The hazards, marked before you sail into them.

The Unsolicited Electronic Messages Act 2007 (UEMA) governs every email, and mobile text message your automations send. It does not cover voice calls, which we treat separately below.

Consent basis. Consent can be express, inferred, or deemed. Deemed consent exists where an electronic address is conspicuously published in a business or official capacity, there’s no statement saying the holder doesn’t want unsolicited messages, and the message genuinely relates to that person’s business role. This is the legal basis for compliant B2B outreach in New Zealand — but if a business’s own site says “no unsolicited emails,” that address is off-limits, full stop.

What every message must carry. Accurate identification of who authorised the message, accurate contact information, a New Zealand postal address, and a functional unsubscribe facility. We keep records of every consent basis — express, inferred, or deemed — for every list we build.

What we won’t do. Address-harvesting software to build contact lists is prohibited under UEMA, and it’s exactly the kind of tooling a lot of offshore automation agencies sell as standard. We build lists from manual research, public directories, and association member lists instead.

Why it matters to your build. Any reactivation or win-back automation we build includes consent-state tracking as standard, and onboarding always confirms the consent basis for your existing customer list before we touch it. Penalties for non-compliance reach up to $500,000, and the burden of proof sits with the sender — so we build the paper trail in from day one.

UEMA explicitly doesn’t cover voice telemarketing calls, which means a voice agent isn’t legally bound by the same consent and unsubscribe requirements as email or text. That doesn’t make it a free-for-all — we apply these as non-negotiable house rules on every voice deployment:

• Always disclose it’s an AI at the start of the call. This matches the Privacy Commissioner’s transparency expectations, and it’s simple commercial sense — being caught pretending an AI is human damages your reputation and ours.
• Always offer a path to a human, on request, at any point in the call.
• Never let a voice agent make a decision that materially affects someone — approving credit, making a clinical judgement, agreeing to a contract variation.
• We check the Marketing Association’s name-removal and do-not-call service and honour suppression lists.
• We record and review calls, with disclosure and consent, for the first month of every deployment.

Voice AI is the most impressive part of any demo and the highest-risk part of any deployment — a bad voice interaction fails loudly and publicly. We deploy it inbound-first, tested extensively against New Zealand accents, and only add outbound calling once inbound is proven and trusted.