Why We Keep a Human in the Loop on Outbound Messages

Why does a person still read outbound messages before they are sent?

Across the brands built on our platform, software can draft a message, but it cannot send one on its own. A person reads the draft first and decides whether it actually goes out. We hold this line because a message that reaches a customer, a vendor, or a partner carries consequences that belong to a person, not to a model. The system can propose wording quickly, but judgment about tone, timing, and accuracy stays with a human reviewer.

What does a human reviewer actually check?

The review is not a rubber stamp. A reviewer reads the draft against the situation it responds to and asks whether the content is accurate, whether the tone fits the relationship, and whether anything in the message could be misread. This is deliberately plain work. It does not require special tools, only attention and context that a generated draft cannot fully supply on its own.

In practice, a reviewer is checking for things like:

  • Whether the draft accurately reflects what actually happened or was agreed
  • Whether the tone matches how the recipient should be treated
  • Whether sensitive details are handled carefully
  • Whether the message is ready to stand on its own without further explanation

Does keeping a person in the loop slow the brands down?

It adds a step, and we accept that tradeoff deliberately. Speed is not the only measure of a working system. A message that goes out fast but wrong creates more work later, in corrections, apologies, and lost trust, than the extra moment a reviewer spends catching it beforehand. We would rather a draft wait briefly for a person than go out unreviewed.

What counts as a receipt for an outbound message?

A receipt is the record of what happened: what was drafted, what source or situation it was based on, and when a person reviewed and approved it. This is the same standard we hold for any result our systems produce. A message without that trail is not something we consider finished, even if the words themselves read fine. The receipt is what lets anyone, later, understand why a message was sent and who stood behind it.

How does this show up differently across brands like Fidomi?

Each brand on our platform owns its own product and its own customers, so each brand decides what its outbound messages should sound like and who on that team reviews them. The platform gives every brand the same discipline, a draft does not become a sent message without a person approving it, but the judgment calls inside that review belong to the team running the brand. We do not centralize that decision, because the people closest to a brand's customers are best placed to judge what a message to them should say.

What happens when a draft fails the review?

It does not go out. A reviewer can send it back for revision, ask for more context, or decide the message should not be sent at all. None of these outcomes are failures of the process, they are the process working as intended. The point of having a person in the loop is precisely to catch the drafts that need a second look before they reach anyone outside the company.

Why does this matter more as the system gets faster?

The easier it becomes to generate a plausible-sounding message, the more important it is that something other than plausibility decides what actually gets sent. Our rule is simple and we do not make exceptions for convenience: nothing goes out without a person approving it. As our systems get better at drafting, that rule becomes more important, not less, because the gap between a good-sounding draft and a message that is actually right can be easy to miss without someone checking.

Frequently asked questions

Does a person review every outbound message, or only some of them?

Every outbound message gets reviewed by a person before it is sent. This applies the same way across the brands on our platform, regardless of how the draft was produced.

Who is responsible for a message once it is sent?

The person who approved it. The system can draft a message, but approval and the responsibility that comes with it rest with a human reviewer.

Can a brand on the platform set its own review process?

Yes, each brand owns its product end to end and decides how its review works. The platform requires that a person approve every outbound message, but the details of that review belong to the brand's own team.

What happens to a message that a reviewer is unsure about?

It does not go out until the reviewer is satisfied. A reviewer can ask for changes, request more context, or hold the message entirely rather than send something they are not confident in.

Back to blog