Bruce Marjoribanks
Built an agency. Sold it. Spent 27 years inside the machine. Now I replace it.
The meeting that ended it for me.
I was on a Zoom with five other people from the agency. SEO. Social. Facebook Ads. Google Ads. A strategist. An account manager. And the client — a national roofing company – sitting through the whole thing.
Six people. Two hundred bucks an hour each. Twelve hundred dollars to be told “this went up 5%, this went down 5%” by five people who’d never spoken to each other before the call started. Each one got their ten minutes. Each one read out generic numbers. Each one then went back to running a generic playbook under the title specialist.
And the strategist? That’s the title agencies hand to their sales team. Their job isn’t to build you a strategy. Their job is to align whatever’s in the agency’s product menu with whatever you’re willing to pay for. That’s not strategy. That’s a quote.
I sat there and realised something I couldn’t unsee: I could do every single one of their jobs. Better. Five times faster. Using AI tools none of them had touched.
That was the moment the agency model died for me.
How I got here.
I started in 1998. Dial-up modems. Internet Explorer 2. Netscape Communicator. If you wanted a website doing real work, you had to learn how to code it, how to design it, how to run the server it sat on, and how to market the thing. There was no SEO specialist. No paid media strategist. There was you, the project, and whatever you could teach yourself.
Twenty-seven years later, I still work that way.
In 2008 I started Digital Organics – team of twelve, websites and SEO when paid traffic was still a sideshow. I ran it for four years and sold it when I moved countries. Back then, every client meeting was face to face.
Then came Head of Search and Performance Media at First Digital – one of the most disciplined data-driven agencies in New Zealand, and a genuinely good company. Real attribution work. Real conversion work. Big clients: Harmoney, Contact Energy, House of Travel, Dymocks, The Co-operative Bank. And even there — at one of the best shops in the country — the analytics team and the paid team barely shared data on the same client. All technicians, heads-down on their own numbers. None of them gave a shit about what makes an ad actually work.
That’s the part of the agency model nobody outside it sees. The walls aren’t between disciplines. They’re between people who treat advertising as a craft and people who treat it as a queue.
I also did six months as Head of Advertising at another agency I won’t name. I quit when I dug into a 12-month-old client’s ad accounts and found them empty. Nothing built. Nothing run. $2,400 a month, twelve months running, not a single thing done. When I raised it with management, they did nothing. A year after I walked, the agency was gone.
Across all of it, I’ve personally managed $47M+ in ad spend across finance, retail, travel, energy, and home services – both sides of the Tasman.
The Google moment.
In 2018, Google published an Insights case study on machine learning in Search. The case study was Harmoney. The campaign was mine — one of the first Target ROAS lead-generation builds in New Zealand. We doubled qualified loan applications year-on-year at a 37% lower cost-per-acquisition.
Here’s the part that didn’t make it into Google’s write-up. I was sitting in a Google Ads team of ten people. Not one of them asked how I’d done it. They went back to their own accounts and did what they’d always done.
That’s not a people problem. That’s what the agency model does to the people inside it.
A word on the Internet Marketing Bros.
"Grow Your Practice With 100 New Patients — Or It's Free AND We Pay You."
These agencies don’t get built by marketers. They get built by people who took a six-week course on how to sell marketing. The pitch is always the same: a generic lead-gen package dressed up in the language of your industry. “Specialised Lead Gen For Dentists.” “For Plumbers.” “For Chiros.” Same playbook every time. Different vertical on the homepage.
The sales pitch is polished. The delivery is panicked. The actual work gets outsourced to a freelancer in the Philippines who’s been given a Loom video and a hope. Average client retention: three months. Check the website 18 months later and half the time you’ll hit a 404.
If your gut has ever told you the polished offer was a bit too clean, the guarantee a bit too good, the testimonial page a bit too templated – your gut was right.
What I believe - and most agencies won't say out loud.
The agency model isn’t broken because the people are bad. It’s broken because fragmentation is the product. Five specialists. Five invoices. Five dashboards. Five sets of incentives. And nobody accountable for whether your business actually grew.
The model was designed for a world where one person genuinely couldn’t run paid, content, email, creative, and analytics at the same time. That world is gone.
I believe one strategist with the right systems beats five specialists who can’t see each other’s screens – every time. Not because the specialists are wrong. Because the structure is.
The AI edge - and why most "AI consultants" are bluffing.
Eighty-one percent of IT professionals think they have the AI skills their job needs. When researchers actually tested them, twelve percent did.
That gap is the entire game right now. Most agencies and consultants who claim “AI capability” added a ChatGPT subscription and called it transformation. Glass facade on a 1970s building. The org chart didn’t change. The billing model didn’t change. The six-week production timeline became a five-week production timeline. The tools sit on top of the old structure – cosmetic, fragile, fooling nobody who looks closely.
That’s not what I do.
For three years I’ve been rebuilding the marketing workflow from the foundation up. Every client gets a knowledge base that compounds – brand voice, audience research, past performance, creative guidelines – so the AI doesn’t start from zero on every brief. Generic AI output needs 80% human rewriting. Knowledge-base-backed output needs 20% refinement. The difference isn’t the model. It’s the architecture underneath it.
Concretely, that means:
- Video ads produced in hours, not weeks – Seedance 2.0 generation, professional editing, ElevenLabs voiceover, all in one pipeline. Not “AI → upload.” AI → craft layer.
- Creative variants generated and tested at volumes a freelance designer can’t match — Midjourney through Nano Banana through Photoshop, not Midjourney through publish.
- Customer research that used to take three months and $75K, now done in days — and faster and deeper, because the AI can read every review, every transcript, every competitor page in parallel.
- Campaign analysis and reporting automated end-to-end, so my time goes to decisions, not screenshots.
The test for whether someone is actually AI-native: if they cancelled every AI tool tomorrow, would their pricing model, delivery timeline, or org chart need to change? If the answer is no, AI was never structural. It was decorative.
Most consultants are still figuring out how to prompt. I’ve built the architecture that makes the prompt almost incidental. That’s why one operator now does what used to take seven.
27 Years
Industry experience (AU/NZ)
Since 1998
$47M+
Ad spend personally managed
Google Insights
Featured case study, 2018
Harmoney
50 → 4,500
Leads per month
RoofBuddy
Built & Sold
Digital Organics
2008–2012
Strategic Marketing
Diploma, NZMA
How I Work
- Strategy first. Always. No dollar of ad spend moves until we’ve agreed where it’s going and why.
- One point of contact. That’s me. No account managers. No juniors. No handoffs.
- Two meetings a month – that’s it. At my last agency I had weeks where I couldn’t find two hours between internal meetings to do actual work. That’s the model. Not here. Your money goes into the work, not the calendar.
- AI-integrated end to end. Not as a gimmick. As an unfair advantage.
Done reading? Good. Now let's look at where your money's actually going.
Fair warning:
Once you see the real cost of fragmentation, you can't unsee it.