Metricsense, on the Avesta Labs stand.

The Avesta Labs stand is at Retail Show Australia, on stand 489, where the conversation has moved past the pilot. The questions on that floor are about labour costs, store productivity, loss prevention and supply chain visibility, and about what it takes to run any of it with AI in a way that can still be checked afterwards.

The Metricsense angle in a retail business is the customer record: app reviews, support tickets, complaints and post-purchase surveys. Most of it gets read as a sample, or skimmed by star rating, which is how a store-level or product-level problem stays invisible while the average holds up. If that sounds like your reporting, come and tell us what you are trying to see.

Avesta Labs ran a free two-hour session at UTS, with UTS Startups, on the gap between the two. Personal use and organisational impact are different things: one saves a person a few minutes at a time, the other changes how the work gets done. The session covered the stages of adoption, how to choose a first agentic use case, how much autonomy to hand over, and the controls that make handing it over safe. No technical background needed.

The Metricsense half of that argument is the unglamorous one: you cannot manage what nobody has read. If you are working out where to start, the conversations you already record are usually the cheapest place to look, because the record of what is going wrong is sitting in them already. That is still the conversation we would rather have, at a stand or on a call.

Most conversations about AI open with which tool to buy. The Avesta Labs stand at the MCEC spent three days asking a better one: which part of your business is still held together by somebody moving data between systems by hand? That is the work agents should be doing. People brought their slowest manual workflows to Stand H5 and we talked each one through, including the ones where the honest answer was not to automate that yet.

This was the show on the list where Metricsense was the thing we came to talk about. Australian contact centres, Australian regulators, and a room full of teams trying to get a pilot into production without handing their recordings to somebody who cannot say where they went. If you run quality assurance on a sample and you have wondered what the rest of the calls would say, that was the conversation at H5, including the part where the honest answer is that a sample is adequate for your volume.

The floor at the Helipad Exhibition Centre was machines: granulators, packaging lines, cleanroom equipment, analytical instruments. What the Avesta Labs stand was there to talk about was the work around them. Deviations, CAPAs and out-of-specification investigations, quotes, orders, customer replies, vendor follow-ups, all of it still done by hand, by the same few experienced people.

The Metricsense end of that is narrower. Complaints, field reports and support calls are a record of what went wrong that sits outside the quality system, and it usually gets read only once something has already escalated. Reading all of it, with the exact words behind every finding, is a different proposition to sampling it.

Most automation stops at the plant floor. The machines run and the line moves, and the decisions around it still rest on people, spreadsheets and a phone that rings all day. That was the argument the stand made in Mumbai.

The Metricsense end of it is narrow and specific: the calls, complaints and tickets around a plant are a written record of what is going wrong, and almost nobody reads all of it. A quality manager reads the ones that escalated, which is the sample that has already failed.

The demo is the easy part. Getting a system to behave the same way on the ten thousandth input as it did on the first is the part that decides whether any of it survives contact with a real business. That is what the stand at Bharat Mandapam was about.

For Metricsense the same point has a narrower form. Reading every conversation is not the hard bit. Reading every conversation and being able to show the exact words behind each answer, months later, to somebody who does not trust you, is the hard bit.