AI for an Australian small business, built to the criteria you sign off

Not a subscription with a chat box. A system built on your documents and your workflow, with the acceptance criteria agreed before anyone builds, and handed over so your team can change it.

Where it usually gets stuck

Most of the AI on offer to a small business is a subscription: a chat widget, a phone bot, a monthly fee. It answers from a script, not from your documents, and when it gets something wrong nobody can tell you why.

The work that actually costs a small team its evenings is specific: finding the answer that is in a PDF somewhere, sorting the inbox, keeping the social accounts fed, chasing the leads that came in last week. Each one has a process, and a process can be built.

What matters is whether the result can be checked: what counts as done, who signs it off, and what happens on the day it gets an answer wrong. Whether AI is involved comes second.

The three lines that apply

Each line states what you need, what we provide and what you receive. The images are finished work, not concepts.

  • grounded-chat: two sentences cited, the third, a planted fabrication, flagged unsupported

    01

    Agents and knowledge systems

    RAG agents: customer service, after-sales and internal knowledge Q&A

    You need
    Staff and customers get answers from the company's own documents, every answer traceable to its source, and the material need not leave your environment.
    We provide
    A knowledge assistant for internal documents: local retrieval, answers with citations, every sentence scored against the retrieved passages, unsupported ones flagged immediately. Private deployment supported.
    You receive
    An agent running on your material, an evaluation dataset with recall and ranking figures, and handover documentation.
    See the work · grounded-chat
  • 概念图:邮件、账号、对话与搜索
    Concept image: mail, accounts, conversation and search

    02

    AI-driven operations

    Email and mailboxes, social accounts, search and AI visibility, organised and run

    You need
    The inbox sorted, the social accounts published on schedule, the website represented accurately in search results and AI answers, without taking up your people.
    We provide
    Sorting, archiving and follow-up rules for email; content scheduling, publishing and review for social accounts; keeping the site visible in search engines and AI answers. Automation runs the routine, people handle only what needs judgement, every step logged.
    You receive
    Operations flows that run, an organisation standard for accounts and mailboxes, weekly review records.
    Describe your case
  • gig-radar: 809 leads fetched, 64 sent after model scoring, 16% reply rate

    03

    Growth and e-commerce operations

    Online strategy, planned and executed: brand, hero images, listings, conversion and product selection

    You need
    Leads surface every day, already scored and drafted per channel, with a person reviewing only the last step.
    We provide
    Lead sourcing, model scoring, drafting per channel, automatic reply scanning after sending; runs daily unattended, every step recorded.
    You receive
    A lead flow that keeps running, with its records. Our own gig-radar runs this way: 809 leads fetched, 64 sent after scoring, 16% reply rate.
    See the work · gig-radar

How it starts

01

Map the workflow

We look at how the work runs today and where the time and money go. Facts only, no proposal yet.

02

Pick one testable case

One scoped case, with the success criteria and the acceptance method agreed in writing before anyone builds.

03

First working version in two to four weeks

A version that runs, not a prototype demo. If it misses the criteria fixed in step two, we keep working until it meets them, within the same scope.

04

Hand it over so you can change it

You get the docs and the acceptance scripts, plus how to change it later. If you stop working with us, it keeps running.

Questions we get asked

  • Do you sell software or build it?We build it, on your material and your workflow, and hand it over with documentation and acceptance scripts. There is no subscription to us; the system keeps running if you stop working with us.
  • How do we know it works?The success criteria and the acceptance method are written down before we build. The first version is accepted against them, not against a demo. The numbers on this site can be reproduced.
  • What does it cost?Priced per scoped case, not per hour. A number comes after the process has been mapped and the case scoped. There is no price list.
  • Where does our data live?Local retrieval and private deployment are supported; company material does not have to leave your environment. Which model handles which step is stated in writing.
  • Where are you?Melbourne. Hours 09:00 to 22:00 AEST, by email, phone or WhatsApp.

Tell us which step costs you the most

Contact us