What AI can actually do for your business
amatis.ai learns a job the way a new staff member would. You show it the real thing, explain what it means, and it writes the job down as a procedure. Then it does that job in the background, and a person checks the result before anything is final.
Most businesses are stuck in the same place
Most owners I talk to fall into one of two camps. Some are excited about AI and plan to use it to grow, but haven’t got much past using a chatbot to help with emails and research. Others have stopped believing the hype and hope to carry on without it.
Both groups are missing the same thing: how AI actually gets a job done inside a business.
Why the first try disappoints
Here’s what usually happens. You give an AI chatbot a real task. It can do it, but it’s slower than doing the task yourself. So you decide AI isn’t worth it.
That first slow attempt isn’t the result. It’s the training. A chat window can’t take you past that point. amatis.ai is built to.
A real example: the gin awards at Matahui
Matahui is a small craft distillery in Katikati. Guests come in to make their own gin, writing their recipe on a sheet we provide, and we make the gin to that recipe. Guests got competitive, so we started a monthly award, and entries quickly grew into the hundreds each month. Every one meant recording the recipe, meeting compliance requirements, collecting tasting notes, adjusting scores so each judge’s are fair against the others, and announcing the winners. It was a surprising amount of admin.
The first try. We automated recipe capture in a chat session. It worked, but each recipe took up to 30 minutes. That was slower than typing it in by hand. If that had been the only option, we would have given up.
How it got better. Each session was also teaching. We gave the agent a recipe sheet and explained what everything on it meant. Each time an unusual case came up, we updated a written standard operating procedure. Once the process ran smoothly, the agent turned that procedure into a skill.
Now. We upload a batch of recipe sheets and leave. The agent sets each one up as a task and works through them in the background. An hour or so later, each recipe is waiting with the photo of the sheet beside what the agent read from it. Someone checks it and clicks Approve, and the agent enters it into the right tables in the database. At the end of the month it adjusts the judges’ scores and names a winner for us to contact. When a past guest wants to reorder, a second skill finds their recipe and adjusts it for the bottle size they’ve ordered.
A batch of ten now takes less of our time than a single recipe used to.
public/images/what-it-does-recipe.jpg — brief in docs/image-prompts.mdHow it works
Show it the real thing.
A form, an email, a spreadsheet: the actual material, not a description of it.
Explain what it means.
In plain conversation, the way you’d explain it to a new staff member.
Write the procedure down.
Each unusual case gets added to a standard operating procedure your business owns.
Turn it into a skill.
Once the procedure runs smoothly, the agent turns it into a repeatable skill.
Hand over as much as you’re comfortable with.
Every task can run entirely manually, with an agent drafting and you approving each step. It can run on a schedule that still asks for sign-off. Or it can run fully on its own once you’ve seen it work reliably. Nothing runs unsupervised until you’ve decided it’s earned that.
The AI Chief of Staff is the newest part of amatis.ai. It holds a profile of your business and starts by mapping how your business runs today, before any AI is involved. From there it guides you through moving each part of the work to AI, and keeps an eye on the whole system so the work stays pointed in the right direction.
What it won’t do
It won’t be faster the first time.
Expect the first runs of a new job to be slow. That time builds the skill.
It won’t choose the right job for you.
Most AI disappointments start with automating the wrong thing. Knowing where to start is the hard part, and that’s what the free AI Audit is for.
It won’t replace the people who make your business what it is.
At Matahui, the gin being made by people is part of what guests pay for. The AI handles the paperwork, and the team has embraced it because it makes their jobs easier.
One shared space for your team
Your team and your AI agents work in the same place, on the same information, and can see and build on each other’s work. Give each person access to just the parts they need. Your marketing person can have the marketing tools, calendar and documents without seeing your accounts.
It connects to what you already run
amatis.ai connects to Xero, Google Analytics, Mautic, CiviCRM and the main social media platforms. You’re not replacing anything or migrating anywhere. If you need a connector that doesn’t exist yet, it gets built when you need it, usually within about a week.
Your data stays in New Zealand
amatis.ai only uses AI models on zero-data-retention terms. The model providers keep none of your data and never train on it. amatis.ai is a New Zealand company operating under New Zealand’s strict privacy laws. We keep our own logs and review them to improve the system. If you’d rather nothing ever left your own server, you’ll be able to run amatis.ai on your own server or ours. Enquire about early access
The architecture keeps sensitive credentials, the passwords and access codes that connect AI tools to your other systems, out of logs and out of a model’s context in the first place, rather than relying on policy or good intentions.
Any model. No lock-in to one vendor.
Use whichever model suits the task. Relying on a single AI provider is a real business risk. In June 2026, a US export-control order led Anthropic to switch off two of its models worldwide, overnight, for every customer. Businesses built around those models lost access with no warning and no say. amatis.ai is built so that if one model becomes unavailable, restricted, or stops being the right tool, you’re not stuck.
Built so you can leave
amatis.ai uses open formats and standards wherever it can, like markdown documents and Agent Skills, for the procedures and skills it builds for your business. If you ever leave, what you’ve built goes with you. The code itself will be public, so you’ll always be able to run it yourself.
It gets better quickly
Feedback you give inside amatis.ai is usually acted on within a week.
Not sure which job to start with? The free AI Audit finds it.