BronVast Messages | Context & appraisal

A message gains meaning through relationships, public tasks and decisions.

An isolated reply may be impossible to interpret without participants, time, subject and work process. BronVast reconstructs demonstrable relationships and exposes possible context without presenting a proposal as an archival decision.

Specialist reviewing digital message source packages
2dA BronVast

Messages

Relationships

Conversation before isolated line

Messages, replies, events, participants and available attachments remain in their demonstrable context.

Context

Source and interpretation remain separate

Case, process and subject relationships retain source, method, confidence and review status.

Appraisal

Authorised people decide

Rules and AI can prepare candidates; value, retention and exceptions are confirmed by people.

Evidence route

From source package to accountable destination

01

Reconstruction

sequence, participants and events

02

Context proposal

case, process, subject and role

03

Description

metadata with source and confidence

04

Framework

selection list, key role or hotspot

05

Review

confirm, reject or escalate

06

Decision log

authority, reason, date and version

From data to understandable information

BronVast preserves not only text but also the data needed to understand sequence, sender, recipients, group changes and attachments.

  • source time and UTC normalisation side by side
  • reply and attachment relationships
  • participants and aliases as reviewable objects
  • source locator for every normalised message

Several appraisal routes may be relevant

A conversation may contain process records, key-role communication, hotspot material or temporarily relevant messages. The valid framework and actual context determine the route.

  • selection list and process application
  • key role where applicable
  • hotspot or significant event
  • FOI, investigation, dispute or other hold
  • exceptions escalated to an authorised reviewer

Local AI as a proposal engine

A local model can propose clustering, entities, sensitivity, process context and review priority. It records version, source, confidence and human correction.

  • no silent replacement of source metadata
  • low confidence sent to a review queue
  • model card and evaluation set per task
  • no automatic decision carrying legal effect
Deliverables

What your organisation receives

Context map

Why information may belong together

Conversations, people, roles, cases, processes and events with visible evidential strength.

Review queue

Uncertainty becomes manageable

Open proposals, exceptions, low confidence and required authority together.

Decision log

Accountability remains

Who decided what, under which framework, for what reason and under which version.

Reference points

Reference points

These references show the professional basis for our choices. The requirements that apply in practice are determined with you from your situation, responsibilities and receiving environment.

ContextNational Archives guidance on creation and use context.Appraisal methodsWork process, key function and hotspot methods may coexist.NIST AI RMFGovernance and evaluation for responsible model use.
FAQ

Questions about this route

Does AI decide which messages are retained?

No. AI may support proposals and prioritisation. The authorised organisation decides under the applicable framework.

Can one conversation contain different retention routes?

Yes. Aggregation level, context, exceptions and operational feasibility must therefore be determined in advance.

Can users see what a model derived?

Yes. Derived proposals retain method or model version, confidence, source reference and review status.

Next step

Start with evidence, not assumptions

A first engagement can be limited to one lawful source package, one process and one receiving profile.