Email organisation · 2026-10-22
How MailTrace turns emails into a clearer picture
From scattered messages to topics, individual matters and timelines: how free automatic grouping and optional AI assistance work together.
Start with the correspondence, not a filing system
You asked for a refund. You sent three reminders. Someone acknowledged your email, and you later contacted a manager. Weeks later, the story is spread across several conversations. MailTrace’s aim is to bring that correspondence together around the problem you were trying to resolve.
This guide explains the automatic organisation workflow being introduced to MailTrace. Availability depends on rollout, your account and whether optional AI processing is enabled.
A topic is a subject. A matter is one particular problem.
A topic gives you a broad way to browse your evidence. A matter brings together correspondence about a specific request, problem or incident. The topics come from the content of your case, rather than a preset list that every customer must fit.
For example, two expense claims might share a topic such as “Travel expenses”. An unpaid train fare and a separate hotel reimbursement would remain different matters, each with its own evidence and history. These examples are fictional; your case’s subjects should come from your own correspondence.
One email can discuss both claims. Relevant passages can support each matter, so the whole message does not have to be forced into a single group. You can return to the original email to check the surrounding context.
What the free engine does automatically
Free grouping compares clues in the evidence: specific reference numbers, recurring phrases, distinctive wording, participants and the way a conversation continues. This is called deterministic processing: it applies consistent comparison rules to the selected correspondence.
The rules describe how to compare messages. The subjects and descriptive names come from the evidence. Sharing a sender or mentioning a complaint is not enough, by itself, to make two messages part of the same matter.
The result is a saved starting organisation: topics, individual matters, supporting passages and timelines. You do not have to create every group or approve each assignment before that organisation is useful. Where the evidence does not support a clear connection, a passage can remain explicitly uncertain.
What optional paid AI adds
A large language model, or LLM, can help interpret wording and context that simpler comparisons miss. For example, “refund the fare charged twice” and “the duplicate train ticket payment” may describe the same problem without using the same words.
AI assistance can refine topic names, connect paraphrased correspondence, separate similar incidents and assess whether a reply addresses the actual concern. It improves the same saved organisation, preserving continuing matter identities and the history of changes.
AI can also misunderstand a message or make an incorrect connection. Its assessments remain linked to evidence you can inspect. If paid processing is unavailable, the automatic free organisation remains available. Large cases may be refined in stages, with the application showing how much evidence has been assessed.
Reminders and escalation belong to the original matter
An initial reimbursement request, three reminders and a complaint to a manager belong together when they concern the same unpaid expenses. Escalation describes something that happened to that matter; it should not become a separate catch-all group that loses the original concern.
The timeline can show when you raised the issue, repeated it, received an acknowledgement, obtained a substantive response or escalated it. A separate concern keeps its own chronology, even when the same people are involved.
“We have received your email” does not automatically mean the concern was answered. The wording and supporting evidence matter when assessing whether something was answered, partly addressed or remains unresolved.
“No response found” is a statement about the evidence
“No response found” means that a relevant response was not found in the available case evidence. It does not establish that someone deliberately ignored you. A phone call, letter or missing message could change the picture.
MailTrace should make uncertainty visible rather than fill gaps with assumptions. Organising a notification history is also separate from deciding whether it establishes a legal breach.
What changes when new emails arrive?
New selected emails are compared with existing matters. A supported follow-up can extend the original timeline; a different problem can become a new matter. Continuing matters keep their identity so you can return to the same history as the case develops.
Changing a category name does not, by itself, move the underlying evidence. You can optionally correct a name, remove an incorrect link or move supporting passages. Those corrections help organise your case, while saved history lets you inspect and undo supported changes.
Where machine learning, embeddings and privacy fit
Machine learning is part of the optional assistance, rather than a requirement for every grouping step. You may also hear about embeddings: numerical descriptions of text that help find potentially related passages. They are a matching aid. Similar wording alone does not establish that two messages concern the same incident or that a reply resolved it.
Free grouping runs within MailTrace without an external language-model call. Optional AI refinement sends relevant case excerpts to the configured provider with your permission. Gmail evidence and some file operations have additional consent requirements. The privacy policy explains processing, retention and your choices.
How to check the organisation
Open a matter, follow its timeline and read the supporting passages. Check that unrelated problems stay separate, that reminders remain with the right request, and that a claimed response deals with the concern you raised.
The purpose is to give you a useful, evidence-linked starting point. Corrections are optional feedback; checking important conclusions against the original correspondence remains a sensible part of using the result.