Service 02 · One-off build · client-owned
AI Agent Deployment
A custom AI built for one job. You own it.
What it is
A purpose-built AI agent for one defined job in your business, scoped to you, trained on your real examples, tested against your actual work, and handed over ready to run. You own it outright. No monthly fees, no subscription, no lock-in to us.
What it solves
There's a job in your business that needs a brain, not just a rule: drafting quotes, qualifying leads, answering the same customer questions. It takes judgement, so it lands on you or your most experienced person, and it eats hours. Generic AI doesn't know your business, your prices, your voice, or your process. A purpose-built agent does.
Workflow vs agent: a workflow follows fixed rules (Service 01). An agent makes a judgement and produces bespoke output, the right tool when a step needs thinking, not just steps.
Agents we build
| Quote drafter | Reads a job description → drafts a quote in your format and pricing → you review & send. |
| Lead qualifier | Reads an enquiry → scores it → asks the right follow-ups → books a call. |
| Customer enquiry agent | Answers FAQs and booking questions in your voice, around the clock. |
| Intake agent | Collects client info, qualifies, routes to the right person. |
| Reorder agent | Watches stock levels → drafts supplier reorder emails. |
What you get
- A custom AI agent built for one defined job
- Trained on your real examples: your format, pricing, voice
- Tested against your actual scenarios before it goes live
- Deployed into your environment (email / your tools)
- Video walkthrough + a plain-English guide
- You own it, no subscription, no dependency on us
- Two-week check-in included
Price
| Package | Price (NZD, GST-excl) |
|---|---|
| Standard Agent: one defined job | $3,000 fixed |
| Advanced Agent: multi-step / autonomous | $6,500 fixed |
| Ongoing management (optional) | see Fractional AI Operations |
How it works
| 1 · Define the job | We write the agent's "job description": inputs, outputs, what good looks like. You sign it off. |
| 2 · Build & train | Built on Claude (Anthropic), trained on your real examples. |
| 3 · Test | Run against your real scenarios until it's reliable. |
| 4 · Deploy & hand over | Live in your tools, walkthrough, plain-English guide. |
| 5 · Settle in | Two-week check-in to tune anything. |
Common questions
- What does "you own it" actually mean?
- The agent's prompts, configuration, and documentation are yours, running in your accounts. The AI model behind it (Claude) is a pay-as-you-go API in your name, typically a few dollars to a few tens of dollars a month depending on volume. You can maintain it yourself, hire anyone, or come back to us.
- Can it make decisions without me?
- Only where you say so. Most agents we build draft-and-queue: the agent does the thinking, a human approves the output. Full autonomy is a choice you make per task, not a default.
- What if it gets things wrong?
- It will sometimes. That's why we test on your real scenarios before go-live and set hard boundaries it can't cross (never quote outside your price book, never promise dates). The two-week check-in tunes whatever the first fortnight surfaces.
Start with the intake formDescribe the job that eats your hours. We'll tell you straight if an agent is the right tool for it.