AI Projects
I don't just advise on AI.
I build with it.
The best technology advice comes from people who build. These are AI products, tools and frameworks I've designed and shipped myself — from live SaaS in market, to a productised website business, to governance tooling for enterprise retailers and the autonomous agent that runs my own practice.
ProductMatch
AI competitive pricing intelligence for Australian retailers
The problem
Retail buyers and merchandisers make real pricing calls every week — on hero SKUs, on category resets, on sale events. The bigger retailers run this as a standing function with a BI team behind it. SMB and mid-market retailers make the same decisions on gut feel, not because their people are worse but because the analyst headcount isn't there and the tech team is heads-down on something else. The existing tools match on keywords and lie about confidence.
What I built
A self-serve SaaS that turns a product image into a defensible pricing recommendation in 90 seconds. Upload a SKU, ProductMatch finds visually similar products at Australian competitors using AI visual reasoning, fetches live prices, and produces a floor / sweet-spot / ceiling recommendation with the reasoning behind each number. Multi-tenant, multi-user, credit-based subscriptions, fully self-serve from signup to first scan.
How it works
- Image-first input — no SKU spreadsheets, no integrations, no analyst on retainer
- Multi-model AI pipeline: Claude Haiku for triage, Claude Sonnet for visual reasoning and similarity scoring, with extended-thinking recommendations
- Live AU competitor coverage — Temple & Webster, Freedom, Amart, West Elm, Castlery, plus 40+ more, with no cached prices older than 24 hours
- Built-in trust mechanics — similarity scores 0–100, human-in-the-loop corrections, full reasoning attached to every recommendation
- Multi-tenant SaaS from day one: Clerk Organisations, Supabase RLS, atomic server-enforced credit ledger, Stripe-hosted Checkout and Billing Portal
The outcome
Live at productmatch.com.au since May 2026. v1 shipped in two weeks — full commercial loop including AI pipeline, multi-tenant auth, and Stripe billing — then iterated through polish and v2 using Claude Code, Claude Design, and a Linear-based planning discipline that treated every architectural decision as a documented artefact. The build itself became a case study in what AI-assisted product development looks like when the human stays in the architect's seat — and the working method now informs how I help clients build their own AI products.
Want this for your own product? Build with me
Digital Starter Kit
Properly built small-business websites — live in days, fully managed
The problem
Small businesses get two bad options: DIY builders that quietly eat their weekends, or agencies with heart-stopping quotes and the clock running on every email. The slow, expensive part of a good website build is exactly the part that doesn't need a human.
What I built
A productised website business that takes a small business from brief to a live, properly built site in days, not the month-plus an agency quotes — then keeps it running. AI tooling I built in-house handles the slow, repetitive part of the build, so human time goes where it matters: understanding the business, getting the design right, and getting them found. One flat monthly fee, everything in the client's name.
How it works
- From brief to a live site in days — the AI-assisted build pipeline removes the weeks of agency turnaround, not the craft
- Fully managed after launch: hosting, security, backups, and the content and design updates — the client never logs into a CMS or chases a developer
- Tiered plans from a single-page presence site to a managed Shopify store
- One flat monthly fee — a service the industry still sells by the billable hour, productised
The outcome
Agency-grade sites live in days and looked after for a flat monthly fee — hands-off for the owner, updates included — only possible because AI carries the build. Currently in pilot, taking on founding customers.
MyWealthTracker
AI-powered digital financial advisor
The problem
Australians navigating financial decisions face a fragmented landscape of tools and advice. Most digital finance tools are either too simplistic to be useful or too complex to be accessible. Professional financial advice is expensive and hard to access at scale.
What I built
A full-stack AI financial advisor built on current Australian tax law, superannuation rules, and financial planning frameworks. It provides personalised, regulation-aware guidance that adapts to each user's situation — not generic calculator outputs.
How it works
- AI reasoning engine trained on current Australian financial legislation
- Personalised advice generation across super, tax, investment, and insurance
- White-label platform enabling mortgage brokers and financial planners to offer AI-powered guidance under their own brand
- Lead generation engine that qualifies prospects before they reach a human advisor
The outcome
A live product serving real users, with white-label capability that turns financial advisors and mortgage brokers into AI-enabled practices — generating qualified leads while their clients get immediate, accurate guidance.
Sally — Autonomous AI Practice Agent
AI infrastructure that runs a consulting practice
The problem
Solo consulting practices face a structural problem: the overhead of project management, context assembly, follow-up tracking, and client preparation consumes the same time regardless of firm size. What used to require a team of support staff either gets done slowly or doesn't get done at all.
What I built
An autonomous AI agent that operates as the operational backbone of my consulting practice. Sally lives in Telegram, manages project pipelines, scaffolds new engagements from a voice note, maintains persistent memory across every client interaction, and surfaces priorities before the first coffee. She doesn't wait to be asked — she operates continuously.
How it works
- Telegram-native interface accessible from anywhere — phone, laptop, watch
- Persistent memory that compounds across every engagement, conversation, and decision
- Voice-to-structure pipeline: a single voice note becomes structured notes, task breakdowns, and routed follow-ups
- Autonomous project scaffolding — "new project" creates a repo, board, and environment in seconds
The outcome
A solo consulting practice that operates like a product company. Clients get senior thinking with the responsiveness of a team — without the team. The gap between conversation and structured deliverable is hours, not days.
AI Capability Matrix & Governance Framework
For retailers evaluating and adopting AI responsibly
The problem
Most retailers know they should be doing something with AI. Very few have a structured way to evaluate where it creates genuine value versus where it introduces risk. The gap between vendor promises and operational reality is wide.
What I built
A structured capability matrix that maps AI use cases against business readiness, data maturity, and governance requirements. Paired with a governance process that gives leadership teams confidence to move forward without exposing the organisation to regulatory or operational risk.
How it works
- AI use case scoring framework mapped to retail operating model
- Data readiness assessment across customer, product, and operational domains
- Governance process covering model risk, bias monitoring, and compliance
- Board-ready reporting templates for AI investment decisions
The outcome
Gives retail leadership teams a clear, defensible path to AI adoption — knowing exactly where to invest, what to defer, and how to govern what they deploy.
AI Project Manager for Startups
Autonomous project governance across departments
The problem
Startups and scale-ups move fast, but fast without structure means projects stall, scope creeps, and teams lose alignment. Traditional project management tools track tasks but don't enforce discipline. Hiring a full-time PM for every workstream isn't viable at this stage.
What I built
An AI project manager that integrates directly with ClickUp and Slack to provide real-time project governance, status tracking, and intervention across multiple departments. It doesn't just track — it actively manages.
How it works
- Deep integration with ClickUp for task, milestone, and dependency management
- Slack-native interface for team updates, blockers, and escalations
- Cross-departmental visibility — engineering, product, design, and operations
- Governance engine that flags scope drift, missed dependencies, and resource conflicts before they become problems
The outcome
Teams ship faster with less overhead. The AI PM enforces the discipline of a seasoned project manager across every workstream, without adding headcount — purpose-built for the pace and constraints of startup and scale-up environments.
Why this matters
Advisors who build give better advice
When I advise a retailer on AI readiness, I'm drawing on the experience of having built and shipped AI products myself — not just evaluated other people's. When I assess whether an AI vendor's claims are credible, I know because I've built similar systems from scratch.
This hands-on capability is what separates practical advice from theoretical frameworks. Every project here involved real users, real constraints, and real decisions about where AI creates value and where it doesn't.
Building AI capability or evaluating where to start?
Whether you need a governance framework, a hands-on AI build, or just a clear-eyed assessment of what's possible — I've done all three. Let's talk about your situation.