The future of AI is ‘agentic’: Arcanum AI founder Asa Cox on the incredible potential of virtual assistants

Imagine having a gifted EA who could handle any low-priority email or routine admin task in your business and worked 24 hours a day, 365 days a year. 

According to Asa Cox from Arcanum AI, that’s the future of AI virtual assistants, which will soon arrive.

“Think about how long you spend writing emails, how long you spend writing internal reports, how long you spend checking external websites for information, how long you spend analysing data,” he told Caffeine.  

“Depending on your role, I think you can kind of see half of your day, you know, being freed up. That’s not unrealistic.” 

According to Cox, who specialises in creating AI agents to help automate work for Kiwi businesses, this is already happening with his clients. 

“We’re working with customers who are freeing up literally 50% of a full time employee’s time with some relatively simple AI implementation. We’re seeing that it’s true. Those kinds of stories are going to come thick and fast over the next 6 to 12 months.” 

The explosive growth of Arcanum AI supports Cox’s vision. He says revenue has gone from $100,000 to around $2 million in the past 12 months, and he expects that to double in the next year.

As Caffeine covered in our conversation with Madeline Newman, NZ AI Forum Chair, Arcanum is far from alone in predicting exponential growth in the AI economy.

A report by consultancy firm PWC argued that AI could contribute up to $15 trillion USD to the global economy in 2030, more than the current output of China and India combined.

Another report from Microsoft, New Zealand’s Generative AI Opportunity, says generative AI adoption could add $76 billion to New Zealand’s economy by 2038, more than 15 percent of GDP. It could also increase worker productivity by 15% in the same time frame.

Built into those rosy economic predictions is how AI systems will progress from pure tools which users interact with, to active participants in our economy.   

Microsoft AI CEO Mustafa Suleyman told Wired recently that AI tools are progressing through three distinct stages. 

First, learning where they can absorb the same information a human can. Second, the stage they're currently in, is developing a long-term, persistent memory that creates a shared understanding over time. 

And third, he argues, is: “AI interacting with third parties by sending instructions and taking actions—to buy things, book things, plan a schedule.”

Of course, even enthusiasts like Cox can see where the edges of responsible use are, and he is not advocating AI being delegated anything with serious consequences in your business just yet.  

“I think we could look at the consequences of failure. So if you put it into the financial system and you go, ‘Great, there’s an API to ANZ bank, we should, therefore, let our AI do all of the payments -  maybe not. The consequences are pretty high.” 

“It’s about looking at, What’s the tolerance for failure? What are the consequences for failure? That’s a rational process to go through to go. There’s a huge amount of value here but also a huge amount of risk.”

But Cox argues that tools are only becoming more capable and reliable at a dizzying speed, and there is no reason to think that the pace will slow down soon. 

Tech titans have not yet run out of data to train frontier AI models on and have also been finding ways to get more from less. Research from Epoch found that “every nine months, the introduction of better algorithms contributes the equivalent of a doubling of computation.”

If the effective computational power of the models grows at a literally exponential rate and will continue to do so for the next two years, what will NZ look like when it translates into increasingly active AI agents?

“I think there’s an opportunity to kind of 10x the growth of companies with only doubling the cost. I think the potential to grow profitably is what we’re all kind of needing, right?,” says Cox.  

“We need to grow the GDP of the country. And we can’t keep on employing humans to do that because we’re gonna run out pretty quickly. So I think there’s a huge opportunity, even at this stage of where the technology is now, let alone where it will be in 18 months, 24 months.” 

Fundamentally, says Cox, the future he wants to see is not about taking humans out of the loop or replacing them across our economy, but freeing them from the kind of tasks we don’t actually want humans to be doing in the first place. 

“We get the ability to grow without having to by default, hire a whole bunch of people to do lower level tasks. And so I think that’s good for businesses.” 

“And it’s good for the humans that are doing those jobs or not doing those jobs anymore, because they can do more meaningful work. Ultimately, it frees us up to spend more time being human, which is what we should all be focused on.”