Why Phone Conversations Are the Gap in AI Governance
Every conversation about AI in client meetings seems to start the same way. A notetaker joins the video call. The client notices, asks what it’s for, and the firm explains what’s being recorded, who can see it, and how long it’s kept.
Over the last year we’ve watched firms get genuinely good at answering that question. Consent is being formalised rather than assumed. That’s real progress, and it’s overdue.
But almost all of it is about one kind of conversation: the one that happens over video, where a bot can actually join. It says very little about the conversation that still makes up a large share of client contact in most firms the phone call (or the forgotten call).
A gap hiding in plain sight
AI meeting assistants have caught on quickly because they slot into existing behaviour. They join the call you were already having, on a platform built to let a third-party tool in. A phone call doesn’t offer that opening. There’s no link to join, no room to enter. So a firm’s AI governance framework might carefully cover what happens when a notetaker sits in on a Teams call, and say nothing at all about the callback made from a solicitor’s mobile, or the quick update given over a landline between meetings.
That’s not a small oversight. In my experience, the calls that fall into this gap are often the ones that matter most later. A client rings with an urgent question. A solicitor calls to talk through risk before a decision gets made. These are exactly the conversations most likely to be remembered differently by each side if something goes wrong and exactly the conversations least likely to have any record beyond a rushed note, if that.
Two questions, not one
“AI and client conversations” is really two separate governance questions that keep getting treated as one. The first is about consent and confidentiality when an AI tool visibly joins a meeting, who’s told, what’s recorded, how it’s stored. The second, quieter question is about coverage: which conversations are actually being captured accurately at all.
Nearly all of the current debate is about the first question. Almost none of it is about the second. But a firm’s audit trail is only as strong as its weakest channel. If the video meeting produces a clean transcript and the phone call produces nothing, the documentation problem hasn’t been solved it’s just moved somewhere less visible.
Why this matters
Disputes can often start because there is no clear record or the record is created in a way that can be argued. They start because the records are incomplete, or two people remember the same conversation differently. A verbatim transcript removes that ambiguity. A hurried note, or no note at all, doesn’t.
If a firm has done the work to make its video meetings defensible but hasn’t asked the same question about its phone calls, it’s solved the more visible problem and left the bigger one alone.
The question worth asking about the forgotten call
Most firms writing AI policy right now are focused on the AI bot now joining the meeting room. It’s the obvious, immediate question. But it’s worth asking a second, quieter one internally: of all the client conversations that happen in a week, how many are on the phone and how many of those currently produce any record beyond memory?
For a lot of firms, it’s worth talking and about considering the forgotten call before a customer complaint makes it urgent.