Quality Assurance for Insurance Brokers: Beyond Random Call Sampling

ASIC secured a record $830 million in court-ordered civil penalties in the 2025-26 financial year, roughly eight times the $104.1 million recorded the year before, with $643.5 million set to be delivered back to tens of thousands of customers and investors through remediation, refunds and payments in connection with ASIC's work, according to ASIC's own enforcement figures. One of the penalties in that total was a $10.3 million order against Mercer Super for systemic reporting failures, including failures to report significant breaches to ASIC under the reportable situations regime, highlighting the consequences that can follow from gaps in an organisation's own reporting and monitoring systems.

Quality assurance, in an insurance brokerage context, means the structured, ongoing review of client calls against defined standards: what was said, how it was said, and whether the conversation met the client's needs and the licensee's obligations at the same time. It sits at the intersection of two things brokerages usually manage separately: staff development and regulatory monitoring. Coaching conversations improve when they are grounded in real call data rather than a manager's memory of a handful of interactions, and the same data can double as evidence that supervision is being carried out consistently.

In many client-facing businesses, including insurance brokerages, QA still means a supervisor pulling a handful of calls each week or month, scoring them against a checklist, and moving on. That approach made sense before better tools existed. It becomes harder to sustain as call volume grows and as ASIC's court-ordered civil penalties continue to climb. The training angle and the compliance angle usually get discussed as though they are separate problems with separate owners. In practice they draw on the same raw material: the calls themselves. A team lead reviewing calls to prepare for a coaching session and a compliance manager reviewing calls to prepare for an audit are often looking at the same conversations for different reasons. Treating QA as a single, shared discipline rather than two disconnected activities is what makes both purposes easier to run well.

A licensee's obligations under the Corporations Act include three separate things: making sure the financial services it provides are delivered efficiently, honestly and fairly; having adequate resources to provide those services and to carry out supervisory arrangements; and, unless the licensee is a body regulated by APRA, maintaining adequate risk management systems.

ASIC's regulatory guide on meeting those obligations describes what the regulator looks for when it assesses whether a licensee's monitoring and supervision arrangements are adequate for the size and complexity of the business.

Contact centre benchmarking puts informal manual sampling at around four calls a month per agent, a figure from older, general call-centre research rather than an insurance-specific study, predating most of today's automated scoring tools. Even taken as a historical reference point rather than a current rule, it illustrates the scale of the gap: for a broker handling twenty or thirty calls a week, a monthly sample of four covers a small fraction of what is actually said to clients. That is not a reflection of effort: reviewing a meaningful proportion of calls manually takes far more compliance resource than most brokerages carry, and building that resource has traditionally meant hiring more reviewers.

The statistical limitation compounds the resourcing one. A small sample can fail to capture a recurring issue, particularly one that only appears with a specific product, representative or type of client conversation. By the time a pattern surfaces in a complaint or an audit, the same conduct may already have occurred many times across calls nobody reviewed. That gap between what is happening on calls and what compliance teams actually see is where the hidden risk in unreviewed conversations tends to build, not in any single call that does get reviewed.

There is also a training standards dimension that sampling struggles to evidence. ASIC's guide on training standards for financial product advisers continues to apply to representatives giving general advice, and personal advice on general insurance and consumer credit insurance products, among the categories ASIC identifies. A small sample of calls can offer some evidence of how a representative applies their training in a given interaction, but it says little about whether that holds consistently across every representative and every product line. Broader call coverage can give a licensee an additional way to assess how representatives apply their training in real conversations, rather than relying on training records alone.

Random call sampling made sense as an approach when reviewing every conversation manually was not realistic. It becomes harder to sustain as ASIC's court-ordered civil penalties continue to climb and as brokerages grow past the point where a handful of reviewed calls a month can represent what is actually happening across a team.

A QA programme built around full call coverage, running training and compliance scoring side by side, gives a brokerage two things at once: a stronger coaching foundation and a stronger evidence base for its compliance arrangements. Neither one has to come at the expense of the other, and neither has to depend on how many hours a reviewer happens to have free in a given week.

The shift from sampling to full coverage is not really a technology decision first. It is a decision about what a brokerage wants to be able to say about its own calls: that a handful were checked and looked fine, or that every one was reviewed against clear training and compliance standards. Callyx.ai is built to run that second version automatically, turning the calls a brokerage already records into the evidence base both its coaching programme and its compliance framework actually need.