Our AI identifies itself as AI
Ask it if it’s a robot and it will tell you it is. You never have to work it out for yourself.
Every LeadGenie voice and chat agent is configured to disclose that it is an artificial intelligence whenever a person asks — directly or indirectly — whether they are speaking with a human, a machine, a recording or a bot. Disclosure is immediate, unambiguous and given in everyday words.
Rather than describe the disclosure, we publish it. This is the actual wording, not a paraphrase:
The agent does not stall, deflect, answer a different question, or say something technically true but misleading such as “I’m here to help you today”. It answers the question that was asked, the first time it is asked.
In chat and SMS, disclosure does not depend on the person asking. Every conversation carries a persistent AI assistant label for its full duration — not a line that scrolls away after the first message — with the option to reach a human beside it.
This behaviour is enforced at the platform level. It is not a setting a customer can disable, and we will not build a deployment that removes it. It reflects Guardrail 7 of the Voluntary AI Safety Standard: that people are informed when AI is used and when it is generating content that affects them.
It never pretends to be a specific person
The AI has its own name. It won’t be dressed up as your receptionist Sarah, and it isn’t a copy of anyone’s voice.
Each agent is given its own identity and is introduced as answering on behalf of the business. We do not configure agents to adopt the name or persona of an identifiable real person, whether an owner, an employee or anyone else.
The synthetic voices we use are generated. They are not cloned from, trained on, or modelled to imitate the voice of any identifiable individual, and we do not offer voice cloning of your staff as a feature. Requests to impersonate a named individual are declined.
Customers may choose the agent’s name, accent and manner. They may not present it as a member of their team.
A human is always available
Anyone can ask for a person and get one. They only have to ask once, and they don’t have to give a reason.
Every conversation includes a route to a human. On voice, a request for a person triggers an immediate message to the business with the caller’s name, number and reason for calling, flagged for callback. On chat, the conversation is handed to the business’s inbox.
The agent never argues, never says “I can help you with that instead”, and never requires the caller to explain why. Asking once is enough.
We do not deploy retention scripts, objection handling, or any loop that returns a caller to the AI after they have asked for a person. This reflects Guardrail 5 of the Voluntary AI Safety Standard — meaningful human oversight, and the ability of an affected person to reach a human.
What we cannot promise on a customer’s behalf is how quickly they ring back. That is a business commitment, and it is theirs to make.
What it is allowed to say about price
It only quotes from the pricing the business gave us. Anything it says is an estimate, and it says so out loud.
Agents draw pricing exclusively from the rates, inclusions and exclusions supplied and approved by the customer during onboarding. Where a price is given, it is framed as an estimate, stated as subject to confirmation, and stated as excluding GST where that is how the customer’s pricing is structured.
Where a scenario falls outside the supplied pricing, the agent does not extrapolate. It states that it will not guess, captures the detail, and refers the enquiry to the business.
Agents do not offer discounts, waive fees, agree to price matches or vary a customer’s terms. Any figure given is an estimate only and is not an offer capable of acceptance. Nothing in this clause limits a business’s obligations to its own customers under the Australian Consumer Law.
Recording, transcripts and consent
If a call is recorded, the caller is told at the start, before any details are taken. If they’d rather not be recorded, they can still be helped.
Where call recording is enabled on an account, the agent gives a recording notice at the beginning of the call, before personal information is collected, and gives the caller an opportunity to decline. A caller who declines is still assisted and their details still taken.
We do not recommend, and will not configure, silent recording.
Recording of telephone conversations in Australia is regulated by the Telecommunications (Interception and Access) Act 1979 (Cth) and by state and territory surveillance devices legislation, which differ between jurisdictions. Because callers may be anywhere in Australia, our default configuration gives notice on every recorded call regardless of the caller’s state.
Transcripts are generated for every conversation, recorded or not, and are visible to the business in its inbox. Retention periods for recordings and transcripts are set out on the
Data, Security & Privacy page.
Whose obligation this is
The recording is of the customer’s callers and is held for the customer’s business purposes. We provide the notice, the mechanism and the default settings. The customer remains responsible for the lawfulness of recording in their circumstances, and for having a privacy policy that reflects it. See clause 09.
What we don't do with your conversations
We don’t sell your calls, we don’t use them to train public AI models, and we don’t mine them for anyone else’s benefit.
Call recordings, transcripts, chat logs and the personal information within them belong to the customer’s business and are processed on that customer’s instructions. Specifically, we do not:
- sell, rent or licence conversations or the personal information within them to any third party;
- contribute conversations to the training of publicly available or third-party foundation models;
- use a customer’s callers’ details to market to them on our own behalf;
- share conversation content with other LeadGenie customers, including competitors in the same trade or area.
We do use aggregated, de-identified operational metrics — how often calls are answered, average duration, how frequently handovers are requested — to monitor and improve service quality. This data does not identify any business or individual and cannot be reconstructed to.
Where we review a specific conversation to fix a fault or improve an agent’s performance, that is done on the customer’s account, for their benefit, by our team, under the access controls described on the
security page.
Where it goes wrong, and who is accountable
It will occasionally get something wrong. We are stating the failure modes up front rather than after they happen to you.
An AI receptionist is not a solved problem and we will not present it as one. The realistic failure modes are:
- Mishearing. Suburb names, surnames, street names and strong accents are where speech recognition most often slips. Contact details are read back and confirmed before a call ends specifically to catch this.
- Missing the point. An unusual, layered or emotionally loaded question can be answered narrowly and unhelpfully.
- Overconfidence. Generative models can state something fluently and be wrong. The constraints in clauses 02 and 04 exist to narrow this to things that do not cost a caller money.
- Poor timing. It can interrupt, or leave a pause that feels long, in a way a person would not.
Who is accountable
We are accountable for the behaviour of the system: the disclosure, the constraints, the escalation paths, and the quality of what we build. The customer remains accountable for their business’s dealings with their own customers, including anything said on their behalf. Where our agent gets something wrong and it costs a customer, we would rather fix the configuration than argue about it. Rights under the
Terms of Service are not limited by this page.
Every conversation is transcribed and available to the customer, so a disputed call is a matter of reading the transcript rather than a matter of opinion.
Automated decisions about people
Our AI takes details and books appointments. It does not decide anything about a person that affects their rights — no approvals, no refusals, no scoring.
Our agents perform capture, qualification and scheduling. They record what a caller tells them, categorise the enquiry, and offer available appointment times. They do not make, and are not configured to make, decisions that could reasonably be expected to significantly affect an individual’s rights or interests — including decisions to grant or refuse a service, set an individual’s price on the basis of personal characteristics, assess creditworthiness, or rank one customer above another.
From
10 December 2026, the
Privacy Act 1988 (Cth) requires APP entities to disclose in their privacy policy where personal information is used in substantially automated decisions that significantly affect individuals. We are stating our position in advance:
we do not use personal information for such decisions in the LeadGenie product. Were that ever to change, it would be disclosed here and in the
Privacy Policy before the capability was released, not after.
Where a customer configures their own automations on top of the platform in a way that does make such decisions, that is their automated decision-making and the disclosure obligation is theirs.
What we require of our customers
Don’t ask us to hide the AI, don’t feed it false pricing, and make sure your own privacy policy says calls may be answered by AI and recorded.
Responsible AI does not work if it is undermined at the customer’s end. As a condition of using the product, a customer must not:
- ask us to remove, weaken or work around the disclosure behaviour in clause 01;
- configure an agent to impersonate a named individual, contrary to clause 02;
- supply pricing, availability or service information they know to be false or misleading, which would risk breaching the Australian Consumer Law;
- use the agents to contact people who have not consented, or who have opted out, contrary to the Spam Act 2003 (Cth) and the Do Not Call Register Act 2006 (Cth);
- deploy the agents in a context requiring professional licensing or regulated advice without appropriate human oversight.
We also require that a customer’s own privacy policy discloses that enquiries may be answered by an AI assistant and that calls may be recorded. We will explain plainly what that needs to cover, though we are not lawyers and cannot draft it.
Lead Genie Au reserves the right to suspend an agent operated in breach of this clause. In practice that begins with a phone call, not a shutdown, unless the breach is serious or unlawful.
How to complain about our AI
If our AI did something it shouldn’t have — to you, or to your customer — tell us. We’ll read the transcript and come back to you in writing.
This applies whether you are a LeadGenie customer or a member of the public who spoke to one of our agents. You do not need to be our customer to complain about our AI.
- Email [email protected] with “AI conduct” in the subject line, or call 0435 095 231. Tell us roughly when the conversation happened and which business it was with.
- We acknowledge within 2 business days and retrieve the transcript and, where it exists, the recording.
- We give a written answer within 30 days: what happened, whether the agent behaved as it should have, and what we are changing if it did not.
- Where the issue concerns how personal information was handled, it is also treated as a privacy complaint under the process in the Privacy Policy, which includes escalation to the Office of the Australian Information Commissioner.
Where a fault is in our configuration rather than a one-off, we fix it across every account it affects, not only the one that complained.
Where AI is and is not used across the product — including the parts that are ordinary software rather than AI — is set out on the
Data, Security & Privacy page alongside retention periods and access controls. Questions about anything on this page can go to
contact us.
