The most valuable hour in your firm happens in a black box, and it always has.

Vinyl now has somewhere between fifty and sixty thousand hours of recordings of that hour. Client meetings, advisory sessions, board meetings, discovery calls. Indexed, transcribed, searchable. Anonymised, and we don't train models on the content. As best I can tell it makes us the largest single repository of what actually gets said between accountants and their clients anywhere in the world. Joe McCord asked me on PM,ID: Pinch Me I'm Dreaming what we've done with it. The honest answer is not enough yet. But sitting with the number for a few months has changed how I see this profession.

The choice nobody made on purpose

The practice management software industry has spent two decades building beautiful tools around financial data and ignoring the thing that actually generates the relationship. Time entries, WIP, lodgements, debtors, jobs, invoices, all real-time, all integrated, all gorgeous. Cloud accounting won that war years ago.

Meanwhile the conversation between you and the client, the thing the entire engagement is built on, has been treated as ephemera. Not tracked. Not searchable. Not surfaced. It sat in the partner's head, or in a notepad, or in nothing at all once the meeting ended.

That wasn't an oversight. It was a choice nobody made on purpose. The profession built systems around what it could measure, and conversation wasn't measurable in 2005, so the industry decided it didn't count. The vendors didn't build for it. The standards didn't require it. The partners didn't demand it. Everyone agreed, quietly, that the most important hour of the client relationship would happen in a black box and nobody would ever look inside.

It's now the single largest unmonitored asset in your firm.

What becomes possible once you can see inside

Three things become possible once you have this data, and the early signal on all three is striking enough to talk about.

The first is something the system is designed to catch and that I expect a lot of firms will find uncomfortable when they start looking. The same opportunity gets raised multiple times, across multiple meetings, across multiple months, and never gets quoted. Not by one partner having a bad week. By different partners, to the same client, sometimes across an entire year. A client mentions restructuring in March. Mentions it again in June. Mentions it in September, this time to a different partner. No scope goes out. No quote gets sent. The opportunity sits in three separate partner memories and quietly dies in all three. That isn't a follow-up problem. It's a firm-level information failure that no human memory was ever going to solve, because no human was ever going to be in all three of those meetings.

The second is that clients telegraph what they need and the profession has been deaf to it. When you ask a business owner how comfortable they are with their cash position and they give a short, hedging answer, that's information. Their financial literacy on that topic is low. They would benefit from support there. You did not need to ask them to rate themselves on a scale. The conversation already told you. Multiply that across every client, every quarter, every topic, and you have something no firm currently possesses, a map of where your client base actually needs you.

The third is the one I find quietly fascinating. Generic AI tools miss most of this because they're horizontal. The classic example is the word extension, a generic model has no idea whether a client means a Chrome extension or a tax extension. That's one example of many. It compounds across every signal in a conversation, and across fifty thousand hours of them, the gap between generic and trained becomes the difference between noise and intelligence.

Firms don't need another dashboard

A lot of AI vendors are about to waste this moment, so I'll be direct about it.

Firms do not need another dashboard. They have enough dashboards. Telling a partner we picked up ten signals in this morning's call and leaving them to act on it is the same problem as before with extra steps. The partner is busy. The partner was busy when they missed the signal the first time. Giving them a list of signals they missed does not solve the problem that made them miss the signals. It just makes them feel worse about it.

What firms need is the next action, drafted and ready for approval. The next-steps email already written. The file note already produced. The scope of work for the new engagement already waiting in the inbox. The board minutes, and this is the one that genuinely surprised me when I first saw firms doing it, produced in seconds rather than the two hours of post-meeting writing that used to be standard. Partners are walking out of board meetings with the minutes already done. They get an evening back.

The unlock isn't spotting the opportunity. It's moving on it before the week swallows you.

Why your clients actually leave

Ask any business owner why they changed accountants and the answer is almost always some version of I tried to call them and they didn't come back to me. I emailed and never heard anything. It's not technical competence. It's not pricing. It's responsiveness, and the feeling that the firm was paying attention.

The reason firms can't fix this isn't that they don't care. It's that the follow-up is buried in an inbox full of compliance reminders, webinar invites and internal noise. The partner means to circle back. They genuinely mean to. Tuesday becomes Thursday becomes never, and three months later the client is sitting across from another firm asking the same question.

AI does not fix the relationship. The relationship is still yours, and it has to be. What it changes is the volume of small, attentive follow-ups your firm can deliver between meetings without burning out the people delivering them. The reminder you said you'd send. The check-in to see whether they did the thing. The follow-up question that shows you were actually listening in March. The firms that figure this out will keep their clients. The firms that don't, won't, and the reason will be exactly what it has always been.

What we've chosen not to build

Sitting on fifty thousand hours of client conversation data is not the same thing as being able to use it. Privacy considerations are real. Client confidentiality obligations are real. The question of who owns a transcript when a client moves firms is genuinely unresolved in Australian law, and the consent question varies by state.

We have deliberately not built the cross-firm intelligence layer that some of our investors would like us to build. The security architecture would have to work in ways we are not yet satisfied with, and the legal framing in this country isn't settled enough for us to be comfortable racing ahead of it. There is a version of this product that doesn't wait for those questions to be answered. That version will get built by someone. It won't be us.

That's worth saying out loud because the loudest AI marketing in this space right now skips past the question entirely, and the profession deserves better from anyone selling into it.

Start treating the conversation as an asset

You don't need to use Vinyl to act on any of this. The argument is bigger than the product.

Record every client meeting, with consent. Make sure the transcript lands somewhere the whole firm can search, so the conversation outlives the partner who attended it. Then start asking the questions you couldn't ask before. What did we promise this client that we haven't billed for. Where are clients showing low confidence on a topic we should be advising on. Which conversations from last quarter contained an opportunity we never followed up on. What does the pattern in our discovery calls actually look like across a year, rather than across the partner who happens to remember best.

The revenue is already in the room. It has been the entire time. The profession has been measuring everything around it for twenty years and ignoring the thing in the middle.

That's the part I think is about to change, and the firms that change with it are going to look meaningfully different to the ones that don't.

The full conversation with Joe McCord is on PM,ID: Pinch Me I'm Dreaming.

More like this