AI agents for daily operations

What an agent is, without the hype

An AI agent is a small program that reads something, decides what it means, and does the next step. That is genuinely it. The interesting part is not the technology — it is that a lot of office admin is exactly that shape, and always has been.

Reading an email and working out whether it is an order, a complaint or a supplier invoice. Pulling the numbers off a PDF and putting them in the right row. Noticing that a job finished three weeks ago and nobody has invoiced it. None of that needs judgement so much as it needs someone to actually do it, every day, without forgetting.

What Kiwitech actually builds

Agents are scoped to one job each. An agent that does one thing reliably is worth more than one that does six things and needs checking.

• Quoting and invoicing — draft a quote from an enquiry, raise the invoice when the job is accepted, and chase the ones that go unpaid.
• Inbox and document handling — sort what arrives, pull the details out of attachments, and file it where it belongs.
• Scheduling and chasing — book the thing, confirm the thing, remind the person who has not replied.
• Data entry — move information between systems that have no integration, without a person in the middle.
• Monitoring and tuning — once it is live, Kiwitech watches it and adjusts. An agent that drifts and nobody notices is worse than no agent.

Where a person stays in the loop

Anything that costs money, commits you to something, or goes to a customer under your name gets a human approval step unless you explicitly decide otherwise. That is the default, and it is the right default.

Kiwitech will also tell you where an agent is the wrong answer. Work that needs real judgement, carries legal weight, or happens twice a year is usually cheaper to keep doing by hand. Automating something rare is a good way to spend money proving a point.

Your data

You own everything the agent touches. Kiwitech builds on your accounts where possible, does not train anything on your data, and will tell you exactly which services your information passes through before anything is built.

For a New Zealand business that also means being straight about where data goes. Some AI services process offshore. If that matters for your business — and for some industries it does — say so at the first call and it changes what gets built.

What it costs and how long it takes

Fixed price, agreed before the work starts. Most first agents are live in two to six weeks, and Kiwitech deliberately picks a small one first so it proves itself before anything bigger gets built.

Ongoing monitoring is a separate, smaller arrangement. Agents need looking after — inboxes change, suppliers change their invoice format, and something that worked in March quietly stops working in June if nobody is watching.

Common questions

What happens when the agent gets something wrong?

It gets caught at the approval step, or by the monitoring. Kiwitech builds agents so that being unsure results in asking a person rather than guessing, and the ones that touch money or customers have a human check by default.

Will this replace someone on my team?

That is not usually what happens or what Kiwitech is asked for. The common outcome is that the admin that was eating a day a week stops eating it, and that person does the work you actually hired them for.

Does my data get used to train AI models?

No. Kiwitech does not train anything on your data, and will name every service your information passes through before the build starts so you can decide whether you are comfortable with it.

Talk to us about it

Send Kiwitech the job you dread most. If it can be automated we will tell you how and what it costs, and if it cannot, we will tell you that too. Get in touch.

The other two things Kiwitech builds: Automated AI marketing funnels, and Bespoke software and integration.