reached for first.

How to read that export properly, including the views worth building and the caveats that come with each field, is covered in detail elsewhere on this blog.

The queue and agent statistics. A separate on-screen report, filtered by date range, call queue and dialled number. The PBX statistics report carries the measures the export does not. Per queue and per number dialled, it reports total calls, how many were answered, how many reached voicemail, how many were abandoned, and the average, shortest and longest wait. Per agent, it reports total calls taken, the share of queue calls each agent answered, how many calls the agent ended first, total and average duration, and the shortest and longest call in the window. It also charts wait times and abandonment against incoming volume in 30-minute blocks. MaxoTel's definition in the queue breakdown is specific and worth holding: the caller hung up before being answered by an agent or leaving a voicemail.

Where a brokerage wants to understand queue performance during a busy week, this report rather than the export is the right place to look.

The recording store. Where the audio lives. Call recording, when enabled, captures inbound and outbound calls on the PBX by default, with per-extension exclusions and per-call controls available. Recordings are retrieved from the same call history screen, and retention runs on two settings working together: a disk space quota, where the oldest recordings are overwritten once the quota is exceeded, and a delete-after period that applies while quota remains. MaxoTel notes that a recording removed by the retention policy cannot be retrieved by its team, and recommends enabling backup to Google Drive, Dropbox or an FTP server, which runs overnight.

One line runs through both reporting surfaces. Every field in the call history export and in the queue and agent statistics describes the container rather than the contents. Duration says a conversation ran for 14 minutes. Wait time says a client held for 48 seconds. The recording flag says audio exists. Agent duration says a broker spent 90 minutes on the phone before lunch. Not one of those fields says anything about what was discussed.

The third place is different in kind, and the distinction is worth drawing precisely. A recording does hold the conversation, in full, which is the entire reason it is worth keeping. What the platform does with that audio is store it, apply a retention policy to it, back it up and hand it back on request. Reading it is left to whoever opens the file.

That is not a gap in MaxoTel. It is what a phone system is. The nearest the current documentation comes to reading content is automated speech to text on voicemail messages, which transcribes the message a caller left rather than the conversation a broker had.

The published feature list documents no transcription, content analysis or sentiment analysis for recorded calls. Routing, connecting and recording calls reliably is a different engineering problem from understanding them, and a platform doing the first job well has no particular reason to reach the second on the same timeline.

Questions about the phone tend to arrive together rather than sorted by which system can answer them, usually in the same meeting and usually in a busy week. Four situations come up often enough to be worth naming.

None of these is a reporting failure. Each one is what happens when a question that needs the audio gets routed to the layer that holds the metadata, and the fix is a routing decision rather than a better spreadsheet.

A brokerage reading both layers well tends to look much the same from the outside. The difference shows up in how quickly a question gets to a real answer.

Callyx.ai sits alongside the phone system rather than replacing any part of it. The PBX keeps routing calls, keeps recording them and keeps producing the same call history and queue statistics it always has. Nothing about how brokers take calls changes, and no step is added to the call itself.

What changes is that the audio becomes readable without anyone having to listen to it. Callyx.ai works on the recordings a business already holds, transcribing and analysing them, so a call that currently reads as 14 minutes and connected on Tuesday becomes a conversation with a summary, extracted action items and a record of what was discussed.

The boundary is worth stating as plainly as the capability. Callyx.ai does not route calls, does not run queues and does not replace the phone system's reporting. The capacity question still belongs to the PBX, and a firm that wants to know when calls arrive and how long people wait should keep reading the queue report. What Callyx.ai adds is the layer the phone system was never built to produce.

A hosted phone system reports more than it is usually asked for. Across the call history export and the queue and agent statistics, MaxoTel produces a record of the calls on the service, how long callers waited, how many abandoned and who took which calls, and where call recording is enabled it keeps the audio alongside that record. For capacity, timing, reachability and staff activity, those two reports cover the calls the system handled rather than a sample someone chose.

The boundary is equally clear. Every field in both reports describes the container. Neither describes the conversation. What was said, whether it was right for the client, whether the client understood: none of it is in any field, because it was never in that data to begin with. The nearest the current documentation comes is automated speech to text on voicemail, which reads messages rather than calls.

A brokerage that keeps those two things separate makes better decisions on both sides. It reads the queue report for the staffing question and gets a real answer. It stops asking the duration column a question about quality that the duration column cannot answer. And it decides deliberately how the content questions get handled, rather than letting them default to whichever number is nearest.

That second layer is where Callyx.ai works. It leaves the phone system doing exactly what it does well, and reads the recordings the business is already keeping, so the call the export counted and the conversation that produced it stop being two separate reading problems.