Fixed scope, quoted in writing before anything starts. An ops teardown is $250 AUD (45-minute call plus a written automation plan, credited toward any build). A one-week automation sprint is $990 AUD. Larger systems usually land between $2k–$4k AUD. No retainers, no hourly billing.

How long does it take from first contact to go-live?

A teardown call is usually booked within a few days. A sprint is one week end to end. Chatbots typically go live in 1–2 weeks; larger workflow systems in 4–8 weeks. You get dated milestones in writing before work starts.

What happens after launch — do you keep running it, or do I take over?

You own it. Everything runs in your accounts, on your keys, and we hand over the code, the docs and the logins at the end of the build. Nothing is locked to us.

If something breaks or needs a tweak, how is that handled and at what cost?

Anything we built that breaks is fixed free for 30 days after handover. After that, fixes are quoted per job or covered by an optional monthly care plan — your choice, and never a condition of the build. New scope is quoted as a new fixed-scope sprint.

How much of my time does this take?

About two hours in total: one 45-minute call, giving us access to the tools involved, and one review round before go-live. We do the rest.

What proof do you have that this works?

Named, checkable builds rather than anonymous case studies: Ebb (Stripe revenue recovery), Australian Power (WhatsApp quote automation), Second to None (painting and renovation ops), Invest With Alison (financial-education funnel) and an AI tools directory. Each one is written up on this site with what was built and what changed.

CASE STUDY / CONTENT INTELLIGENCE

A system that notices before anyone else does.

We built and we run a content intelligence system for a property media business in Australia. It watches a whole content landscape without being asked, works out what is actually breaking out, explains why, and hands the operator a draft in their own voice. It has been running in production since 2026.

Build something like this→See the rest of our work

CLIENT NOT NAMED AND SOURCES NOT LISTED, BY AGREEMENT. WHAT FOLLOWS IS THE SHAPE OF THE SYSTEM, NOT ITS CONTENTS.

TaskDriver builds and operates production AI systems for businesses that would rather own the machine than rent another dashboard.

TaskDriver.ai is an AI automation and web studio in Auckland, New Zealand. We build workflow automation, chatbots, CRM integrations and websites for small and mid-sized businesses across New Zealand and Australia — fixed scope, no retainers.

THE PROBLEM

Content teams do not lose to better ideas. They lose to latency.

PROBLEM 01

The signal arrives too late.

A piece of content breaking out is only useful while it is still breaking out. By the time anyone notices it by scrolling, the window to respond has closed and the idea is everywhere.

PROBLEM 02

Watching is a full-time job nobody wants.

Keeping an eye on a whole content landscape, across several platforms and two languages, is hours of manual scrolling a day. It is the first task a small team drops, and the one that quietly costs them the most.

PROBLEM 03

Knowing is not the same as acting.

Spotting that something worked does not tell you why it worked, or what your version of it should say. That gap between the observation and the draft is where most content teams stall.

HOW IT WORKS

Watch, detect, explain, draft, deliver.

WATCH01

Continuous collection.

The system runs around the clock on its own infrastructure, pulling public performance data on a fixed cadence. It never sleeps, never forgets a source, and keeps a full history so today can be judged against what normal actually looks like.

DETECT02

Breakout detection, not vanity metrics.

Raw numbers lie. A post is measured against its own source's baseline at the same age, so a modest account having an unusually big day is caught, and a large account having an ordinary day is ignored. Platforms that hide view counts are handled on their own terms rather than faked.

ANALYSE03

AI explains why it worked.

Anything that clears the bar goes to Claude for a structured breakdown: the hook, the structure, the emotional lever, the reason this one travelled. The output is a briefing a human can argue with, not a score.

ACT04

Draft-ready, in the client's voice.

The same pass produces adapted hooks, titles, and a full script, written against the client's own brand guidelines and back catalogue. The operator picks a direction and edits, instead of starting from a blank page.

DELIVER05

One channel: WhatsApp.

No new dashboard to remember, no login. The alert lands in WhatsApp, the operator replies in WhatsApp, and the whole pick-a-hook, pick-a-title, approve-the-script flow happens in the thread. A web dashboard exists for the deeper view, but the day-to-day never requires it.

REPORT06

A monthly market layer.

A second system collects published market data as it is released and compiles it into a client-ready monthly report. Research that used to be a full day of chasing PDFs is assembled automatically and reviewed instead of written.

WHAT IT CHANGED

The research team is now a phone.

24/7UNATTENDED OPERATION

Self-hosted, monitored, with heartbeat checks on every scheduled job.

MinutesDETECTION TO BRIEFING

From a post crossing the bar to a full analysis in the operator's hand.

5PLATFORMS, TWO LANGUAGES

Including platforms that publish no view counts at all.

1PERSON TO RUN IT

The output of a research team, operated from a phone.

THE BUILD

Six layers, one owned system.

No per-seat SaaS stack, no vendor holding the data. The client owns the system, the history, and the cost curve, and we operate it.

COLLECTION LAYER

Scheduled workers per platform with per-source history, retry handling, and heartbeat reporting so a silent failure surfaces the same day instead of a month later.

DETECTION LAYER

Age-matched baselines per source with hard floors and cold-start handling, tuned so the operator gets a small number of alerts worth reading rather than a feed worth muting.

AI LAYER

Model routing by task, cheap models for triage and expensive ones only for deep analysis, with prompt caching on the stable instructions. The cost per analysed post is measured and kept deliberately low.

OPERATOR LAYER

A WhatsApp bot with access control, claim locks so two people cannot work the same alert, quiet hours, and a state machine that walks an alert from detection to approved script.

KNOWLEDGE LAYER

The client's own back catalogue, transcripts, and past scripts indexed and searchable, so generated drafts are grounded in what they have actually said before.

REPORTING LAYER

Automated collection of published market data plus generation of a formatted monthly report, delivered ready for review.

Most of this is not about content.

Watch a stream of public data, decide what is unusual, explain it, and put the answer in front of the one person who can act on it. That pattern fits pricing, hiring, support queues, and compliance just as well. If you have a stream nobody has time to watch, that is the conversation.

Book a call→What we build

How we work with you

The answers people ask for before they book — cost, timeline, ownership, and what happens when something breaks.

What does it cost?: Fixed scope, quoted in writing before anything starts. An ops teardown is $250 AUD (45-minute call plus a written automation plan, credited toward any build). A one-week automation sprint is $990 AUD. Larger systems usually land between $2k–$4k AUD. No retainers, no hourly billing.
How long does it take from first contact to go-live?: A teardown call is usually booked within a few days. A sprint is one week end to end. Chatbots typically go live in 1–2 weeks; larger workflow systems in 4–8 weeks. You get dated milestones in writing before work starts.
What happens after launch — do you keep running it, or do I take over?: You own it. Everything runs in your accounts, on your keys, and we hand over the code, the docs and the logins at the end of the build. Nothing is locked to us.
If something breaks or needs a tweak, how is that handled and at what cost?: Anything we built that breaks is fixed free for 30 days after handover. After that, fixes are quoted per job or covered by an optional monthly care plan — your choice, and never a condition of the build. New scope is quoted as a new fixed-scope sprint.
How much of my time does this take?: About two hours in total: one 45-minute call, giving us access to the tools involved, and one review round before go-live. We do the rest.
What proof do you have that this works?: Named, checkable builds rather than anonymous case studies: Ebb (Stripe revenue recovery), Australian Power (WhatsApp quote automation), Second to None (painting and renovation ops), Invest With Alison (financial-education funnel) and an AI tools directory. Each one is written up on this site with what was built and what changed.

Who built this work

Robert Iuoras

Founder & lead architect

Scopes, designs and ships every TaskDriver build — the person on your first call is the person writing the code. No account managers, no handoffs between strategy and delivery.

Built Ebb, a Stripe revenue-recovery product, end to end
Shipped WhatsApp quote automation for Australian Power
Delivered ops automation and web for Second to None and Invest With Alison