Machine learning and AI
Most businesses do not need their own AI model. They need the right jobs handed to AI, connected to the systems they already use, and checked by people. We help New Zealand businesses work out where AI pays for itself, then build it.
The problem
Licences are bought before anyone has decided which job the AI is meant to do.
A demo works on ten tidy examples and falls over on real documents.
The system sounds sure of itself, and there is no way to see where an answer came from.
Customer details end up in a free chatbot because nobody offered an approved alternative.
What we build
Advice when you are deciding, and a build when you have decided. Most projects start small and grow.
A short, practical review of where AI will save time or money in your business, and where it will not. You get a ranked list of jobs worth automating, what each would cost to build and to run, and the risks to manage.
Custom AI built into your product or your back office: tools that read documents and pull out the facts, search that understands a question, assistants that draft and summarise. We built Mindvault, an AI study platform where every answer cites its source and running costs are tracked per subscriber.
AI connected to the software you already run, so it can look things up and take action: your CRM, inbox, accounting system, help desk and website. We use models from providers such as Anthropic and OpenAI through their business APIs.
Chatbots that answer from your own content, in your tone, and hand over to a person when they should. On your website, in your app or inside the tools your staff use.
Agents that carry a task through several steps on their own: read the enquiry, look up the order, draft the reply, update the record. You decide which steps they may take alone and which wait for a person.
For the cases where a ready-made model is not enough: a forecast, a score or a classification learned from your own data, such as demand, churn or risk. We check first whether your data can support it, and say so plainly if it cannot.
Worth knowing first
The projects that pay back start with one job that is repetitive, well understood and measurable: answering the same enquiries, reading the same forms, writing the same first draft. Pick the job, measure how long it takes now, then decide whether AI should do it.
Strategy comes after the first result, when you know what the technology does with your own data.
If Microsoft Copilot, ChatGPT or a feature inside software you already pay for does the job, use it. It will cost less than anything we can build, and we will tell you that.
Building makes sense when the AI has to use your own data, take action in your systems, follow your rules or sit inside your product. That is the work we do.
AI has two costs: building it and running it. Every request to a model is charged by the amount of text read and written, so a tool used a thousand times a day costs more to run than one used ten times.
We estimate the running cost before we build, show it on a dashboard afterwards, and put limits in place so a busy month does not bring a surprise bill.
How we work
The plan
Tell us which jobs eat your team's time. We ask how they are done now and what a good result looks like.
We build a small working version on your real examples, so you can see whether it is accurate enough before you commit.
We connect it to your systems, add the checks and approvals, and test it with the people who will use it.
We put it live, watch accuracy and cost, and tune it as you use it.
Good to know
Something else on your mind? Call (09) 951-9554 or send us a message.
Looks at how work gets done in your business, finds the jobs where AI would save real time or money, and tells you which are worth doing first and which are not worth doing at all. You should come away with a short list, costs and risks, not a slide deck.
Almost never. Models from providers such as Anthropic and OpenAI, given your own documents and data at the moment they answer, cover most business uses. Training your own model makes sense only when you have a lot of your own data and a specific prediction to make.
Machine learning is the method: software that learns patterns from examples instead of following written rules. AI is the wider term, and today it usually means large language models that read and write text. Most business projects now use an existing model. Custom machine learning is for forecasts and scores from your own numbers.
We use business APIs whose terms say your data is not used to train their models, send only what the task needs, and tell you where it is processed. Sensitive fields can be removed before anything leaves your systems.
Mostly Claude from Anthropic and the GPT models from OpenAI, chosen for the job: accuracy, speed, cost and where the data is processed. We build so the model can be changed later.
Usually. If your CRM, accounting system, help desk or website has an API, an AI assistant or agent can read from it and act in it, within limits you set.
By having it answer from your own documents and show the source, by testing on real examples before launch, and by sending anything it is unsure about to a person instead of guessing.
There are two costs: building it and running it. We quote the build at a fixed price and estimate the monthly running cost before you commit.
A first working version of a single job usually takes a few weeks. Connecting it to your systems, testing and rollout come after that. You get a timeframe in writing with your quote.
Tell us which jobs eat your team's time. We will tell you which ones AI can take, and what it would cost.
Goes well with
Tell us what you need. We’ll come back with ideas, a timeframe and a fixed quote.
