Published on the 05/08/2025 | Written by Heather Wright

Changing the capital allocation equation…

John-Daniel Trask has a blunt message to business – and one that many won’t want to hear: Get used to spending money.

The founder and CEO of application monitoring company Raygun and new addition Autohive, is talking AI. Autohive, afterall is an AI platform to help businesses create AI agents.

“It is an everybody technology and everybody should benefit from it.”

He’s open that LLMs aren’t cheap, but says it is ‘really important’ for businesses to get used to spending the money and to understand that token usage approximates to outputs and productivity.

It’s a stance Trask has already taken in his businesses, where use of LLMs is tracked per team member and the narrative of the company is to spend more. It is, he admits, something of a head fake for employees more used to being told not to waste company resources.

“Get more done. Scale your company incrementally, by the token, not the headcount, and achieve more,” he told iStart.

As an example, Trask cites one of Raygun’s lead engineers who now works with seven coding agents simultaneously. Every day he ships three to five bug fixes and one to two new features.

“That would have been a few people for a week a year ago.”

Trask believes AI is going to be ‘a bigger gift than the internet’ and he’s keen to see New Zealand embrace the technology to improve productivity.

“The ability to get more done with the same resources is always a huge benefit of any technological change.”

While many are still wary about the big claims being made about AI and its potential productivity gains, Trask says his own experience shows how the technology can provide big benefits, but he’s worried that the number eight wire mentality of old is a thing of the past, and Kiwis are becoming too timid, waiting for other companies to make the first moves and show that success can be achieved.

Back in late 2022 when ChatGPT dropped, Trask and the Raygun team saw it as ‘another internet moment’. Unlike when the internet arrived, Trask now has the resources behind him to pursue the AI dream, and told the team in early 2023 he saw a big future for AI. Raygun would not backfill leavers, he announced, and he wanted to ensure the people it held on to were the most AI-native.

By May the company was having its first AI week, stopping all work, with the exception of customer service, to focus on the technology.

“We had the entire company do our engineering onboarding test that we would use for hiring and everybody could do it, including the finance people, designers, people who had never written code in their life.

“This was the realisation they could achieve more now than they ever could before.”

By 2024 though, Trask says he was sitting at the edge of AI, wondering what to do next.

“That’s when we reached the conclusion we were going to have to build what is next.”

Next, for Trask, is agents. The company began building agents to run its Adwords account, handle marketing and audit inbound trials.

If ChatGPT is the email of the internet days, he says agentic is the web stage. Few would say that email was the best thing about the internet. Most would point to ecommerce, the web, the digital rails that business now runs on as being the true value. So too, Trask believes, it will be with agents.

But for now at least, there’s a lack of on-ramps, particularly for non-tech businesses.

“The reality is those software products don’t really exist yet. Well, they kinda do – there are some legacy workflow tools that have tried to adopt it.”

Autohive, which launched earlier this year, was Trask’s response to the issue.

“It’s us trying to say what would the dashboard look like to run a modern AI native company where agents are executing various parts of the business for us, and the team manage those agents through a really nice interface that feels natural.”

The company has an expanding library of extensible integrations and uses MCP, a standard being adopted for enabling tool sharing between agents.

Current integrations include Hubspot, Adwords, Google Calendar, Gmail, Github and Discord.

Agents can also write a call to an API for other systems, as yet unsupported by Autohive.

It’s also collaborative, providing a messaging platform akin to Slack and Teams, to enable colleagues to discuss issues – and @mention agents who can join the conversation and provide input just as another colleague would.

“Autohive is the response of saying this is a superpower technology, but as with a lot of new inventions, it’s currently pretty nerdy. How do we make it accessible to everybody – because it is an everybody technology and everybody should benefit from it.”

Ironically, Trask says when it comes to AI and agents, the revolution won’t be lead by engineering or tech teams.

“I will tell you right now there are just as many people on the scepticism bus on the technology side as there are elsewhere. When I go around meeting companies, it’s not a role that gets AI.”

Instead, it’s often those from other areas of the business, such as HR and finance, who are playing with agents.

“It’s not based on a role. Make them the champion in your organisation and really hail them.”

The perils of overhype

But while Trask admits to being ‘extremely optimistic’ about AI and agentic, he also says we need to be real about current limitations.

“I’m excited for what the future can bring. But let’s be real.”

He says he often hears companies outline grand plans for agents, such as agents that act as an expert EA for every person in the company.

“You’d want better capabilities than the current frontier even has. This doesn’t exist yet guys.”

Instead, he’s urging businesses to start small.

“Think of a task that takes you an hour or more every week or month that you just bloody hate and automate that piece – just build an agent that does that one little bit.”

Then build on it. And keep building on it.

He notes a personal example – his use of an agent to manage his Gmail. It started as a simple tool to summarise the day’s emails into a report. Then the ability to auto-archive some was added. Then forwarding those with invoices to the accounts team automatically.

To be fair, some of that can be done using rules, he admits. But more labouriously.

Trask offers up three tips for businesses when it comes to AI and agentic:

1. Focus on the culture that will enable experimentation and celebrating of the people who are actually driving innovation in your business. “They are the ones who are actually working to ensure your business still exists in the future. They are important.”
2. You are going to have to open the chequebook. “This stuff isn’t cheap. One way you might think about it is if you plan to hire 10 people, maybe you hire nine and that $15k month that was going to be a salary is going to go into tokens to enable that other nine to perform like they are 18. That’s how you want to be thinking about capital allocation in the C-suite – that tokens are the super power behind people.”
3. Encourage boldness and internally focus it. “Everybody has a great excuse for sitting on the side and waiting. But that’s typically not the attitude of a winner. I want to see more winning companies in New Zealand. But secondly think about the opportunities to reduce toil for your existing team. that’s the internal part.” While some companies are keen to build external facing agents he says that’s an area where you could be setting yourself up for too much risk. “You lower the risk by doing something internally rather than making it externally focused at the very beginning.”