So, I'm Building a Team of AI Agents

Over the past few weeks, I've begun experimenting with building AI agents. It's still early — but some of what's happening is surprising enough that I wanted to start writing about it. So here we are. A blog. Updates, learnings, and honest notes as things unfold.

A little context on me, if we haven't met

I've spent most of my career at Google and Meta, on the business and operational side. For years, I sat across the wall from product and engineering teams — close enough to understand how things got built, far enough to see how much of what makes a company grow has nothing to do with the product itself.

After leaving Meta in 2023, I wanted to get closer to the builders. I created Elsinore Ventures as a way to do that — part venture advisory, part startup studio. For the last year, I've been embedded inside an AI-first venture studio, working alongside early-stage founders.

Distribution is Hard

That experience reinforced something I already knew in my heart: the thing most founders struggle with, even the best ones, is distribution.

Sometimes the challenge is the basics — simply defining what problem you're solving, for whom, how, and why it's better than the alternatives. That's what advisory work is built for: coaching and supporting technical founders who need to think in first principles about the building blocks of a great go-to-market foundation.

In later stages, founders face different challenges: pricing, managing unit economics, or building a clear and defensible conversion funnel to generate and convert leads.

But across all stages, the one nagging and costly GTM challenge I see founders face is the seemingly simple process of setting up their go-to-market infrastructure.

A website. Conversion tracking. Systems integrations. The tools that enable scaled founder-led inbound and outbound motions.

For a technical founder, it's a waste of development hours. For a commercial founder, it's either outsourced or a massive learning curve that pulls them away from selling. The typical outcome: hire a growth marketer, outsource the GTM tech, hire an agency, and pay $15–30K a month.

The hard truth of the matter is: most of that work is operational, low-value, and API-driven. Systems integration, account setup, daily optimization, bug resolutions. It's junior-level work dressed up as strategy. And nearly all of it runs on APIs that an agent can call just as easily as a human clicking through a dashboard.

Autonomous Agents

Now — enter autonomous agents.

The prospect of having an always-on, highly-skilled, autonomous agent that knows everything about each of these systems, and can just do the task on repeat? That's a real solution to a nagging problem.

So, I figured, why not see if it can be done.

Here's what I'm trying to find out:

Can a team of AI agents do this work — not just as well as a human, but better? Faster, more consistent, available around the clock? For significantly lower cost?

I'm pressure-testing the idea by building a team of agents and pointing them at my own company's website first.

Honestly, not many people visit my site right now — which makes it the perfect test bed. Low risk, real stakes. If the agents can turn it into a functioning B2B growth engine, that's a meaningful proof point. If they can't, I'll know exactly where the limits are.

This is Day Zero. I'll document what happens from here — what works, what breaks, what surprises me.

Next up: how I chose my first agents and what I asked them to do on day one.

Angus Birchall is co-founder of Elsinore Ventures, an AI-native venture advisory and startup studio. Follow along as he builds a team of AI agents from scratch — starting with his own website.