The AI Era of Service.com.au Has Started. Here's What We've Shipped and Where It Goes.

Less than a few minutes after we switched on our AI quote assistant in late April, a homeowner used it to post a lawn-mowing job. The whole conversation took her about 30 seconds. Five minutes after the job went live, the first response landed.

That one interaction is the entire thesis. Posting a job used to mean a long static form: pick a category from a list you don't understand, describe a problem you can't name, hope it routes to the right trades. Now it's a conversation. A month after launch, 80% of homeowners posting jobs on Service.com.au are doing it through the assistant.

This is the first wave of a much larger rebuild, and I want to lay out what's actually shipped, what the early numbers say, and where it's heading.

What's live right now

A conversational way to post a job. The AI Quote Request Assistant walks a homeowner through posting via natural conversation instead of a form. Send it a photo of the cracked tile and it works out what it's looking at, then asks smarter follow-ups. A bathroom reno gets different questions than a garden tidy-up. It resolves the right trade category itself, so the homeowner never has to know whether their problem is "plumbing" or "roofing" or both. Need a plumber and a gardener? It splits the conversation into separate requests, then asks if there's anything else.

AI summaries on every new job. Each quote request now arrives in a tradie's inbox with a generated title and one-line summary. A vendor reads one line and decides in seconds whether the job is worth quoting, instead of opening every request and parsing a homeowner's free-text description.

A redesigned jobs inbox. Rebuilt around those summaries, with clearer distance display and an at-a-glance count of new jobs waiting. The summaries and the inbox shipped as a pair on purpose: better job intelligence is worthless if the surface can't carry it.

Featured Reviews. Tradies can pin their best review to the top of their listing and have it lead their directory card. Not an AI feature, but part of the same release wave, because the goal across all of it is identical: put the strongest signal in front of the right person at the moment they're deciding.

Behind the customer-facing pieces sits instrumentation you'll never see. Conversation-level tracing tells us exactly where homeowners hesitate or drop out of the assistant flow, which is how the next iteration gets designed from evidence rather than opinion.

The scorecard

The completion numbers favour the assistant: in our split test against the static form, homeowners who start the AI flow finish at a higher rate, and they get matched faster. The 80% adoption figure a month in tells its own story. We're still tuning the entry point on the homepage, but the data that matters, completed submissions turning into real jobs, says we're on the right side of this rebuild.

Why a tradie marketplace, and why now

Marco Zappacosta at Thumbtack made a point recently that I keep coming back to: it's easy to see what AI substitutes, and much harder to see what it complements. Home services is firmly in the second category. No model is going to clear your blocked drain. What AI can do is dissolve the friction around the work, the describing, the matching, the triage, the admin, that wastes hours on both sides of every job.

Most home projects don't start with clarity. They start with a leak, a crack, a noise. The old internet's answer was to hand that uncertainty to the homeowner: you go figure out what trade you need, then fill in our form. The assistant flips it. Describe the problem in your own words, show a photo, and let the platform carry the burden of turning that into a well-routed job.

On the tradie side the equation is even simpler. A tradie's scarcest resource is time off the tools. Every minute spent reading vague job descriptions is a minute not quoting, not working, not invoicing. AI summaries exist so a sparky in the ute between jobs can triage the inbox at a red light.

An AI layer is only as good as the data underneath it, and ours sits on top of hundreds of thousands of real quote requests since 2015, millions of notifications for work sent to tradies, and a network of 30,000+ businesses whose ABN, licence, and insurance have been checked against the governing body for each trade in each state. Smart matching plus regulator-backed credentials beats smart matching alone, every time.

Where this goes

The pieces that shipped in April and May are infrastructure for what comes next. Three threads, in rough order:

Better leads, properly qualified. The same intelligence that summarises a job can score it, urgency, completeness, fit, so the matching between a job and the tradies who see it gets sharper. The endpoint is fewer, better leads rather than more, noisier ones.

A homeowner experience that doesn't end at "job posted." We're prototyping what happens after submission: a quieter copilot that tracks quotes, answers questions, and keeps a multi-job project organised, rather than a dashboard you have to remember to check.

Less admin for the businesses we serve. AI-drafted listing profiles already cut setup time for new businesses joining the platform, and drafted review replies are next in the queue. The pattern repeats: anywhere a tradie types when they could be working, we want the machine to do the first 80%.

The vision fits in a sentence, better leads, less admin, more jobs won, but the build behind it is the most ambitious stretch of product work in this company's 13 years. Watch a first quote arrive minutes after a homeowner finishes a 30-second conversation, and you can see the shape of where the whole category is going.