AI Disclosure

1. What the Vecta AI agent is

The Vecta AI agent is our proprietary intelligence layer. It combines structured queries over your aggregated Financial Data (balances, transactions, holdings across your linked institutions) with Amazon Bedrock (Amazon Nova Lite) running in O&C Technology’s own AWS account in Asia Pacific (Sydney), which generates natural-language answers.

When you ask a question, our orchestration layer determines what data is needed, retrieves it from your synchronised records, performs calculations in conventional software where precision matters, and uses the language model to explain the result in plain language.

2. What the agent can and cannot do

• It can answer questions about your own connected data: spending, cash flow, balances, portfolio allocation, unusual activity and similar topics.
• It writes daily, weekly and monthly insights from patterns in your data.
• It cannot move money, place trades, initiate payments or change anything at your institutions. The Service is read-only by design.
• It does not provide personal financial product advice, tax advice, credit assistance or recommendations to buy, sell or hold any financial product.
• It only sees data from institutions you have connected. If an account is not linked, or a sync is delayed, answers will not reflect it.

3. Tokens: how usage is metered

Agent usage is measured in Tokens. Each plan includes a monthly allowance: Free includes 20 Tokens (about 20 typical messages), Basic includes 50 Tokens, and Pro includes 100 Tokens. Allowances reset each billing cycle and unused allowance does not roll over.

The Token cost of a message is not fixed. It depends on:

• The type of question. Simple banking queries (such as a balance or a category total) are the lightest. Investment and portfolio analysis is heavier because it involves holdings, pricing and performance calculations. Superannuation and multi-domain questions that combine banking, investments and super are heavier again.
• The number of institutions involved. A question that spans many linked institutions requires more data retrieval and computation than the same question over one account, so it consumes more Tokens.
• The depth of analysis. Multi-month trends, anomaly scans and portfolio concentration analysis consume more than single-figure lookups.

Where practicable, the app shows indicative Token costs before you run heavier analyses, and your remaining balance is always visible in your account area.

4. Extra Tokens

Subscribers on the Basic, Pro and Premium plans can purchase Extra Tokens at any time from the account area of the app. Extra Tokens are not offered on the Free plan and are intentionally not sold on the public pricing page; they are an in-app top-up for subscribers, similar to usage top-ups on AI platforms.

• There is no limit on how many Extra Tokens you can purchase.
• Extra Tokens are consumed only after your monthly plan allowance is exhausted.
• Extra Tokens do not expire while your account remains open.
• Extra Tokens have no cash value and are not transferable. Refund treatment is set out in the Billing and Refunds Policy.

5. Accuracy and limitations

We engineer the agent to ground every answer in your actual data, to compute material figures in conventional code rather than by language model, and to state uncertainty rather than guess. Despite this, you must understand the following limitations:

• AI output can be wrong. Language models can occasionally misinterpret a question, mislabel a merchant or category, or summarise data incorrectly.
• Data can be stale or incomplete. Institution and data partner feeds may be delayed, and some institutions provide partial data.
• Estimates are estimates. Projections, trends and categorisations are indicative, not statements of fact.
• The agent has no knowledge of your finances beyond connected data and what you tell it in a conversation.

Always verify material figures with your institution before acting on them. Do not use agent output as the basis for tax filings, credit applications, legal proceedings or other consequential submissions without independent verification.

6. Data handling in AI processing

• Only the data reasonably necessary to answer your question is provided to the model.
• Inference is Amazon Bedrock (Amazon Nova Lite) in our AWS account in Sydney (ap-southeast-2). CDR-derived prompts are not sent to Anthropic or any other external AI API.
• AWS Bedrock terms restrict use of customer content to providing the service and do not permit training on that content.
• Your data is never sold.
• Agent conversations are logged for security, quality assurance and Token accounting, and are handled under our Privacy Policy.
• The application is hosted on Vercel (United States). CDR storage and AI inference are in Australia. See the Privacy Policy, overseas disclosure.

7. Prohibited uses of the agent

• Attempting to make the agent act outside its read-only design or produce advice it is not permitted to give.
• Attempting to extract other users' data, system prompts or proprietary orchestration through prompt injection or similar techniques.
• Using the agent to generate unlawful, deceptive or harmful content.
• Automated bulk querying designed to circumvent Token metering or abuse fair use.

Breaches are handled under the Acceptable Use Policy and may result in suspension or termination.

8. Human review and feedback

We review aggregate quality metrics and investigate reported errors. If an answer looks wrong, use the in-app feedback control or email contact@vectapp.io. Reports of incorrect figures are prioritised, and confirmed defects in Token-consuming responses are eligible for Token credits as described in the Billing and Refunds Policy.

9. Changes

As models, costs and capabilities evolve, we may update the agent, its indicative Token costs and this disclosure. Material changes will be notified by email or in-app notification, and the current version is always available in the Legal hub.