AI Pulse · ANZ AI Watch

A Salesforce survey of 2,132 ANZ knowledge workers has surfaced a sharp contrast in how businesses are handing AI agents access to company information. In New Zealand, 47% of respondents said AI agents have full access to sensitive data. That compares with 32% in Australia and 28% globally. At the same time, 38% named lack of accountability for AI mistakes as their biggest concern.

That combination should make any owner pause. An AI agent with access to customer records, financial information, staff files, or internal documents can save time. It can also make a damaging mistake quickly if its permissions are too broad or nobody owns the outcome. The issue is not whether an agent makes an error. People do too. The issue is whether your business can identify what happened, stop it, correct it, and decide who was responsible for allowing it.

Regulation is also slowing deployment. It was cited as a constraint by 83% of Australian service professionals and 74% in New Zealand, according to Salesforce’s survey findings. Before giving an agent broad access, set clear permission levels, assign a business owner for each use case, and keep a record of what systems and data it can reach. This is the kind of thing we build into an AI command centre, so AI activity is visible and someone is accountable when decisions affect the business.