Laid off from eBay, she built 27 AI agents

27 AI – After a layoff from eBay in 2024, Linara Bozieva launched a marketing agency run by a three-layer AI workflow featuring 27 custom agents. She oversees strategy, tests the system across client profiles, and keeps key human judgment—especially when there isn’t e

When Linara Bozieva lost her job at eBay. the timing felt especially brutal: her family had moved from Switzerland to the US. and she was staring at a job market crowded with candidates and short on openings.. After 11 years at the company. she decided to build something small instead—at least for a couple of years—to support her household.

A few months after the layoff in 2024, she launched a traditional marketing agency with an unusual structure. Bozieva said she doesn’t have a marketing background, but she built a three-layer AI workflow with 27 custom AI agents that she runs under her oversight.

Her automation started as she watched AI tools become mainstream. When AI became mainstream, she used ChatGPT in her browser and tools like Midjourney for video work. She then gravitated toward closed-loop systems where agents can act autonomously, and she set out to build something similar.

The stack changed as she ran into practical constraints.. She initially built her system on Google’s Antigravity platform because it was user-friendly. but she switched to Claude Code after token limits with Gemini Pro.. She said she loves Claude, but she has started hitting token limits there as well.. Even so. she described her prior experience as useful in shaping the system: she worked in analytics at eBay. and that background helped when building the architecture and creating guidelines.

Bozieva also said AI wrote much of the system itself. With markdown files, skills, agent files, and scripts, she used plain-language instructions to tell the model what she wanted it to do, then tweaked as needed.

Her system’s 27 agents are organized into three layers: directives, orchestration, and execution.. In the directives layer, the setup defines who the agents are, what they know, and how they operate.. The orchestration layer chooses which agents act and when.. Bozieva said there are six agents in the orchestration layer that function as the “brain” and think before other agents act: a market researcher agent. a data analyst agent. a creative director agent. a finance agent. a legal agent. and an orchestrator agent that routes tasks to the execution layer.

Once tasks are routed, three agents build the technical groundwork. Ten agents work on driving traffic and awareness, and the last five agents convert traffic into revenue. The execution layer includes scripts designed to guide repeatable tasks.

Cost is also part of the picture. Bozieva said the setup costs her under $1,000 a month, covering Claude Code, Codex, and ChatGPT. She also uses specialized tools including HeyGen and ElevenLabs, and she pays for APIs because she gives Claude Code access to other models and tools through APIs.

One of her biggest challenges, she said, wasn’t building the system—it was knowing when it actually worked. The workflow has been trained on 14 clients in total. She currently has five clients, and she also tested the workflow on her own projects and did free work for friends to improve the agents.

She said she tested 14 client profiles to make sure the system understood her approach and frameworks and delivered high-quality results every time.. When she turns to ads. she described a workflow meant to reduce stiffness in the output: she has Gemini and Claude generate ad copy separately. then compares and combines the results to produce what feels more human.. She also added scripts in the execution layer that pull customer pain points from Reddit discussions to sharpen the strategy.

Still, the human element shows up most sharply in the day-to-day interactions. With her business, Bozieva said the AI agents run ads, analyze performance, improve creatives, and then give her daily reports. Clients receive weekly reports, and Bozieva and the client do broader monthly calls.

She said AI can process a transcript of a client call. but it cannot fully read the room or identify where a client seems most nervous.. That judgment and client management remain with her.. She’s also heavily involved in strategy when there isn’t enough incoming data. because her agents and she reach a crossroads about which path to choose.. She said she continually checks whether what the agents propose makes sense and is realistically doable as a strategy.

She drew a boundary around what she believes AI can handle without domain expertise.. Bozieva said AI can build a system. but you need to understand the domain; she gave an example about healthcare. saying she wouldn’t be able to build a healthcare agent that treats colds and flu in children because she doesn’t have that knowledge and wouldn’t know which guidelines to include and which agents she would need. or how to better update the system as she goes.

She also described her view of the future workload. Bozieva said all repetitive tasks can and will be done by AI in the future, but she believes the expertise needed for specific departments or domains will still sit with humans acting as VPs over AI architecture.

Her agency has split into two tracks. She said she still has a few traditional marketing clients that she serves with freelancers, and because she has been serving them before, she can’t say she isn’t doing that anymore. Those clients are part of a smaller, separate branch of her business.

For new clients, she said she now provides only full-cycle marketing through the AI system.. After onboarding, she spends about two hours a week overseeing their account.. She estimated that if she focused only on client oversight, she could manage 20 to 25 clients on her own.. Once she reaches that number. she said she would hire people like her—operators who oversee the agents and guide strategy.

Bozieva framed the work as personal as much as professional: building something on her own has been “incredibly fulfilling and promising,” and she said it changed not just how she works, but what she believes one person can build.

eBay layoff AI agents marketing agency tiny teams Claude Code ChatGPT Gemini Pro HeyGen ElevenLabs APIs ad copy