Anna Mowbray has built start-up Zeil with AI tech and support from Google. Photo / Carson Bluck
Anna Mowbray staged something of an AI beauty contest as she prepared to launch her latest project - Zeil, a Gen Z-friendly app pitched as the “Tinder for recruitment”.
The entrepreneur assessed AI platforms and technologies offered by Amazon, Microsoft-backed OpenAI, Facebook and Google as she seized the opportunity that a start-up provides to incorporate artificial intelligence from the ground up (more on which shortly).
While Google platforms are the base, all software development is in-house via 28-all local staff and counting.
“My goal behind this project is to build another unicorn from New Zealand, using all-New Zealand talent to power it up,” Mowbray says.
“Unicorn” is start-up-speak for a private firm that hits a $1 billion valuation. Often such ambitions have to be taken with a pinch of salt, but Mowbray created Zuru Toys with her brothers Mat and Nick; the NBR put the siblings’ collective worth at $3.2 billion on its 2023 rich list. Zeil is bootstrapped, 100 per cent owned and funded by Anna Mowbray.
Within a couple of years, most of the workforce will soon be the so-called digital natives: Millennials or, snapping on their heels, Gen Z.
“We need to talk to that demographic,” Mowbray says. Applicants can swipe left on a picture-heavy job list if it’s not for them, or swipe right to apply or shortlist.
She also wants to help the emerging workforce talk to employers. AI-assisted cover letters and CV-builders will be added to Zeil later this year. The aim is to help applicants communicate well, and include all the necessary elements, but still maintain their own voice.
As part of the process, Zeil has consulted with the Blue Light programme in South Auckland and Le Va, a Pasifika-led non-profit that offers youth and community support.
“One of my most important goals through this platform was to look at how you could democratise the path to employment for our youth and those who are underprivileged.”
In part, that will be through using AI to help candidates find their voice and present themselves well when creating a CV and cover letter.
“That’s a really challenging aspect for a lot of people,” Mowbray said. “I talk about the privileged upbringing that I had and one of those privileges was that I had parents that were able to support me to build and to put together my personal brand and profile. We don’t all have that luxury.”
“We’ve found that around 15 per cent of profiles on the platform don’t necessarily follow best-in-class practices,” Mowbray adds. The Zeil crew is experimenting with various conversational prompts to help users write a good application, and help them articulate the skills they have for any given role - without it becoming cookie cutter.
“I want to be very careful and considerate about how we approach this, because I think it’s important a candidate’s profile is authentic.”
More bluntly, she’s also undercutting Seek on price.
Mowbray is targeting 250,000 downloads and 25 per cent market share in 18 months.
Since its August 2023 launch, Zeil has had 60,000 downloads, she says.
So how are things going, six months in?
And as of earlier this week, there were around 2500 jobs on the platform. Early supporters range from Fonterra, Downer, Westpac, Fisher & Paykel Appliances, 2degrees, Les Mills to hospitality outfits like Savor and Mitre 10 to hot start-ups like Tend and Partly. And yes, Zuru uses it too. Freebie campaigns are offered as a taster). Mowbray puts that at around 10 per cent market share, allowing for crossover with rival platforms.
On March 15, Seek said it had 20,990 New Zealand listings, Trade Me 15,579 and LinkedIn 10,627.
“We did a lot of research. We tested five different models, everything from Facebook’s Llama to Amazon’s Titan, OpenAI and obviously Google, which at the time was Bison. We built out use cases in the HR [human resource] space - CV prompts, job description prompts, candidate matching -and we really tested every platform to see which had the best output for us. I was very happy that we settled on Google because they have got an incredible team down here in New Zealand; some of the best engineers and technical people.”
Mowbray said: “Our first foray into AI was utilising it for job descriptions. A general job description will take an HR manager one to two hours from scratch. We wanted to look at how we could do that 10 times faster and reduce costs.
“You prompt the model with some basic insights around job title, work type or style, and seniority of the role. And then that gets coupled with intel from our platform around your organsation’s background, their perks and benefits.”
The AI takes those elements, then presents them in your organisation’s tone-of-voice, which it picks up from your brand page, and other sources.
Off the back of the pilot, Zeil started building support for salary recommendations.
“One of the big things for us is to encourage that transparency. We know that candidates are 70 per cent more likely to apply for a job if they see the salary on it. That’s what our data is telling us,” Mowbray said.
Beyond customer-facing AI, creating a new business from scratch has been an opportunity to incorporate AI from the ground up.
“We’re really driven to ensure we’re exploring AI technology in every function of our business,” Mowbray said.
“So we’ve also set up an AI marketing assistant using Google’s Gemini model, but also the RAG framework - so looking at retrieval, augmented generation and how we can use that to build creative assets, build creative copy for us and layer over our own tone-of-voice and have that fully on-demand.”
RAG, pronounced like the bit of cloth, stands for retrieval augmented generation. It’s a framework designed to make large language models like ChatGPT Gemini - which can sometimes “hallucinate” incorrect facts or go rogue with an inappropriate image - more reliable by pulling in relevant, up-to-date data directly related to a user’s query, often from an inhouse database.
RAG is supported by Gemini, plus another Google tool that’s in Zeil’s mix, Vertex AI, which is used for rapidly prototyping and testing generative AI models. It helps train, deploy and monitor an AI app. It can help an organisation put its own parameters around how the AI will respond to user queries.
“We’ve also built - again using Google and Gemini - an AI-powered data insights tool. We’ve reconstructed our data warehouse to ensure that we can use conversational search over the top of that. So we can ask the model, for example, ‘What does user acquisition look like for the last four weeks? And what were the key inputs that drove that success?”
“We’ve been working very, very closely with Caro [Google NZ country manager Caroline Rainsford] and her team, which is incredible because to partner with an organisation that’s so collaborative and really wants to bring novelty and an edge to a category - which is really important for us, because if we want to stand out above and beyond the competition, we’ve got to be constantly thinking differently,” Mowbray said.
“We’ve also worked with Google on understanding the competitive landscape and figuring out where we can go next.”
“We’ll go to Australia, relatively quickly. And then we’ll be looking further afield from there.”
Chris Keall is an Auckland-based member of the Herald’s business team. He joined the Herald in 2018 and is the technology editor and a senior business writer.
Auditors PwC: Cited disruptions to operations from that IT upgrade.