Why I’m Building a One-Person, Billion-Dollar Agency

I’m building a $1B agency with a headcount of N=1.

I know, I know - my cringe detector is going off too. If I saw that sentence on my timeline, my immediate reflex would be to mute the account and launch its author into the sun.

However.

Obviously, the $1B valuation is a complete hallucination. I didn’t open a Jupyter notebook, run a Monte Carlo simulation on my SaaS margins, and converge on a clean ten-figure exit. I picked $1B purely as an attention-grabbing hyperparameter - it’s a high-salience token designed to spike your attention so you actually read this blog post.

But.

The structural bet underneath the clickbait is 100% real.

For the last half-century, the agency industry has scaled by a simple, brute-force law: complex problems require massive clusters of humans. If a Fortune 500 company wants a global campaign, they wire millions to a legacy giant like WPP or Ogilvy. But if you actually profile where any of that human-compute goes, what are they paying for? They’re paying for 19 layers of lossy middle-management routing, three weeks of synchronization calls, four junior copywriters OOM-crashing in the bathroom, and a $400,000 invoice to update the typography on a slide deck.

Most of what the corporate world calls “scale” is really just unoptimized human I/O friction.

For the first time in history, we can prune nearly 100% of that graph. The raw generation phase - brainstorming candidate concepts, drafting copy, running A/B permutations, and compiling final deliverables - used to require an army of caffeinated 20-somethings grinding away.

Today, the marginal cost of that token generation has collapsed to roughly zero.

An LLM doesn’t have an internal world model for what’s actually cool. It doesn’t have a reward function that penalizes an idea for feeling slightly cringe, or trying too hard, or lacking the emotional resonance that makes a human actually care about a product. Base models and RLHF assistants are basically a firehose of high-probability, median-quality slop. To get 99th-percentile output, you need a human in the loop acting as a ruthless discriminator.

My thesis is pretty straightforward: a single high-taste human orchestrating a fleet of a few dozen specialized AI agents can now execute the exact workloads that legacy agencies charge millions to slow-walk. Zero office politics. Zero synchronization overhead where you have a status meeting to prep for the next status meeting.

Just pure, unthrottled leverage.

That’s the goal for SELF.

What clients pay for

A big agency wins your business with ideas. Once you sign, it bills you for coordination.

A strategist writes a brief and assigns it to a creative team. The creative team writes a few concepts and assigns them to an account manager. The account manager presents them to you, collects your notes, and carries them back down the chain. A producer books the shoot; a project manager chases the producer; someone else spends the week moving files between folders. Each person who takes the work from the last one loses a little of what you meant. The team holds a Big Meeting™️ at every handoff, and the agency bills you for each one.

If you cut the chain down, you need two kinds of people: someone who understands your business problem, and someone who can find the idea that solves it. Clients paid for everyone else because, until recently, nobody could turn an idea into finished work without a team of copywriters, designers, producers and planners. Agency owners also had good reason to keep those teams large. They billed by the seat and the hour, so every extra person on your account raised the invoice, whether the work got better or not.

I can now hand the drafting, formatting, resizing, media planning, and file wrangling to AI agents. I can do in an afternoon what that chain did in a fortnight, with no meetings.

That leaves me three jobs:

1. Working out what the client needs,
2. Finding an idea nobody has had, and
3. Judging what comes back.

A holding company can’t get better at that by hiring more managers.

The giants are shrinking.

WPP reported a 4.7% fall in like-for-like revenue less pass-through costs for the first half of 2026. Its leaders have cut about 11,000 jobs since early 2025 and plan up to 1,000 more by the end of the year. They have also put Ogilvy, VML and AKQA under one umbrella called WPP Creative.

Omnicom’s leaders bought IPG, then retired three of the oldest advertising names: DDB, FCB and MullenLowe. They cut 4,000 jobs in the first wave and expect to cut about 15,000 roles in all.

Meanwhile, Meta’s leaders have said they want brands to upload a product photo, set a budget, and let Meta’s AI write the copy, make the images and video, and pick the audience, all by the end of 2026. Meta used to sell agencies their ad space. Now it wants to make the ads too.

Why they can’t fix it

I don’t think the people running these companies are stupid.

I’m not that hubristic.

As hubristic as I am…

They can see what AI will do to their business as well as I can. They can’t fix it from the inside.

A holding company makes money by selling people’s time. When its staff uses AI to do in an afternoon what five people used to do in a week, the company loses billable hours. Its leaders can build the best AI platform in the industry and still lose, because they can’t bill for any hour their staff saves with it.

They can pass the savings to clients and watch revenue shrink. They can also keep charging the old fees and hope clients don’t notice. Procurement teams at big brands know what the tools cost so that the second plan won’t last long.

Inside a holding company, executives measure their power in headcount. A plan that cuts an account team from forty people to four asks those executives to vote against their own careers. Very few of them will.

They also pay for offices in expensive cities, layers of management, and pensions. WPP alone still pays about 97,000 salaries. I pay for none of that.

What I’m building

I’m exceptionally good at narrative, messaging, positioning, editing, and speaking to a room. I can sit with a founder or CEO, work out what the company believes, and turn that into a story people repeat.

I do the thinking, the judging, and the client relationship myself. I use agents for production.

Here’s how I run a launch:

1. I spend a few hours with the founder and the people who sell the product.
2. I write the core story: what the company believes, who it serves, and what it wants buyers to repeat.
3. Then I brief the agents. One set pulls competitor messaging, customer reviews, and sales call notes. A second set drafts the site copy, the sales deck, the launch post, and forty social posts. A third builds a media plan and a reporting template.
4. I read everything, cut most of it, and rewrite the parts that carry the story.
5. The client sees one draft from me, not five rounds from a committee.

A network would staff the same job with a strategist, two copywriters, a designer, a producer, a media planner, and an account lead. They would hold most of their meetings to align before the client saw anything; the client would pay for every single one.

I run the business on five rules:

1. I sell outcomes, not hours. A client buys a repositioning, a launch story, a crisis response, or a year of messaging, at a price they know up front. If I find a faster way to deliver it, I keep the margin.
2. I think; agents do. I never hand a client anything I haven’t read and approved.
3. I take fewer clients and charge each one more. I grow by making each engagement worth more, not by adding accounts.
4. I don’t hire. When the work outgrows me, I build better agents, or I say no.
5. I don’t take a meeting a written update could replace. My clients get faster answers; I get my mornings back.

Holding companies can’t adopt the first rule without cutting their own revenue, or the fourth without firing their own managers.

Where this could go wrong

Marketing chiefs at big brands hire big agencies because it feels safe. A CMO who hires a network and fails can blame the network. One who hires a solo operator and fails takes the blame personally. I can’t remove that fear. I can shrink it with a track record, references, and work clients can see before they sign.

I still catch mistakes in agent output every day. Clients don’t pay me to ship those mistakes; they pay me to catch them. If I stop reading closely, I become a worse version of the agencies I compete with.

I can only hold so many client relationships myself, and I can’t hand my judgment to an agent. Some clients will want a room full of humans. I’ll send them elsewhere.

I could also be wrong about the timing. The holding companies have money, data, and relationships, and WPP’s leaders say they’ll return the company to growth in 2027. Maybe they will. If they do, I’ll still have built a high-margin business with no staff, doing work I enjoy. I can live with that.

Why now

Holding company leaders move slowly while they restructure. Every month they spend merging units and retiring brands, their clients compare the bills with what the tools cost and start asking questions.

I can already replace most production work with today’s tools. Most clients haven’t yet decided who to hire instead. Senior strategists and creatives keep leaving the networks to start small shops. In two or three years, I expect clients to have made up their minds: they’ll know which small operators they trust, and those operators will have case studies to prove it. I want to earn a place among them before clients finish choosing.

If you’re a strategist or creative director inside a network, write down the parts of your job a client would pay you for directly. If the list is long, you may not need the network anymore.

If you run marketing at a company that pays a network retainer, count the people on your account you’ve actually met. Then ask what you’d pay for just those people.

As for the billion dollars, I’ll report back…