Services · AI workflows
We look at how work actually moves through your business — enquiries, quotes, emails, data entry, content — and build AI into the steps where it saves real time. Not because it's fashionable. Because it pays for itself.
How we think about it
Most AI projects that disappoint started with "we should be using AI" rather than "this job takes nine hours a week". We start from the job.
We map how the job is done today, where the time goes and where mistakes creep in — then decide which steps AI should take on.
AI drafts, sorts and summarises. Anything that commits your business — a price, a refund, a promise — waits for a person.
Built around your process, not one vendor. If a better or cheaper model comes along, we swap it without rebuilding the workflow.
Each step only sees what it needs. Sensitive steps can stay on NZ-hosted or self-hosted systems, under the Privacy Act 2020.
What we build
Every business is different, but these are the workflows we're asked about most. Each can start small and grow once it's proved itself.
Incoming emails and web enquiries sorted, summarised and routed — with a draft reply ready to approve.
Example: a trades business gets 40 enquiries a week. Each one is tagged by job type and suburb, urgent ones are flagged, and a tailored first reply is drafted for the owner to send with one click.
Site notes, photos or a call summary turned into a first-draft quote or proposal in your own template.
Example: after a site visit, voice notes and photos become a structured quote using your price list and terms. You check the numbers, adjust, and send — instead of typing it up at night.
Invoices, dockets, order forms and PDFs read and keyed into your systems, with anything odd flagged.
Example: supplier invoices arrive by email; line items are extracted, matched to purchase orders and prepared for your accounting system. Mismatches go to a person instead of slipping through.
Product descriptions, FAQs, alt text and structured data drafted at scale — reviewed before anything goes live.
Example: a WooCommerce store with 2,000 thin product pages gets consistent descriptions and schema markup drafted from supplier data, queued for review in batches, then published.
Multi-step jobs across your website, CRM, stock and accounts — joined up through APIs and webhooks.
Example: a new online order checks stock, books a courier, updates the customer record and drafts a personal thank-you — only stepping out to a person when something doesn't add up.
Staff ask questions in plain English and get answers from your own manuals, policies and price lists.
Example: new staff ask "what's our returns policy on custom orders?" and get the answer with a link to the source document — instead of interrupting whoever's been there longest.
How a workflow gets built
No big-bang rollout. Each workflow proves itself on real work before it grows — and if it doesn't, we stop and tell you why.
Discuss an AI workflowOn a Teams or Zoom call, you walk us through how the work is done today: who touches it, which systems, how long it takes and where it goes wrong.
We mark which steps AI should handle, which need a person's approval, what data each step needs, and the likely running cost per job.
We build one workflow with clear limits, run it alongside how you work now, and measure: time saved, accuracy, and anything it got wrong.
If the numbers stack up, we tighten it, widen it, or move on to the next job. If they don't, we switch it off and you've lost a pilot, not a project.
Fit
Questions
Whichever suits the job. We work with the major providers, including OpenAI, Anthropic's Claude and Google's Gemini, and with smaller or self-hosted models where data or cost makes that the better call. We design the workflow so the model can be swapped later without rebuilding everything.
Only what the workflow needs. We map exactly which data each step touches before anything is built, and sensitive steps can stay on NZ-hosted or self-hosted systems. It's your data under the Privacy Act 2020, and we treat it that way.
Wherever it matters, yes. Drafts can wait for approval, anything outside set limits is flagged to a person, and every action is logged so you can see what happened and why. How much runs automatically is your decision, and it can change as trust builds.
It depends on how many systems the workflow touches, how much volume runs through it and how much review it needs. Every build is scoped and quoted at a fixed price before work starts, and we estimate the ongoing AI usage costs up front. All prices exclude 15% GST.
Then we'll say so. Plenty of jobs are better solved with a simple rule, a better form or an integration between two systems, which is often cheaper and more reliable. You'll get the honest recommendation, even when it's the smaller job.
With a conversation. Tell us the job that takes the most time or causes the most errors, and we'll set up a Teams or Zoom call to walk through how it works today. From there we can tell you whether it's a good candidate and what a small pilot would look like.
Next step
Tell us the job that eats the most time. We'll set up a short Teams or Zoom call, walk through how it works today, and tell you honestly whether AI will help.