Strategy
A ranked list of the jobs where AI pays, with the effort beside each. The rest get a plain no.
AI consulting · New Zealand
AI consulting that knows when to say no
We find the jobs in your business where AI earns its place, build agents that do them, and keep them running. Sometimes the honest answer is no AI at all.
An agent can read, sort and draft faster than anyone on your team. It cannot be held to account. So we build it to do the legwork, and a person keeps the decision.
What we do
From a ranked list to a running agent
Four kinds of AI work. Take one on its own, or all four together.
-
-
Audits
A review of an AI tool you already run, with the fixes in order.
-
Build
Agents wired into the tools you already use, each with a short list of allowed tools and a person at the gate.
-
Care
Runs watched, faults fixed, changes made as your business moves.
Where AI belongs
Where AI pays, and where it does not
AI pays where the input is messy and the volume is steady: emails, PDFs, call notes. It does not pay where a plain rule already works, or where one wrong answer costs more than the hours saved.
Limits
An agent can only touch what you allow
Four limits sit around every agent we build. You approve them in the scope before anything runs.
-
Tools
A named list: read the inbox, find an order, draft a reply. Nothing else exists for the agent.
-
Data
Only the tables and folders the task needs. Everything else stays out of the prompt.
-
Approval
Sending, paying and deleting wait for a person. The agent drafts, you press go.
-
Log
Every request, tool call and result is written down, so you can see why.
Method
Four steps, and something in hand after each
You get a written scope and one price first. Nothing starts until you say yes.
Care
Models change. Someone should be watching.
Providers update their models and your data moves on. A care plan means someone is watching and fixing.
Questions
Quick answers
Open any question
What is an AI agent?
An AI agent is software that uses a language model to read a request, decide the next step and use tools to take it, such as searching a mailbox or creating a record. A chatbot only replies. An agent acts, which is why each one needs set limits and a log.
How can a small business use AI?
Start with reading and writing work that arrives in volume: sorting an inbox, drafting replies, pulling figures out of PDFs, summarising call notes. These jobs are messy for a rule and quick for a person, which is where a model helps. Pick one job, not a platform.
How much does AI consulting cost?
Jobs run from about $500 to $200,000 and up. The price moves with how many systems the agent touches and how many steps need a person's approval. You get a written scope and one price before anything is built. See how we price a job.
Is it safe to put customer data into an AI model?
It can be, if the setup is chosen for it. Personal information in a prompt is still personal information under the Privacy Act 2020, and terms differ between providers and between consumer and business plans. We write down which data goes to which model before we build. The Office of the Privacy Commissioner publishes guidance.
Which AI models do you use?
Whichever fits the task. That might be a hosted model from a provider such as OpenAI, Anthropic or Google, or an open model run on a server you own. We name the model and the reason in the scope, so you know where your data goes.
Can an agent send or pay on its own?
Only if you decide it should. By default an agent drafts and a person approves anything that sends, pays or deletes. Once a task has run cleanly for a while, you can loosen a step. The log shows what the agent did, so you can tighten it again.
What does an AI audit cover?
An audit lists where AI would pay in your business, where a plain workflow is enough, and where to keep a person. You get it as a ranked list with the effort beside each. Already running an AI tool? We review its instructions and access, then put the fixes in order.
Can you take over an AI tool someone else built?
Yes, if we can read it. We look at the instructions, the tools it can use and its logs, then tell you plainly whether to repair it or start again. You get that verdict in writing, with a price, before we touch anything.
Can an agent use Zoho and our other tools?
Yes. An agent can connect to any tool with an API or an inbox, including Zoho, Xero, Google Workspace and Microsoft 365. Each connection is listed in the scope, with what the agent may read and what it may write.
Name a job a model could take over
A few lines is plenty. We reply within 2 business days with a couple of questions and an honest view on whether AI fits.