Most organisations sort their communication by tone. The sort that matters is by consequence.
Every professional knows the difference between a quick message to a colleague and a letter that commits the firm. We treat that difference as a matter of register, how carefully we write, which greeting we use, whether we sign off with a name or an initial.
That instinct is right about the stakes and wrong about the remedy. The difference isn’t tone. It’s category.
Some communication creates obligations. It establishes what was advised, what was instructed, what was agreed, and when. It can be produced in a dispute years later. It carries legal, regulatory, or fiduciary weight the moment it is sent, regardless of how casually it was written.
That is 'formal communication', and it has been treated as a style rather than a category for as long as we have had email.
What it looks like in practice
A law firm confirming the scope of an engagement. A partner giving instructions on a filing deadline. An asset manager recording a client’s risk tolerance, or receiving an instruction to move money. A trustee documenting the reasoning behind a distribution decision. A bank confirming a change to an account mandate.
None of these are unusual. All of them happen daily, in ordinary language, through ordinary tools. And each one may need to be reconstructed, in full, by someone hostile, long after everyone involved has forgotten the details.
The category is not defined by formality of language. It is defined by what happens when someone asks you to prove it.
Five tests
A channel carries formal communication properly if it passes all five. Most channels in daily professional use pass one or two.
1. Verified counterparty. You can prove who was on the other end, not because they told you, and not because the display name matched, but because their identity was verified before the exchange began.
2. The firm is the unit. The account, the history, and the authority belong to the organisation, not to an individual’s personal device or phone number. A person leaving does not take the record with them.
3. Record by default. The exchange is captured because that is how the channel works, not because someone remembered to archive it. Nothing important depends on a participant’s discipline.
4. Retention by design. How long records are kept is set by obligation, not by preference, and cannot be quietly overridden by a disappearing-message setting or a user pressing delete.
5. Known jurisdiction. You can say which country’s laws govern the data and which courts can compel its production. Not where the server sits, which legal system reaches the company operating it.
What fails, and why
Consumer messengers fail tests 1, 2, 3 and 4, not through poor engineering, but through excellent engineering aimed at a different goal. They are built for individuals: personal accounts, personal control, personal convenience. Every property that makes them good at that makes them wrong for a firm.
Email fails test 1 completely. The sender line is a claim, not a fact, and it always has been. It half-passes 2 and 3, and fails 5 for most organisations, who could not say today which jurisdictions can reach their mail provider.
Phone calls fail 3 and 4 outright. The most consequential sentence in a matter is often the one spoken aloud and never written down.
Document management systems pass 3, 4 and often 5, for documents. They were never designed to hold the exchanges around those documents: the instruction, the caveat, the approval. The vault is secure. The corridor is not.
The result is not one broken channel. It is a fragmented process in which a single obligation-creating exchange might begin in email, continue in chat, be confirmed by phone, and end in a document store, with no single record of any of it.
Why this is surfacing now
Three pressures are converging.
Regulation has moved from encouraging documentation to requiring it, with clients and supervisors entitled to demand records within fixed periods. Documentation duties are now enforceable obligations rather than good practice.
Impersonation has become cheap thanks to AI. Systems that can write in a specific person’s style and clone their voice have removed the tells professionals were trained to spot. A channel that cannot verify identity is now a channel that cannot be trusted with instructions.
And the tooling has multiplied. Every firm now runs more communication channels than it has policies for, and the gap between what is used and what is governed keeps widening.
AI makes impersonation cheaper than ever. The OpenAI Astra AI model is extremely sophisticated in computational tasks; imitation a person or character is an easy job.
Naming it
Categories get built when someone writes the definition. Secure messaging, document management, e-signature, each became a category because a problem was named clearly enough that buyers could ask for it.
Formal communication has not had that. It has been split across a dozen products, none of which was designed for the whole of it, and defended by policies that lose to convenience every time.
The test for whether it deserves a name is simple: can you describe a requirement that no existing category satisfies? Verified counterparty, firm as the unit, record by default, retention by design, known jurisdiction, no messenger, mail client, or document store satisfies all five, and nothing about the list is exotic. It is what regulated work has always implicitly required and never explicitly demanded.
That gap is the category. It is worth naming, because you cannot procure what you cannot describe.
.png)