I'm Not an AI Company, Entry 02: What I've Actually Built — and What It Taught Me
By André McKenzie, Founder, Silverhouse HD / Silverhouse AI · · Technology · 10 min read
Entry 01 was the why. This is the what. A working list of the AI systems I've designed, shipped, and learned from over three years — the avatar pipelines, the automations, the agents, the software — without the brand-name tool list, because the thinking is the part that actually transfers.
In Entry 01 I tried to explain the why — why a media company I've run for eighteen years quietly turned into an AI obsession I can't switch off. The most common reply I got was a fair one: okay, but what have you actually built? This is the answer. Consider it the receipts.
I'm writing it as a living document, because the whole point of doing this in public is that it keeps growing. Everything below is real — shipped, or built and learned from — and current as of today. I'll keep adding to it.
I'll describe what I built and what it does. I'm deliberately not naming the specific models, platforms, and apps I chained together to do it — and I want to be honest about why. Two reasons. First, that exact stack is the part of this that's genuinely mine; it took three years and a lot of failed afternoons to assemble, and it's the edge I bring to clients. Second, it would be out of date by the time you finished reading. The tools change every month. **What doesn't change is the thinking** — how you break a problem into a pipeline, what to automate, and where a human still has to sit. That part transfers, and that's the part worth sharing.
The tool is the easy part. Knowing what to point it at is the whole job.
I've produced more than fifty AI avatar videos — scripted, directed, and delivered in different brand voices, in both English and French, for different markets and audiences. The one I'm most proud of: a complete bilingual video pipeline for a Toronto and Montreal real estate developer — eleven long-format videos, eleven short vertical cuts, and nineteen carousel sets, across two buildings and two languages, each with its own on-brand presenter and a voice cloned and localized per market.
**What it taught me:** the hard part of AI video isn't generating a single clip. It's building a system that can produce forty consistent assets that all feel like one brand — then doing it twice, in two languages, without the wheels coming off. That's pipeline work, not prompt work.
I've produced sixty to seventy virtually staged real estate images — empty rooms turned into fully furnished spaces, in multiple styles, from multiple angles, at a fraction of the cost and time of physical staging. I also built a relighting automation that runs on its own: a raw photo lands in a folder the moment it's shot, gets sent through a generative model with the right instructions, and comes back transformed — no clicks in between.
**What it taught me:** the magic isn't the one good image. It's removing the human from the boring middle steps so the good image happens by default, every time, while I'm still on set.
I've put a live, branded AI avatar on a website — loaded with everything it needs to know about the business, holding real-time conversations with visitors about services and pricing. And I've built an inbound phone agent: a real number you can call and speak to, answered by a cloned, on-brand voice that can actually represent the company on demand.
**What it taught me:** the technology to have a machine answer the phone in your own voice already exists and works. The interesting questions are now about judgment — when it should answer, what it should never say, and where it hands off to a person.
A lot of my AI work isn't visible at all. A job lands on my calendar with the right marker and a draft invoice writes itself, priced correctly, ready for me to approve — billing that used to eat an afternoon a week now takes minutes. Content I make gets published across roughly ten platforms at once. There's research-and-outreach infrastructure that finds the right businesses to talk to and runs the first lap of the work before I'm ever involved.
**What it taught me:** the highest-return AI in a small business is almost never the flashy creative stuff. It's the silent plumbing — the handoffs between your calendar, your email, your payments, and your publishing — that you never realized was costing you hours every week.
I've built around ten websites and three full SaaS platforms — including Carousel Studio Pro, where you type a single sentence and it generates a complete, branded social carousel set, with paid plans and real user accounts. All of it self-built, with no outside engineering team.
**What it taught me:** the line between a person with an idea and a person who shipped the thing has basically dissolved. I'm a creative director, not a software engineer — and I've put working, paid software into the world. That still amazes me, and it's the single biggest thing I'd want another business owner to understand.
If Entry 01 answered why and the list above answers what, this is the part people are really asking about: what's in my head now that wasn't there three years ago. Honestly, it comes down to a handful of things.
None of that is a certificate. It's reps — three years of building, breaking, and rebuilding real things for a real business with real clients.
Here's what I think separates this from most AI-guy stories: I'm not coming at it from software. I'm coming at it from eighteen years and more than five thousand delivered media projects — for partners like Sotheby's, Chestnut Park, Minto, Westbank, and Ferris Rafauli — and an audience I built to forty-nine thousand subscribers and over twelve million views. The creative judgment came first. The AI is what I bolted onto a craft I already knew cold. That order matters — it's the difference between someone who can generate output and someone who knows whether the output is any good.
This is Silverhouse AI — the thing that grew out of the media company. For businesses that want help finding where to start, I consult. For businesses that want their own visual and video AI pipelines, I build them. For the ones that just want the output without thinking about any of it, I deliver. If you've read this far, you already know I'd rather show you the work than talk about it.
More entries are coming. This is me, learning in public — and there's a lot left to build.