How would you know if the real record is in the inbox?
There is a quick test. Choose a matter that closed six months ago, worked on by someone who is out today, and ask a colleague to answer three questions from the file alone: what was the client told, what did the client decide, and why was that the advice. If the honest answer is that they would need to go through somebody's email to be sure, then the file is not the record. It is a summary of the record, and the record sits in a mailbox only one person can read.
Most firms of 20 to 200 people know this and treat it as normal. Mail gets filed when there is time. Attachments get saved when they matter. The thread where the client changed their mind on a Friday afternoon never makes it across, because nobody thought of it as a document. The file ends up holding the formal outputs, the letters, the reports, the engagement terms, while the reasoning that connects them stays in email.
That gap is not a tidiness problem. It is three separate liabilities the firm has never put a number against.
What does it cost the first time somebody is unavailable?
Illness, resignation, holiday, a client who rings about a matter nobody can place. Each turns a private record into an operational problem in a single morning. Someone needs access to a mailbox, which usually means a partner's sign-off and an IT ticket. Then they have to read it, and reading a mailbox is slow, because email is organised by time rather than by subject. One thread carries three matters and a lunch arrangement.
The cost shows up as time, not as loss, which is why it goes unnoticed. Two people spend half a day piecing together a story that was complete in one head a fortnight ago. The client gets a holding response. We have written separately about what else walks out of the door when an experienced person leaves, and this is the most immediate part of it: the knowledge is not even in people's heads, it is in an account.
Why does supervision quietly stop working?
Supervision depends on being able to look. A partner checking how a junior is handling an engagement looks at the file. If the file holds the outputs but not the exchanges, they are reviewing the finished letter and not the advice that led to it. They cannot see the question the client asked twice, or the caveat dropped between the second and third draft.
Professional bodies across these sectors, the SRA for solicitors, ICAEW and ACCA for accountants, the FCA for advice firms, RICS for surveyors, all expect work to be supervised and records kept in a way that supports it. None of them cares where your mail sits. What matters is whether the firm can demonstrate what happened. A firm whose demonstrable record is thinner than its actual record has made its own evidence problem. When a complaint lands two years later, the defence rests on files, and files assembled from memory are weak ones. It is the same structural fault as rework: the job is done properly and the firm fails to capture that it was.
What happens when the client asks for their data?
This is the part that usually moves a partner from mild interest to action. Under UK GDPR and the Data Protection Act 2018, individuals can ask what personal data a firm holds about them. If the only reliable answer involves searching several mailboxes, the firm has a problem of scale and a problem of accuracy. Mailboxes hold personal data about people who are not the client: the other side, a relative, a witness, a candidate, a tenant. They hold drafts and asides never intended for a file. They hold duplicates, so nobody can be confident everything has been found.
Retention has the same shape. A firm that has set a retention period for matter files has usually not applied it to mail, so material that should have gone years ago sits in a folder marked 2019. If the firm is also starting to use AI on client work, that sprawl is the first thing anyone sensible asks about, which is why data protection on AI client work begins with knowing where the data lives, not with the tool.
Why doesn't a filing policy fix it?
Every firm with this problem already has a policy saying mail must be filed to the matter. It is not being ignored out of laziness. Filing is unpaid work that happens after the real work, at the exact moment the fee earner is already late for the next thing. Any process that depends on a tired person doing an extra step at the end of the day will fail, and the failures cluster on the busiest matters, which are the ones that most need a record.
More training does not change that arithmetic. Nor does a better folder structure. What changes it is removing the extra step, so the file is produced by the work rather than after it. That is the argument in a tool is not a process, applied to records. You can buy storage, search and retention controls. None of them decides what belongs on the file.
What does it look like when the file is a by-product?
Three things change in practice.
- Mail is sorted on arrival, not at the end of the week. Messages are matched to a matter, the ones carrying an instruction or a decision are flagged as such, and the rest are treated as correspondence. That is the job email triage does.
- The note is drafted from what happened. Instead of a fee earner retelling a week of exchanges, the running record is drafted in the firm's own format from the material itself, and the fee earner corrects and approves it. A named person signs off anything client-facing. That is file notes and case notes.
- The trail builds itself. Who decided what, when, on what information, captured as the matter runs rather than assembled at audit time. See audit trails and compliance evidence.
The effect is that the inbox becomes what it should be, a channel rather than an archive. Mail is how things arrive. The file is what the firm holds. In most firms those have been the same object for twenty years, and separating them is the whole job.
Where should a firm start?
Not with a migration. Moving mail into a document system reproduces the problem in a tidier place. Start by deciding, for one type of matter, what the file must contain for someone else to pick it up cold. Six or eight items, not forty. Then rebuild the single step where those items are currently lost, which is usually either the handling of incoming mail or the note after a conversation.
To size it, ask fee earners how long they spend each week filing, hunting for a message somebody else sent, or rebuilding a history for a colleague. Multiply that out using the method in the unbillable hours guide, or let the calculator do it. The continuity and data protection exposure sits on top of that figure and is harder to price, which is no reason to carry it. The audit will tell you whether this is the process to rebuild first.