What does AI in business look like? 

We want to get the real robots to do the robot work humans have to do, explains Asa Cox. 

He’s the CEO and founder of Wellington-based Arcanum, formerly Intela AI, a startup integrating artificial intelligence (AI) into SaaS products to simplify back-office operations for small-to-medium businesses. 

The company offers customers two AI assistant solutions: Archie, aimed at financial services, and Arcos, for customer support services, building the ‘smarts’ into business software. 

It also provides a free canvas and guide to assist businesses with AI adoption. 

Having founded Arcanum in 2016, Cox has long been riding the AI wave, first catching a glimpse of the artificially intelligent future in Silicon Valley ten years ago. 

“When I left, it was obvious this was coming. I thought, ‘Why not be one of the people to start talking about it’?” 

With AI still a fledgling space back then, Cox bootstrapped the startup through its first years, with the occasional boost from early innovation and R&D projects. 

A notable project, commissioned by Rugby New Zealand for the 2019 World Cup, used AI and machine learning to digitally chart players' movements, helping identify the opposition's strengths and weaknesses. Its success led to the AI sporting venture Play in the Grey, a collaboration between Arcanum and Dot Loves Data.

As the interest in AI rose, the startup gained the support of Callaghan Innovation and, since 2019, has successfully carried out three pre-seed investment rounds, attracting early-stage investment from the likes of WNT Ventures, Quidnet Ventures, NZVC, and Ice House Ventures. 

Now the “market is here” for AI, Cox is planning a $2 million seed round in the first half of this year to support the business's expansion – which is gaining traction. 

Following the announcement of a multi-year collaboration with cloud computing company Amazon Web Services (AWS) last November, Arcanum will be introducing three new AI assistants, aimed at resolving staffing and recruitment issues. 

The collaboration will also support Arcanum in its plans to break into the United States and double the size of the business. 

‍

‍

How are you using AI in your business operations? 

We have to adopt the technology that we want others to adopt and do it early. 

In terms of our internal operations, we use AI to deal with inbound sales inquiries. It will qualify somebody when they send an email to see if they’re a good fit and then book a sales meeting. We also use it in our accounting to extract information from invoices and put it into Xero. 

It provides internal support as well. We have a lot of technical documentation, so we make that available to the AI, and it can answer questions for us. 

We’re using and refining everything we’re taking to market.

‍

What AI models are you using? 

We started on pre-trained models 18 months ago. Before these models, you had to train a machine learning model, which was a big barrier to entry and quite expensive. 

Now, because of the GPTs [generative pre-training transformers], or those that have already been pre-trained, the knowledge and capability is built in. 

The first one we used was an AWS pre-trained document processing model. It was the next generation of optical character recognition [the process that converts a text image into a machine-readable text format], which has been around for a long time. 

Then, we started using OpenAI and GPT-3, the precursor to ChatGPT. At the time, everyone thought it was wizardry; they had never seen anything like it. 

You can build a lot on OpenAI, and people do, but there are other options out there with different flavours and capabilities. That can be hard if you don’t know what you’re doing. Thankfully, we do. 

Now, we’ve fully migrated onto AWS because it has all the other open-source generative AI options, of which there are lots. We get much more control over it and are not beholden to a single commercial company. 

‍

How do you foster AI literacy in your organisation? 

Really, everyone’s learning as they go. 

We have a couple of PhDs and Master’s students who come from the machine learning world, but even that knowledge is only useful to a certain degree because it was yesterday’s machine learning when it was all about data. Today is more about software and how to interact with AI models. 

We have software engineers who had very little to no AI experience 12 months ago, and they’ve been figuring it out as they go. 

So, no, you don’t need AI expertise in terms of the technicalities. What’s needed is an understanding of how to apply AI that generates business value. The ability for us, as a company, to help our technical teams understand the business problems they’re trying to solve, that journey of education, is the most important. 

‍

What’s been your biggest challenge with AI? 

The biggest challenge up until this year has been convincing people that AI is worth exploring. Now, it’s how quickly we can build the technology to cope with the coming demand.

We’ve moved from a market education challenge to a technology challenge – but it’s great because we’re in control of the technology, whereas before, we weren’t in control of the market. 

‍

How about successes? 

We've done some amazing work in the past that we've lived off for a few years, for instance, with the 2019 Rugby World Cup.

Most recently, we’ve been chosen as AWS’s only generative AI partner in the [Asia Pacific, excluding Japan]. 

It’s taken a long time to build that relationship, but now it’s beginning to pay off, and that will put us in a strong position in terms of being the partner that’s doing all the ‘super sexy stuff’ with AWS. We’re very proud of that.

‍‍

How do you manage privacy while using AI? 

From a technology perspective, privacy is pretty much a solved problem for most people. 

If you are using the free version of ChatGPT, then we all know it’s free because we are the product. But if you get the paid version, simply tick the button that says, ‘Don’t share my stuff with OpenAI’. If you pay for it, it will be private. 

That’s the same broadly speaking across AI now, but people should always check that it is set up to be private before using it. 

‍

‍

Does the New Zealand government’s legislation help or hinder your efforts? 

Unfortunately for the New Zealand government, the knee-jerk reaction is to stop usage. 

About six months ago, they advised not to use AI in public-facing applications. They were working off the assumption that the privacy/security thing would be a system problem, which it isn’t. From that, there are still pretty big agencies that aren’t allowed to use ChatGPT and the like, so they’re missing out on huge opportunities for productivity. 

In terms of broader regulation, it’s going to be interesting. 

The AI Act was passed in Europe recently and the European ambassadors were in New Zealand a few months ago. I would imagine that New Zealand isn’t going to reinvent the wheel in AI regulation. It doesn’t have the skills, people or scale to invest in it. 

Our regulations will likely have a flavour of the European Union (EU) Act, which has good intentions but has some holes and overreach. A few aspects will need to be considered for New Zealand, particularly around Māori data sovereignty. 

Another point is that the EU is approaching AI from a ‘protect its citizens’ angle and is less focused on economic value. As New Zealand’s software industry is a growing export, regulations may be a little looser. 

But, so long as we have something similar to the EU, then we will be closer to a global standard, which will be useful for companies like us trying to export and also for companies to get the confidence to adopt AI in their businesses. 

‍

What benefits can small businesses get from adopting AI? 

We all know we have a productivity challenge here in New Zealand. We also know that it’s not easy to hire and retain staff at the lower levels, particularly in areas of repetitive manual effort, which significantly impedes businesses. That’s where I see the opportunity. 

Whether you’re a small company of two or three people or a relatively large organisation, it’s not necessarily a big, sexy transformation of a company but automating the slower areas or places where you can’t hire humans to do the work. 

‍

AI in Action is a fortnightly series spotlighting New Zealand startups using artificial intelligence (AI) in day-to-day business. Contact mary@caffeinedaily.co if you are an AI company interested in participating.