Maintaining $340k/yr revenue while halving agency workload and headcount

Tom Garcy grew Garcy Studio to $340k/yr and kept it there while cutting back on headcount and workload by leveraging AI in his processes. And while he was busy optimizing this business, his other business got acquired.

Here's Tom on how he did it. 👇

Using AI as leverage

I'm Tom Garcy, and I've been running Garcy Studio for the past seven years. During that time, I launched and sold my startup, Kyugo, which focused on time management. A private investor acquired it last year.

I previously ran venture studios as a creative director and head of design, and I know the advertising agency business quite well. I founded my own studio to do things a bit differently, with a smaller team and more boutique projects. We quickly focused on branding, digital applications, and websites for digital identity.

We're at around $340k/year. This is with a core team of just three people, plus some other specialists hired per project. We've been at this revenue for two years, while we scaled down our team and workload by 50%. As a result, we now earn almost double or triple the money per project hour. AI makes our processes much faster and easier to manage and automate, especially back-office tasks, client management, task management, and general project management. This allows us to focus more on the work itself.

How to hire right

If I did it again, I would avoid hiring many people simultaneously, as I did early on. We lacked the capacity to align with them. Managing a team is very hard when it grows too fast.

And I would always hire one A+ person over three B-level people, even if three B-level people seem cheaper overall. The A+ person will save a huge amount of money in the long term by nailing everything on the first try, requiring little management and working almost autonomously.

It's never worth saving money on people; always hire someone you believe is at the absolute top level, even if they seem very expensive, because this will save a shit ton of money in the long term.

Dogfooding a new product

We are currently gathering all our AI processes and integrating them with the latest AI advancements (AI agents and subagents). And we're turning that into an easy-to-access product to manage all your client knowledge base. This knowledge base serves as the agent's memory, allowing you to reach out to your client at any time — the agent's knowledge will mimic the client itself.

The system can automatically send proven questions and questionnaires for the client to answer; it will then automatically save their responses to the client's memory. Any team member can access the same client brain to solve tasks and find answers. The agent will also create a unified brief based on our proven experience, which each team member will understand.

We're building based on what we've learned, and we're dogfooding it every step of the way. Eventually, we plan to make the tool available to others.

Recurring payments

Our studio clients prefer a monthly fee, which helps both clients and us know what to expect each month. So, we start each project with a project fee and then transition to a monthly fee for a few months, ideally for longer-term projects. That means our revenue is always growing because we don't just hand over projects; we offer assistance in implementing and improving them. It's a better experience on both sides.

We plan a similar model for our new product. We'll play with:

• lifetime

• one-time payment

• monthly payments

• quarterly payments

In many projects, especially with intensive branding, weekly payments even make sense. Our product allows weekly use because users typically need only a week or two to create something, and then they can pause. We can always resume a week-long sprint and return later.

A comprehensive stack

Here's our stack.

Frontend:

• Next.js 14 (React 18, TypeScript)

• Tailwind CSS (styling; we use vanilla CSS for zero dependencies)

• Supabase Auth (email/password authentication)

Backend:

• Supabase (PostgreSQL database + auth)

• Next.js API Routes (serverless functions)

• Supabase RLS (Row-Level Security for data privacy)

Database:

• PostgreSQL (via Supabase)

• 11 tables: Users, Clients, Knowledge Base, Briefs, Agents, Chat, Locked Styles, Agent Learnings, Team Members, Permissions, Questionnaire Responses

• Indexes for performance

• RLS policies for security

Payment:

• Stripe (restricted API key for checkout)

Agent Integration:

• OpenClaw agents (spawned as isolated instances per client). We use our own Openclaw in our dedicated AI room in our studio, which uses Anthropic models (Claude Code, Claude Sonnet, Claude Haiku, and Claude Opus).

• Agent memory: JSON snapshot of client KB

• Cross-client learnings: Aggregated insights table

Deployment:

• Vercel (frontend + serverless functions)

• Supabase (managed database + auth)

Development:

• Node.js + npm

• TypeScript for type safety

• Turbo monorepo (apps/survey + apps/dashboard + shared/types)

Tools:

• Git for version control

• GitHub for repo hosting

• SQL for database schema

Word of mouth

99% of our business is word of mouth. We always build the best possible client relationships, and our clients are our best references.

Other than that, we publish our work on LinkedIn sporadically. We send a couple of newsletters with our projects.

The tech-first, design-later era is done

This is a great time to start, given the rapid advancements in AI. Being in the middle of your career can be scary if you started with an old tech stack and now must completely change. But if you're just starting out, I would focus on learning as much as possible about agentic AI and how to build and ship products independently.

Also, explore ways to improve design and UX, because standard AI only achieves a standard level in these areas. Explore how to make beautiful and functional products, rather than just shipping them. The era of tech-first, design-later is over; people now care about how things look and work. Beautifully designed products always win, even before users know how they could benefit from them. Conversely, a great idea with excellent tech, but poor design/UX will never succeed.

Embrace change

And one more piece of advice: Embrace new things, especially with AI.

In the design and creative industries, for example, many people, even big names, still say that if AI disappeared tomorrow, they wouldn't care and their work wouldn't change. This is scary. The way forward is to embrace AI. Don't fear it. Use it as a second set of hands.

What's next?

From here, our goal is to make our processes much faster and produce in a day what usually takes two weeks.

To do that, we're building an internal product that we hope to make widely available soon.

I'm starting to publish more on my personal LinkedIn, so you can connect with me there. And here's the Garcy Studio website and Instagram.