AI automation for busy NZ teams

If your team spends hours every week sorting email, copying numbers out of PDFs and putting together the same report every Monday, that's the work AI automation does well. We build AI automations and AI agents for New Zealand businesses, connect them to the systems you already use, and keep a person in charge of anything that matters.

What AI automation looks like in practice

These are the kinds of jobs we automate:

• sorting incoming email, tagging it, sending it to the right person and drafting a reply for someone to check
• pulling suppliers, dates, amounts and line items out of PDFs and invoices
• classifying documents into the categories your business uses
• drafting quotes from a customer's request and your price list, ready for a person to review and send
• pulling numbers from several systems into one weekly report
• giving staff one place to ask questions across several data sources, such as company data connected to Claude Desktop

AI workflow automation, step by step

1. We go through the process with you and find the steps that take the most time.
2. We propose what to automate, with a price and a time frame.
3. We build the automation and connect it to your systems. If something unexpected comes up, we check with you before going further.
4. Once it's running, we look after it, either as issues come up or on a retainer.

AI agents for business: what to trust them with

An AI agent is an AI model that can take actions as well as answer questions. It might read a mailbox, look something up in your systems or update a record.

Agents are good at reading, sorting and summarising, at pulling data out of messy documents, and at gathering information from several places into a first draft. They're poor at knowing when they're wrong. A model can produce a confident answer that's simply made up, so anything with real consequences needs a person to check it.

We wouldn't let an agent do these on its own:

• pay or move money
• send anything to a customer without a person reading it first
• delete records
• make decisions with legal or financial consequences

The safest agents have a narrow job and only the access that job needs. Our guide to AI agents for business covers where they fail and how to keep one safe.

AI chatbots for business, AI receptionists and AI voice agents

We build website chatbots, AI receptionists and AI voice agents that answer the phone. They earn their place when they answer the same questions every day: opening hours, prices, availability, how to book. They also take messages after hours and book appointments straight into a calendar.

They annoy customers when there's no way to reach a person, when they guess at answers they don't have, and when they're put in front of complaints or anything sensitive. A good one answers from your own information, says so when it isn't sure, and hands over to a person easily.

Business automation that connects to your systems

An automation is only useful if it reaches the systems where the work happens. We've connected systems to Xero before, and we build connections to the other software your business runs on, so that data an AI pulls out of an invoice or an email ends up where your team needs it.

Keeping a person in the loop

AI prepares the work and a person approves it. In practice, that means a person approves anything that leaves the business, changes money or can't be undone, and you can see what the automation did after the fact. Over time, you can decide which steps you trust it to do without checking.

What AI automation costs to run

There are ongoing costs beyond the build. Most AI providers charge by usage, so you pay for the amount of work the automation does. Self-hosted open models avoid those per-request charges, but they need hardware to run on. Someone also needs to keep an eye on the automation and adjust it when your process or your systems change.

We're flexible about how this is set up. AI usage can sit on your own provider accounts, and monitoring can be part of a retainer or handled as it's needed. We work with OpenAI models, Anthropic's Claude, Amazon Bedrock and open models we host ourselves.

Examples of AI automation we've built

For a legacy publishing business, we brought AI into the development team for code review, bug fixes, refactoring and data analysis, and connected the company's data to AI models through an MCP server so staff can query it from their AI agent.

We've also built automations that classify documents into categories, and connected company data to Claude Desktop so staff can combine several data sources in a single question.

If you're not sure where AI fits yet, start with our AI consulting service. If the automation needs a full application around it, that's custom software.

AI automation questions we get asked