Why Apollo

I built Apollo while working on a consulting project — automating a repetitive back-office task with AI. The kind of thing someone did manually every day that was obviously automatable.

So I built it. And it worked. But along the way I learned something more interesting about how AI agents behave.

The more rules I gave the AI — “always validate this,” “format it like that,” “check these three things first” — the worse it performed. It over-engineered everything. It added unnecessary steps, defensive checks, multiple passes over the same data. Costs went up. Quality went down.

When I stripped all of that away and just said “here's the context, figure it out” — everything improved. Fewer iterations, lower costs, better results. The AI reasoned more naturally when I stopped telling it how to reason.

That insight changed what I was building. The task didn't matter anymore. What mattered was the pattern — give an AI the right context and tools, don't prescribe behavior, and it just works. For anything.

Today, Apollo is how I work. I email it like I'd email a colleague. Pull a report. Chase someone on an invoice. Build a spreadsheet. Check my QuickBooks. Research something. It remembers what I've told it, follows up when things are due, and connects to the tools I actually use.

There's no app to learn. No dashboard to check. You just email, and it handles it.

It's been genuinely useful for me, and I wanted to put it out there for anyone who might feel the same way.

— E