Work & proof · for Melbourne Innovation Centre
For a Digital Business Advisor, the real question isn't whether I know the tools, it's whether I can deliver them for a small business without overwhelming the owner. So here's what I'd build for your clients, each one paired with the actual work that shows I've done it before.
For your clients
The "AI Starter" 1:1: a half-hour intake, pick one or two tools that fit the owner's actual workflow, a one-page action plan the same week, a 30-day check-back. The point is a working result, not a list of possibilities.
Proof
At Silvatron I led the change management that took a cross-functional team to 3 to 5 times their previous efficiency with AI workflows, by addressing fears one-on-one and reframing AI as the grunt work a person still supervises. Output went from roughly one feature a month to about five.
A "Safe AI" workshop: which tools meet basic privacy obligations, what should never be pasted into a hosted model, where Australian-hosted options sit, and the security baseline that has to come first. It bridges the two pillars owners currently meet separately.
Proof
My Monash IT major is Computer Network Security. I've built systems where data residency mattered (a Victorian Government agency engagement via Silvatron Pty Ltd, solved with AWS hosting in Melbourne), and my own platform is Sydney-hosted with secrets kept out of version control and opt-in redaction for sensitive sources.
A 60-minute teardown of the owner's site, ad account and follow-up, ending in three ranked actions to take this week. Specific to their business, not generic advice. Try the 2-minute version.
Proof
I drove a 20% lift in e-commerce revenue in-house at Fujian Footwear (Australia): Meta Ads, SEO, GA4 and Looker Studio across four brand sites, plus managing the outsourced agencies and keeping them honest. Real P&L impact, not slideware.
A voice agent that handles enquiries and takes a booking out of hours, in the business's own voice. The kind of thing I'd stand up with a client in a single session, then reskin to the next one.
Proof
I built two working voice agents for this application, each with a knowledge base and a booking step. Talk to them on the Session page. The same build reskins to a café, clinic, retailer or trades business by swapping the knowledge base.
A workflow that turns one good piece into several channel-ready ones without losing voice, with a human edit pass that's non-negotiable. This is where the "39% worry about AI" owners are won over: a gate, not autopilot.
Proof
I architected a governed AI content pipeline with configurable per-brand voice profiles, redaction gates and a human-in-the-loop approval step, built specifically to avoid generic "AI slop". On PolicyAi, every answer is cited and currency-flagged, with zero cross-role data leaks across a 90-day delivery.
A light intake-to-reporting pipeline (form, lightweight CRM, scheduling, automated comms, an outcome tag per session) so funder reporting is a saved view, not a quarterly spreadsheet day. Built on what's already in use, removing the manual seams.
Proof
I run this shape of pipeline in production: n8n with a CRM and automated comms. Sentinel, my AI code-review and DevOps bot, saves a team 20-plus hours a month. The judgement is mostly what not to add, so a human stays in the moments that matter.
Why I can say all that
Not a sandbox. A self-hosted production stack I operate end to end, in Australia.
Honest note: parts of the content pipeline are phased and still in build, and some of the platform is deliberately kept simple. I'd rather show you what's genuinely running than overstate it.
Selected live work