Heidi's 18% job cut: AI vendor checks for NZ businesses

What Heidi announced

Heidi is the Australian company behind the AI medical scribe many New Zealand clinicians now use. In the first week of November it told staff it would cut about 18% of its global workforce. Startup Daily reported that Heidi has more than 600 employees, so at least 108 roles will go. Two of the three co-founders are leaving at the same time: chief technology officer Yu Liu and chief financial officer Waleed Mussa. According to The Medical Republic, chief executive Dr Tom Kelly is now the only co-founder left in the business.

The cuts came about two months after Heidi raised a lot of money. In September the company closed a US$100 million Series C led by Blackbird at a A$1.26 billion valuation. It also took a separate US$240 million debt facility from General Catalyst's Customer Value Fund to pay for customer acquisition, according to Startup Daily. Kelly told The Medical Republic that Heidi had been too optimistic about how quickly large health systems would buy. He said the company had over-resourced enterprise roles that were waiting on slow hospital deals, and that investors had pushed for a more predictable business. Under the new structure, customer support moves closer to the product team and research and development runs on a single roadmap. The Medical Republic also cites Australian Financial Review reporting that the cuts fall heavily on Heidi's Melbourne R&D, support and sales teams, and that some roles are moving offshore.

Heidi says it still supports three million patient interactions a week and is growing quickly. In the same week, Canva announced a restructure affecting about 22% of its workforce, roughly 1,270 people, Startup Daily noted. Two of Australia's best-known tech companies redrew their org charts within days of each other.

Why the cuts matter in New Zealand

Health NZ is one of Heidi's most visible public sector customers on either side of the Tasman. RNZ reported in November 2025 that Health NZ bought an initial 1,000 licences for frontline emergency department staff after trials at Hawke's Bay and Whanganui hospitals. Heidi says average note-writing time in the trial fell from 17 minutes to 4 minutes. The Hawke's Bay trial reviewed the experience of eight clinicians, so the evidence base for a national rollout was small.

Neither Heidi nor Health NZ has said publicly whether the restructure changes support arrangements, account management or the roadmap for New Zealand customers. We don't see anything in the reporting that suggests the product is at risk. Heidi has fresh capital and a large user base. The more useful point for New Zealand businesses is how a well-funded AI vendor reorganised itself once its sales assumptions met the pace of real procurement.

What the restructure tells AI buyers

Most New Zealand organisations adopting AI are buying from vendors that are only a few years old. Those vendors are often based in Australia or the United States and are under pressure from investors to show predictable revenue. Heidi's case shows which parts of a vendor tend to change first when growth plans get revised. Sales and account management roles hired ahead of slow enterprise deals were among the first cut. Support was folded into the product team. Two roadmaps became one. Every one of those changes lands on customers, even when the software keeps running exactly as before.

A buyer in New Zealand is usually a small account served from another country. When a vendor tightens its focus, smaller markets can end up with less local contact, slower answers to configuration questions, and features built first for the markets the investors care about most. Kelly named competition in the US as Heidi's key pressure point, and The Medical Republic pointed out that Epic launched its own AI scribe there in February 2026. A vendor that's fighting for share in the US will set its priorities around that fight.

The Heidi story also shows something about measurement. Health NZ's rollout rested on time-saving figures from a short trial reported mostly by the vendor. Many private sector AI pilots in New Zealand rely on similar evidence. If you can't show what a tool delivers in your own data, you won't be in a good position to negotiate when the vendor's pricing, support model or product direction shifts.

None of this argues against buying from young AI companies. They often build the best tools for narrow jobs like clinical notes, contract review or call summaries. The risk sits in the contract and in how much of your workflow depends on one supplier you have little sway over.

Checks to run on your current AI vendors

• Find out who supports you and from where. Ask each AI vendor to confirm in writing which team handles your support tickets, in which time zone, and what response times apply. If support has recently moved or merged with another team, ask what changed.
• Read the exit and data clauses. Confirm you can export your prompts, outputs, templates and audit logs in a usable format. Check how long the vendor keeps your data after termination. If the tool handles personal information, check where it's processed so you can meet your Privacy Act 2020 obligations.
• Ask for the roadmap that applies to you. A vendor consolidating its roadmap should be able to tell you which features for your region or industry are still planned and which have been dropped.
• Measure the gains yourself. Run a short internal baseline before and after rollout using your own data, and record the method. Your own numbers will hold up better than a vendor case study at renewal time or in a board report.
• List your single points of failure. Note which business processes would stop if the tool went offline for a week. For anything critical, write down the manual fallback and test it once.
• Watch the vendor's signals. Leadership departures, a restructure or a large debt facility aimed at customer acquisition don't mean trouble on their own. They are good reasons to bring a review meeting forward.

Where we'd start

If your organisation has rolled out more than a couple of AI tools, put them all on one register. Record the owner, the contract end date, the data each tool touches and the fallback for each. That register also gives you a base for AI governance once Australian rules firm up and New Zealand clients start asking suppliers how they manage AI risk. Our AI consultancy work covers vendor assessment and governance. For organisations that want to cut their dependence on offshore providers for sensitive workloads, we also explain how a private AI server can work.

Heidi says it will set out its new structure and priorities in the coming weeks. Health NZ and the private clinics using the tool will be watching that update, and any New Zealand business with an AI supplier in the middle of restructuring should ask for the same detail.