AI work for Australian businesses, with the receipts.

Problem

A small concierge team was responsible for hundreds of high-value member requests per week across email and WhatsApp. Response time pressure was eroding the experience that members were paying for, and after-hours coverage was carried by individual concierges working off the clock.

Approach

We designed and built a domain-tuned triage agent that ingests member messages from email and WhatsApp, classifies intent, drafts a contextual first response in the firm's voice, and surfaces the request to the right concierge with the right context already attached. Edge cases (vulnerable member signals, VIP escalations, anything outside policy) are routed straight to a human with no AI response sent.

Outcomes

• ·First-response time on routine member requests reduced from hours to minutes
• ·After-hours coverage shifted from individual goodwill to system default
• ·Concierges concentrate on the high-judgement requests that justify the membership fee

Problem

Australian workforce conversation was being shaped by US-centric data, and most existing AI exposure scores were either too coarse to be useful or built on assumptions that did not reflect a trade-heavy economy. Policymakers, careers advisors, and the media needed a credible Australian view.

Approach

We built an interactive analysis of 358 ANZSCO occupations using Jobs and Skills Australia data, ABS labour force figures, live SEEK job ad scraping, and a multi-model consensus across Claude, GPT, Gemini, and Grok. Each occupation got an exposure score, a workforce-size figure, and a short narrative on whether AI was enhancing or replacing the work.

Outcomes

• ·Featured on The Morning Show (Channel 7), April 2026
• ·349 of 358 occupations identified as augmentation rather than replacement
• ·Picked up by careers and policy stakeholders as the Australian reference dataset

Problem

A long-established luxury events business had grown beyond what its older website could carry. Enquiry quality was inconsistent, brand presentation no longer matched the standard of the work, and the team had no leverage from the digital footprint when pitching premium clients.

Approach

Working from a refreshed brand style guide, we rebuilt the site in Next.js with deliberate restraint: strong typography, considered motion, photographic editorial imagery rather than stock. We instrumented enquiry tracking and sequenced AI-assisted content drafting against the brand voice so the team could keep the site moving without going off-tone.

Outcomes

• ·Enquiry quality lifted measurably in the first quarter post-launch
• ·Editorial cadence sustained without external copywriting spend
• ·Site holds up next to the international peers the client is competing with

Problem

A construction quality management platform held a rich but under-used dataset of inspections, defects, and incidents across hundreds of active projects. Customers wanted insight rather than reports, but existing analytics gave them little more than counts and pie charts.

Approach

We worked with the product team to identify the questions head contractors were actually asking (where defects cluster on a job, which trades concentrate risk, what changed since last week) and built AI-assisted intelligence into the existing platform. The work included data foundations, prompt evaluation harnesses, and a deployment pattern that respected per-project commercial separation.

Outcomes

• ·Inspection and defect data turned into project-level intelligence customers act on
• ·Cross-project pattern detection delivered without breaching commercial confidentiality
• ·Product team set up to extend the AI surface without external help

Problem

A property and construction group was running multiple major projects simultaneously, with cost, program, and risk data scattered across project management, finance, and contractor systems. Executive leadership was reconciling four reports a week to get a portfolio view that should have been one screen.

Approach

We built a portfolio dashboard that consolidated project data into a single executive view, with AI-assisted commentary highlighting drift, risk concentration, and exception handling priorities. The system was designed to integrate with the existing tooling rather than replace it, with role-based access for executive, project director, and commercial users.

Outcomes

• ·Executive portfolio review compressed from a multi-day cycle to a live dashboard
• ·Cost and program drift surfaced earlier, with commercial impact quantified
• ·Project directors retained their existing systems, with no workflow disruption

Problem

An aged care operator was spending hundreds of staff hours per accreditation cycle assembling evidence against the Aged Care Quality Standards. Quality leadership was on a treadmill of report compilation rather than actually improving practice. Incident reporting suffered the same drag.

Approach

We designed an AI-assisted layer over the operator's existing care management and incident systems that drafted standards-aligned evidence packs as a by-product of normal operation. Quality and clinical leadership reviewed and signed rather than building from scratch. Privacy impact assessment, clinical governance review, and an audit trail were built in from day one.

Outcomes

• ·Accreditation evidence preparation reduced from project to ongoing readiness
• ·Quality team time redirected from data entry to pattern detection and remediation
• ·Clinical governance and ACQSC documentation defensible under audit