The Compounding Conversation

Regulated businesses can have thousands of customer conversations every week. Most of these conversations just disappear: sometimes reviewed, little ever gained, and then gone forever. The business keeps having the same conversations, making the same mistakes, missing the same opportunities. The ones that pull away from their competitors are not the ones having more conversations. They are the ones getting better from every conversation they have.

There is a concept in finance that most people understand intuitively: compounding. Returns that build on themselves. A small difference in the rate of improvement, sustained long enough, producing an enormous difference in outcome.

This applies to conversations too. Specifically, to the regulated customer conversations that Spoke was built to manage. The calls and messages that financial advisors, insurance adjusters, care coordinators, and high-value sales teams have with customers every day, in environments where what they say matters legally, commercially, and ethically.

After several years of building the infrastructure and watching what happens when it works, most regulated businesses are leaving almost all of that compounding on the table.

It's not because they are not trying. It's because the infrastructure they are using was not built to capture the learning and reuse it.

‍

What does a non-compounding conversation system look like?

Most regulated businesses are running what you would call a non-compounding conversation system. The conversations happen. Some are captured. A fraction of the captured ones are reviewed. A fraction of the reviewed ones generate coaching. The coaching produces some improvement in some advisors. The improvement is real but it is not systematic, it is not measured, and it does not feed back into the next conversation in any structured way.

The system looks like this in practice: Take a financial advisory firm with two hundred advisors having an average of twenty-five client conversations a week. That's five thousand conversations a week, or roughly a quarter of a million a year. The compliance, QA team, and line managers can review roughly 3% of them, or one hundred and fifty calls a week. That's 242,500 missed opportunities dropped on the floor each and every year.

What's worse, the findings from those 150 call reviews go into key decision making:

• Risk reports that go to the compliance committee
• Coaching insights that inform individual feedback for the advisors
• Inputs for key operational decisions

The other four thousand eight hundred and fifty conversations that week produce nothing beyond the call itself.

The 150 conversations that were reviewed do not actually truly inform you of anything. The compliance finding from January's sample does not automatically update the coaching program in February, and even if it did, you are making decisions with stastically irrelevant information. The technique that the best advisor used to close a deal or handle the new product's most common objection, is sitting in the four thousand eight hundred and fifty unreviewed conversations no one looks at. Nothing is really learned and little gets taught to anyone else.

Each conversation, in this system, is mostly a one-time event. It produces an outcome and a record. It does not produce systematic learning or opportunity for improvent in the business.

5,000 financial advisor conversations a week and roughly 150 are reviewed. The other 4,850 produce no coaching, no compliance intelligence, and no learning that changes how the next week's conversations will go. (Source: Spoke analysis of customer data across regulated financial services firms, 2025–2026)

The business is not getting worse from this, usually. It is just not getting better in any systematic way. When every conversation goes roughly the same as the one before it, it looks like stability. In a competitive regulated environment, stable is actually a slow relative decline, because the firms whose conversations are compounding are getting better at roughly the same rate that everyone else is staying the same.

‍

What does a compounding conversation system actually require?

It is easy to describe the outcome (conversations and outcomes getting better over time), without being honest about what it takes to build the system that produces it. The compounding conversation is not a philosophy. It is an infrastructure problem. And the infrastructure has four specific requirements that have to be in place before the compounding effect starts.

Full capture

The compounding loop starts with the conversation. Not a sample of it. Not the conversations that happened on monitored channels. Every conversation, across every channel that the business uses to talk to customers — desk calls, mobile calls, messages, WhatsApp — needs to be inside the system before the loop can start.

This sounds obvious until you look at how most regulated businesses actually communicate. We have written about the mobile compliance gap in detail elsewhere. The short version: in most regulated businesses, a significant proportion of customer conversations happen on channels the compliance infrastructure does not reach. For a financial advisory firm that might be 40–60% of advisor conversations occuring on a mobile phone. For an insurance firm with field-based adjusters it is often higher. You cannot learn from conversations you cannot see. Full coverage is not a compliance objective here, it is a prerequisite for everything that follows. Your communications infrastructure needs to be everywhere, especially on mobile phones.

Consistent review

Full capture gets you the data. Consistent review at full coverage (e.g. reviewing 100% of conversations) requires automation. There is no headcount model in which a human call review or call QA team can review every conversation. A regulated business has thousands of long and often complex conversaitons with its customers. The economics simple do not work and the consistency does not hold at scale.

What automated review provides, when built correctly, is not just coverage. It is consistency. Two human analysts scoring the same call against the same rubric will produce different scores. The same automated system scoring ten thousand calls applies the rubric identically to all of them. The intelligence that comes out is comparable in ways that human QA scores are not.

To be clear about what automated review does not do. It does not replace human judgment in complex situations. It does not assess tone with the nuance a human reviewer brings. It does not catch everything. What it does, consistently and at scale, is identify whether required disclosures were made, whether prohibited language was used, whether the conversation followed the required structure, and whether the patterns of performance across the team are getting better or worse. That is the foundation the compounding loop runs on.

Targeted coaching

Full capture and consistent review are necessary but not sufficient:

1. The learning has to get back to the people who can act on it quickly enough while fresh in mind
2. Just enough informaiton needs to be provided that they know what to change and how to change, but not so much it is overwhelming
3. Everyone needs to get feedback, not just the ones whose calls the manager happened to review
4. Feedback needs to be personalized, incremental, and regular for improvent to take hold

The lag between a conversation and the coaching that addresses it is one of the most significant drags on improvement in regulated environments. A compliance issue identified in week one should be in a coaching conversation by week two. In manual QA systems the lag is typically two to four weeks. In automated systems that generate coaching outputs directly, it closes to days or even hours.

The specificity of the coaching is the other lever. Generic feedback "work on your disclosure timing" does not produce consistent behaviour change. Specific feedback grounded in a specific conversation does. Automated call review software can identify the specific moment. It can identify what peers do well in the same situation, and it can be the guide for the manager's conversation. The manager's time is now spent coaching, not finding calls, listening to them, and creating a coaching plan.

Measurement that closes the loop

The fourth requirement is the one that turns improvement from a "hope", into a system: measurement specific enough to tell you whether the coaching is working, visible enough that the team can see their own progress, and continuous enough that it compounds rather than resets.

Most regulated businesses measure outcomes. Conversion rates, complaint rates, and compliance scores on a monthly or quarterly basis. These are important measures. But they are not the measures that drive the compounding value loop, because they are too aggregate and too lagged. By the time a monthly compliance score shows improvement, the coaching that produced it has been running for weeks without feedback.

The measures that drive compounding value are conversation-level and near-real-time: did the disclosure rate improve this week compared to last, is the improvement consistent across the team or concentrated in advisors who had specific coaching interventions, are the improvements persisting or reverting after the first few weeks. These measures exist only if the review system is generating them continuously. And they close the loop. Improvement is measured, the measurement informs the next coaching intervention, the intervention produces improvement that is measured again, and so on.

‍

What does the compounding effect actually look like in practice?

I am going to describe what we have seen in regulated businesses that have built all four requirements and have been running them long enough to see the compounding effect take hold. Not as a prediction of what every business will experience, but as an honest account of what the pattern looks like when the system is working.

The first three months

The first three months are characterized as 'visibility', rather than improvement. The compliance gaps that were invisible, the scale of disclosure omissions that no one noticed, are now visible.  This is not comfortable.

The first reaction of many compliance teams when they see full coverage data is concern that the situation is worse than they thought. In most cases the situation is roughly the same as it always was, it is just visible for the first time. The disclosure omission rate that appears in the full coverage data were present all along, the manual QA sampling process just did not find it.

The compliance improvement in the first three months comes from closing the obvious gaps, coaching that addresses the most common compliance failures, process changes that remove the friction producing the most consistent errors. These improvements are real and measurable, and they tend to be faster than teams expect, because the coaching is now based on complete data rather than a sample.

While having more visibility into the state of compliance is often uncomfortable, this newly acquired visibility also bring opportunity. New visibility into missed revenue, performance coaching (outside compliance), and process optimization, open up endless possibilities that were previously not possible before 100% coverage and analysis.

Three to twelve months

The compounding effect starts to become visible in months three to twelve. This is when the improvement patterns change character, from closing specific gaps to systematically raising the floor. In compliance, sales, process, and more.

The floor is the baseline performance of the advisors in the middle 80% of the team's distribution. In a non-compounding system, the floor is set by whoever the manager least coached. In a compounding system, the floor rises because the coaching reaches everyone, it is personalized and specific for each individual, and the measurement tracks everyone's improvement against the same standard. That said, these systems are unlikely to radically improve your best people, or your worst. They are very good, however, at taking the 80% of your middle performers and improving them by 10% to 20%. A twenty percent improvement across 80% of your people is a material business uplift.

40% — Typical improvement in required disclosure adherence across a regulated team in the first six months of full-coverage automated review and targeted coaching. The improvement is concentrated in the lower half of the team's performance distribution — the advisors whose conversations were previously reviewed least frequently. (Source: Spoke customer operational data, 2025–2026. Anonymized.)

The gap between top performers and average performers narrows. Not because the top performers plateau, they typically continue to improve. But because the coaching that used to be concentrated on the advisors whose calls the manager happened to review now reaches the advisors whose calls were previously invisible. The technique the top performer uses to handle the most common objection, which previously existed only in their conversations, is now documented, teachable, and being coached.

Beyond twelve months

The compounding effect beyond twelve months is what separates the firms running these systems from the ones that are not. The improvement rate does not slow down the way it does in conventional training and coaching programs, because the system is continuously finding new things to improve. The rubric gets sharper as the coaching team develops a better understanding of what the data is showing them. The coaching interventions get more targeted as the patterns become clearer. The measurement gets more granular as the team develops the confidence to ask more specific questions of the data.

The businesses we have seen run these systems for two or more years have compliance metrics, performance metrics, and manager efficiency metrics that are not marginally better than industry benchmarks. They are significantly better — in ways that are difficult to replicate quickly because they are the product of compounding improvement over time, not a single intervention.