How we use AI

AI runs the mechanical layer of our work. People own every decision that matters. This page says exactly where the line is — because no business should hand its operations to a system it can't inspect.

What our agents do

Read, summarise, classify and file documents and correspondence
Move data between systems, keep records current, reconcile routine items
Track diaries, deadlines and obligations, and prepare reminders
Prepare drafts — reports, correspondence, management information — for human review
Sweep, screen and organise research with evidence attached
Answer "has this been done?" instantly from the work ledger

What our agents never do

Submit anything to a regulator
Sign, or commit V3TR4 or a client to, anything contractual
Move or commit money
Send external communications above routine tier without human approval
Act outside their scoped access, or hold credentials for critical actions

These are not policies we ask agents to follow — they are structural. Credentials for critical actions are held by named people, so an agent cannot perform them even in error.

The permission tiers

Tier	What happens	Example
Routine	Agent acts automatically; every action logged	Filing a document, updating a record
Sensitive	Agent prepares; a person reviews and approves	Client-facing correspondence
Critical	Prepared for a named person, who executes it themselves	Anything regulated, contractual or financial

Logging and audit

Every action — agent or human — lands in the work ledger with who, what, when and links to the evidence. The audit trail is append-only: it cannot be quietly edited afterwards. If you ask "who did what, when?", the answer already exists, and clients can see it.

Your data

Client data is classified by sensitivity tier; agent access is scoped per tier and per engagement
Access is least-privilege: an agent working your correspondence cannot see another client's
We operate under UK GDPR; confidentiality is a hard constraint that convenience never overrides
You can see exactly what our agents may and may not touch in your systems, and revoke access at any time

Human accountability

Every engagement has a named V3TR4 specialist accountable for the outcome. AI failure is never an excuse: if an agent gets something wrong, that's our specialist's responsibility, backed by the audit trail that shows exactly what happened.

We hold ourselves to this because we live on it

V3TR4's own company administration runs on these same agents, tiers and ledger — client zero is us. Every rule on this page governs our own business before it governs any client's.