I'm the strategist who builds and the builder who sells. I build new revenue lines inside companies that already work.
Most companies don't have an idea problem. They have a build problem.
The consultancy ships a deck. The dev team ships tickets. Neither ships a business. The deck costs a fortune and never gets looked at again. Something always feels missing, something never quite right. The dev team builds what they're told and asks who's specifying it. Nine months later there's no new revenue line, just an invoice and a Jira board.
I'm the person in the gap. Think like a founder, build like an engineer, sell like an operator. Fifteen years of running operations, scaling businesses, and shipping products end-to-end. That's the job.
Dozens shipped. The work below is a sample. Some I built solo from nothing. Some I built inside operating businesses with their teams. All of it shipped.
Four clusters where one pipeline feeds several products. The rest are standalones. Some live, some mid-build, all real. Read in any order.
Two products, one database. Small Flat White collects the price. Espresso of Interest logs the taste.
AI voice agent calls Australian cafés during off-peak hours. Asks the price of a small flat white. Rotates four scripts so no two calls sound the same. Transcribes, extracts price, size, roaster, alt-milk surcharge, and a confidence score. Hangs up. Calls the next one.
The biggest flat white index in Australia.
Read how it works →SmallFlatWhite knows what a coffee costs. EspressoOfInterest is where you log how it tasted. It ships with a pre-built catalogue of Australian roasters and beans, lets you record extraction parameters for every shot you pull, and visualises the point at which your recipe converges on a good one. Same data model as SmallFlatWhite. Same Oscar-collected roaster data flowing in underneath.
One pipeline. Five front-ends.
F1 produces more data than any one site can show well. So the pipeline ingests all of it once, then five different sites surface it for five different audiences. Results for casual fans. Data for the stat obsessives. Info for insiders. Insights for the strategy nerds. Expo for the event side of the sport.
One Turborepo monorepo. One Hono API. One cron worker pulling from F1DB. One Postgres database. Five independent Next.js front-ends on five different domains, each optimised for a different SEO surface — Results for casual fans, Data for stat obsessives, Info for insiders, Insights for strategy nerds, Expo for the event side of the sport.
The interesting bit isn't any one of the sites. It's that the marginal cost of the next site, after you've built the first one, is about two days.
A Shopify storefront dropshipping die-cast F1 models into the F1 Network's traffic. Supplier-driven inventory, daily CSV feed, prices marked up on a strict formula (supplier cost + $1, doubled, rounded up to the nearest X9). Manually running this store is a full-time job. I didn't want a full-time job. So I built the next thing.
The operations brain behind the die-cast store. It ingests the supplier's CSV daily, detects every price, stock, and product change, runs pricing rules deterministically, proposes AI-generated product descriptions and images for approval, and pushes approved changes to Shopify through a platform-agnostic adapter — so tomorrow the same system can run WooCommerce or anything else.
Three-layer decision model: deterministic (rules engine), agent (AI proposes), human (clicks approve). No money moves without the last step.
Three directory sites, one shared ABS Census pipeline. Built the data foundation once, plugged in three different audiences.
Every Australian suburb, scored 0–100 across six dimensions — Safety, Affordability, Lifestyle, Family, Connectivity, Growth — using only government open data and a published methodology page. Compare any two suburbs side by side. Radar charts. No marketing fluff layered on top of the numbers. Currently live for Queensland; state adapters queued for NSW, Victoria, SA, WA.
Same ABS pipeline, schools on top. A directory of primary, secondary, and combined schools indexed by suburb, enriched with state education-department data. The marginal cost of this property was a weekend — the pipeline was already there from PostcodeIQ.
Same ABS pipeline, childcare on top. Every ACECQA-registered centre in Australia, indexed by suburb, enriched with local under-five population data. Third leg of the same stack. Schools, suburbs, childcare — same families, same data, same pipeline, three different sites.
Oscar is the name I give to the AI agents I run. One identity across several different jobs.
Oscar started as the voice on SmallFlatWhite, calling cafés. He's now four things — a voice agent (calls, transcribes, extracts), a meeting agent (joins, notes, follows up), a coding agent (operates inside my own repos), and a control centre that orchestrates the other three.
Most of this is for me, not for sale. But Oscar is how I move faster than a one-person operation should — and when a client's engagement needs a particular agent pointed at a particular problem, Oscar is the shape of thing I point at it.
Each of these is its own thing. No cluster. Some earn a deeper read.
SerpOS is my internal "Semrush, but custom-built for my sites." A nightly pipeline that measures every LogoLess property against Google, runs Lighthouse audits, pulls Search Console data, runs keyword research through DataForSEO, hands the results to Claude, and generates concrete action items. The recommendations live in a REST API that every site in the portfolio polls in the morning.
It was supposed to save me the cost of a fifth Semrush seat. It ended up quietly running SEO across six different sites simultaneously. Currently the single most valuable piece of infrastructure I've built.
Amazon's famous 6–12 graph format for Weekly Business Reviews, productised. Plug in GA4, Search Console, and Stripe. Pick your six to twelve metrics. Every Monday morning, SixTwelve generates the charts and a Claude-written narrative underneath explaining what moved, what didn't, and what's worth paying attention to. Currently being dogfooded across my own portfolio before the public SaaS opens up.
For operators who run weekly reviews and are tired of assembling the deck by hand every Sunday night.
A phone repair directory for Australia, built programmatically across the full state × suburb × device × repair-type matrix. Shop data from Google Maps via Apify. Prices scraped from each shop's own website and extracted into structured rows by Claude. The URL architecture — state, suburb, device, brand, model, repair type — is designed to saturate the long tail of "where do I get my Galaxy S23 screen replaced in Bundaberg."
Phase 1 live. Phases 2 and 3 (lead-gen quote comparison, then marketplace) queued.
The Foundry is the most ambitious thing on this page. It's a self-operating business that manufactures other AI-powered businesses.
Here's the idea. Most small business software is over-generalised. What people actually want is a preconfigured company for their exact niche — the agents, the SOPs, the prompts, the rule-engines — already assembled and ready to run. The Foundry's pipeline identifies niches with a proactive scanner, runs them through a validation gate (minimum 7/10 score), designs the agent roster, writes the skills, assembles the pack, and publishes it for sale. Seven internal agents run the production line. Editorial review at the end is me.
If it works, I'll be selling downloadable AI companies at volume. If it doesn't, I'll have learned something useful about whether this category can exist.
I also run That Automation Agency — an APAC Zapier Platinum Solution Partner, 450+ clients, 10k-strong community. A few years in, the obvious problem was that the delivery team was doing a lot of high-volume transactional work that didn't compound. Audits, one-off builds, project work. Valuable to the client, but not the kind of engagement that turns into a retainer.
So I built Zap Suite for it. Four linked products on one common core. ZapAudit semi-automatically audits a client's Zapier account against best-practice checks and generates a findings report — the starting point for every conversion conversation. ZapCatalog is a live database of every Zapier app, trigger, action, and field, with an AI flow designer that composes workflows visually and then builds them. ZapDetect is a Chrome extension that tells the sales team whether the website they're looking at supports Zapier. ZapLeads is the upstream view — a live pipeline and opportunities report that keeps the Zapier partnership team in the loop on deals the agency is working, so the partner relationship compounds instead of drifting.
The whole suite is sales-and-delivery infrastructure for a services business that had already maxed out what transactional work could earn. Built once, used daily. A real example of the thing this page is otherwise trying to describe in the abstract.
A multi-tenant agency platform for AI-generated Instagram content at scale. Agencies train custom LoRA models, generate images in a Studio interface, rate them through a feedback loop that injects negative prompts into the next generation, and publish through the Late API. Four-role permission system, 26-table database, 203 unit tests, shipping as v1.12.x.
Originally built for one agency client. The product pattern turned out to be general.
Every one of those started with nothing. No distribution, no customers, no existing revenue to lean on. If you're reading this with a business that already has those things, your starting position is better than any of mine. That's the interesting conversation.
Think like a founder, build like an engineer, sell like an operator. Fifteen years of doing all three at once.
Spent a decade as Operating Partner at an investment group, where I helped scale an education business from one company to nine. Led the acquisitions, the integrations, and stepped in as Interim GM across new businesses to professionalise operations and run turnarounds. The job was figuring out why things didn't work and fixing them. It still is.
Founded That Automation Agency in 2014. First Zapier Certified Expert in APAC. Trained hundreds of professionals. Built thousands of workflows. Saved clients millions of minutes.
Brisbane base, specialty coffee habit, garage of F1 diecast, long list of projects probably about to start.
The services arm. Automation across Zapier, Make, n8n, plus process improvements and fractional ops. The work you'd hire a COO for. We connect systems and streamline processes for hundreds of clients.
The product arm. New revenue lines built end-to-end inside businesses that already work. Solo or with your team, scoped engagements, your IP at the end.
If you need automation, that's That Automation Agency. If you need a thing built that isn't getting built, you're on the right page.
I take on a few of these a year. The fit is usually a business with real revenue, real customers, and a thing they want built that isn't getting built. Scoped engagements, your IP, your name on it.
If that's you, send a note. Tell me what you're trying to build and what's in the way.
Or reach out elsewhere: LinkedIn, That Automation Agency, or [email protected].