We're Owen and Jessica.

Two different paths

On paper we have nothing in common. In practice we think the same way.

Jessica trained as an engineer and came through the university route. Owen spent five years helping a family friend grow a construction company, from being on the tools to running the operation, doing concrete, waterproofing, fireproofing and epoxy floors.

Good systems start with the outcome. What are we trying to achieve? Where does it usually go wrong? What does the person doing the work know that nobody's written down yet? That way of thinking matters more in AI than people expect, because the model is never the whole system. The real work is turning the way your business runs into something clear enough for AI to follow.

Why we started

It started with a leaking tap. Jessica was working for a startup at the time and got tasked with organising a plumber. She spent the better part of a day calling around. Most went to voicemail. The rest couldn't give a straight answer without coming out to look first.

None of them seemed like bad businesses. Probably the opposite: busy, skilled, and already stretched. But from the customer's side, the front door was a phone nobody could reliably answer.

That stuck with us. Service businesses lose work in quiet ways. A missed call. A slow reply. A lead that lands after hours. A customer who needed one clear answer before they'd book.

ChatGPT didn't even have voice back then, but the gap was obvious, so we started building. Voice AI looked like part of the answer. Then real calls showed us why most of it lets people down.

What the demos hide

We're wary of polished AI demos. They show one clear caller, perfect audio, a simple question, no consequences. The agent sounds effortless because the situation has been made effortless.

Real calls aren't like that. People interrupt. They mumble. They call from bad reception. They change their mind halfway through.

One call taught us this better than any demo could. We'd built an agent for a client based in Auckland, so it answered the way the business would: "Hi, this is Auckland Services, how can I help?" The caller asked the usual things. Services, pricing, when someone could come out. It went smoothly. Then, right as the agent moved to book the job, they mentioned they were in Invercargill.

And it's never just one moment. On a real call, the agent has to handle:

That's where the work is. Not making an agent sound clever in a demo, but making it useful when the call gets messy.

The voice is the easy part

A voice on its own is just a smarter voicemail. It can answer and sound natural, but if it doesn't understand your business, it can't do much that matters.

A useful agent needs two things: the operating knowledge behind the call (your services, prices, service areas, booking rules, common exceptions, and handover points) and the systems behind the conversation (your calendar, CRM, inbox, and job management tools).

No model arrives knowing that Tuesdays are full or that one suburb sits outside your area. That knowledge lives in your business.

This is the part we care about most. The best agents don't come from throwing a generic bot at your phone and hoping the model works it out. By the time you read this there'll be another dozen models and a hundred new tools, with more next month. The bottleneck was never the technology. It's clarity: how a good staff member handles the call, what they ask first, where mistakes usually happen, and what should never be promised.

Once that's clear, the agent earns its keep. It answers the common questions, captures the right details, qualifies the lead, books the job, and sends your team a clean summary. We build the AI around how your business actually runs.

Where AI earns its place

AI earns its place where the work repeats, the rules are clear, the information exists somewhere, and a person can step in when it matters.

The phone is the sharpest example, and it's our specialty. A missed call becomes a missed job; a rushed one becomes a messy booking. Handled properly, the agent becomes a reliable front door: not a replacement for your people, but a layer that answers, filters, books, and hands over cleanly.

The same thinking runs through the quieter work behind the scenes, the lead follow-ups, the quote reminders, the admin nobody should have to type twice. The phone is where most businesses feel the pain first. The real value comes from connecting that call to everything that happens next.

A human stays accountable

We're firm on one rule: a human stays accountable. AI can't own a mistake, and "sorry, the AI did that" isn't a line your customer wants to hear twice. So we build systems you can understand, review and improve, not black boxes.

On custom builds, the work doesn't stop at launch. We review real calls after they happen, with tooling that flags the moments where the agent hesitated, misread someone, overpromised, or left a gap in the workflow. We tighten it from there, week after week. That's how it gets better over time: not by pretending it's perfect on day one, but by learning from real calls without putting your business at risk.