AI-Assisted Compliance

AI is optional.
Off by default. Cited when on.

Some venues love using AI for incident narratives, monthly review summaries, and quick regulatory questions. Others prefer the platform without it. Venue Axis works cleanly either way — every AI feature is off by default, switched on per-feature with consent, and you can turn the lot off anytime. Patron names and IDs never reach the model. (One documented exception covers AI-assisted migration — see below.)

01
Off by default for every new club.
Activation never enables AI features automatically. Each feature is opt-in per profile, recorded as a consent event with a 90-day re-consent prompt on Help.
02
Turn-off-able at any time, two ways today.
Three independent kill-switches. Master AI switch under Settings → AI turns every feature off in one click. Set your monthly AI budget to $0 (Settings → Billing) — no model calls fire. Or revoke per-feature consent and the gates close.
03
All patron-bearing AI runs in Australia.
Every AI feature that touches patron data — vision (paper-register OCR, KYC document extraction) and text (incident narrative drafting, board-pack commentary, AML controls interpretation, natural-language search) — runs on an Australian-hosted model (Google Vertex AI in Sydney). Patron data does not leave Australia for any AI feature. Patron names, dates of birth, driver-licence numbers, phone numbers, Medicare and passport numbers are still scrubbed at the gateway before any AI call as defence-in-depth, even though the destination is on-shore. The only AI features routed to US-hosted models are the contextual help assistant and product feedback triage, neither of which carry patron data.
04
Costs capped pre-dispatch.
$20/month default per-club spending cap, configurable up or down. Typical club usage is well under that — the cap is a safety against runaway cost, not a target spend. Per-feature daily call limits and a monthly alarm sit on top.
Bounded workflows only
Model calls happen inside specific surfaces — incident narrative, compliance Q&A, controls interpretation, board pack commentary. Never a free-floating chatbot.
Patron data stays in Australia (every AI feature)
Every AI feature that touches patron data runs on an Australian-hosted model. Patron names, dates of birth, driver-licence numbers, phone numbers, Medicare and passport numbers are still scrubbed before the AI call as defence-in-depth, even though the destination is on-shore. The two vision features (AI-assisted migration, KYC document extraction) process raw patron content because extraction needs it — and that content stays in Australia too.
$
Cost capped pre-dispatch
$20/month default per-club spending cap, configurable up or down. A typical club uses well under that — the cap is a safety against runaway cost, not a target spend. Per-feature daily call limits and a monthly alarm sit on top.
Opt-in by feature, citation-backed
Each AI feature is opted into separately, per staff member. Help re-prompts for consent every 90 days. Regulatory citations must trace back to a real source — the AI never invents a rule number.
The vision exceptions

AI-assisted migration + KYC document extraction.

Two AI features process raw patron content by design: migration (read your existing spreadsheets, paper registers, certificate stacks and turn them into Venue Axis records) and KYC document extraction (read a driver licence or passport and pre-fill the CDD form). Both feed the AI raw images / files because extraction can't work without them. Both run on an Australian-hosted model so patron data does not leave Australia. Here's what we do for these paths.

  • The migration consent dialog spells it out. The first time you use a migration feature, the consent dialog explicitly states that patron names, dates of birth, and licence numbers are not redacted for this feature. You opt in once per profile; you can withdraw consent anytime from Settings.
  • Australian hosting. Migration calls run on an Australian-hosted model. Patron data does not leave Australia for this feature.
  • Your data stays yours.Once import completes, the records live in your club's workspace. Other clubs cannot see your imported rows; your own staff see only what their role permits.
  • Audited and reversible. Every migration is logged. Every batch can be rolled back in one click — the rows you imported get deleted; the batch record keeps its history marked as rolled back so the audit trail stays intact.
The seven features

Each one earns its place.

AI assistance is available only where enabled, with human review before anything is saved or sent. For text features, patron names and IDs are scrubbed before any cross-border call. Vision features (migration, KYC document extraction) run on an Australian-hosted model so patron content stays in Australia. All seven are in production today.

On the floor

Speak the incident

Staff reviews before submission

The RGO speaks for thirty seconds. The system drafts a structured incident narrative in the format regulators expect. Staff reviews, edits, and submits. Never auto-files. The single most-used AI feature on the floor.

FormatVoice + Text
Daily limit50/day
Where it fitsA duty manager finishes a shift with six incidents to log. Each takes 30 seconds of voice instead of 4 minutes of typing.
Need help mid-shift

Plain-English help

No patron identifiers required

"How do I respond to a confirmed FRT match?" "What does enhanced CDD require?" Plain-English answers from your obligation library — every answer comes with citations back to the actual rule.

FormatText
Daily limitUnlimited
Where it fitsA new RGO mid-shift hits an unfamiliar situation and needs guidance without leaving the workflow.
Feedback inbox

Smart feedback triage

No patron identifiers required

Floor and management feedback gets sorted automatically — bug, feature request, compliance question, support issue — and routed to the right queue. Voice option for staff who don't want to type.

FormatVoice + Text
Daily limitEffectively unmetered
Where it fitsEvery venue contributes feedback. The system makes sure nothing important gets buried.
Onboarding

Auto-fill from your existing docs

Staff reviews before save

Upload a PDF, Word document, or image up to 10MB. The system extracts structured fields — staff details, certificate expiry dates, AML controls, board minutes — for onboarding and migration. You review before anything is saved.

FormatText (file upload)
Daily limitPer-document budget cap
Where it fitsMigrating from a paper-based system or another platform. Bulk-extract 100 staff records from PDFs in minutes.
Ad-hoc reporting

Ask your data a question

Read-only — never modifies state

"Show me patrons with three or more $5k checks this month." "Which staff have expired RCG?" Plain-English in, your data out. Read-only — the system never modifies state.

FormatText
Daily limit50/day
Where it fitsGM has a question about patron risk distribution and doesn't want to wait for a developer to write a report.
Board

Board pack commentary

CEO reviews before distribution

Drafts the executive narrative for each section of the monthly board pack — incident trends, AML status, FRT accuracy, training compliance, upcoming deadlines. CEO reviews and approves before distribution.

FormatText
Daily limitPer-pack budget
Where it fitsBoard meets monthly. CEO needs a one-page narrative per section, not a data dump. AI drafts; CEO edits; pack goes out.
AML/CTF

Gap-check your AML program

Compliance Officer reviews before sign-off

Reads your AML controls and tells you where they fall short of AUSTRAC guidance, in plain English. The AI never invents a rule number — every reference traces back to a real source.

FormatText
Daily limit20/day
Where it fitsThe Compliance Officer drafts a new control and wants a sanity-check against AUSTRAC guidance before sending it for sign-off.
What we don't do

Lines we don't cross.

AI in Venue Axis runs inside specific tasks where the output can be checked. Here's what it won't do — and these aren't product decisions we'd revisit later, they're commitments we lock into the architecture.

  • No free-floating chatbot. Generic AI advisors age badly and staff stop trusting them. AI shows up only inside specific workflows where you can verify what it produced.
  • No invented rule numbers.Every regulatory citation the AI returns has to be traceable to a real source. If it isn't, the system won't show it.
  • Not a substitute for counsel. The AI gap-check on your AML program is a sanity-check, not legal advice. It flags issues; a person decides what to do about them.
  • No AI looking at faces. Match decisions come from your chosen facial-recognition vendor — we never run face models ourselves, and we reject biometric template payloads at the boundary.
  • Nothing goes to AUSTRAC without a person signing.Suspicious-activity narratives can be AI-drafted; the two-person sign-off ensures a real human signs every submission.
AI data handling

Where AI calls go.

AI processing is routed by feature. Every patron-bearing AI feature — both vision (migration, KYC document extraction) and text (incident narratives, board-pack commentary, AML controls interpreter, natural-language search) — runs on an Australian-hosted model, so patron data does not leave Australia. Patron identifiers are still scrubbed at the gateway for text features as defence-in-depth, even though the destination is on-shore. Help and feedback features carry no patron identifiers at all and remain on a US-hosted model.

Every AI feature is off by default; staff opt in per profile with consent recorded as an event. Daily call limits and the $20/month default budget cap apply per club. AI never auto-files records — a person signs every regulator-facing submission.

See the AI features in context.