How we automated an entire email marketing engine with AI

Email is one of the highest leverage channels a brand owns, and also one of the most manual. Someone has to research the audience, write on brand copy, schedule the send, read the results, and decide what to do next. Week after week. For Soul Guide 427, an in house Human Design brand, that work sat on senior people and capped how often the brand could show up in the inbox.

So we built a system that runs the whole loop itself. This is how it works and what it changed.

Before automation, every campaign meant the same sequence by hand: pull audience signals, decide on an angle, write the subject line and body, format it in the email platform, send, then manually read opens and clicks and guess at the next move. It was slow, it was inconsistent, and it did not compound. Every campaign started from a blank page.

The brief was simple. Keep the human taste and brand voice, but take the repeatable work off people and let the system improve on its own.

We designed the engine around four jobs that used to be done manually, each handled by an agent that reads, decides and acts rather than following a rigid script.

1. Research

Agents mine audience signals and source material before each campaign, so every send starts from real insight about what the audience is responding to, not a guess.

2. Copywriting

On brand copy is drafted end to end, from subject lines to body to calls to action, in the Soul Guide voice. The team reviews and approves rather than writing from scratch.

3. Data analysis

Every send is analysed automatically (opens, clicks, segments) and the results become the input for the next decision instead of sitting in a dashboard nobody reads.

4. The learning loop

This is the part that makes it compound. Every campaign the team approves and every result that comes back feeds into the system, so the copy and the targeting get sharper the more it runs. The brand's best work turns into an asset that keeps paying off.

We pulled the numbers straight from Klaviyo across 102 campaigns over twelve months, split at the point the AI agent took over. The clean, defensible wins:

Open rate nearly doubled on a healthier, more engaged list. Unsubscribes were cut in half, moving from above the 1.5% danger line to well under it. On the engaged segment offers, each email drove roughly four times more conversions than the old mass blast approach, so the brand earned more from a smaller, healthier list. Spam complaints and bounces both dropped too.

The reason this beats a typical automation is that the agents handle the judgment, not just the busywork. A rules engine can schedule a send. It cannot read a transcript, understand what an audience cares about, and write copy that lands. Agentic systems can, and because they learn from every approved output, they do not plateau.

Email marketing is one of dozens of repeatable workflows most companies run by hand. Lead capture, proposals, reporting, onboarding, invoicing: the same pattern applies. If your team does something the same way every week, it can usually be automated or AI assisted, built to production standard, and handed to you to own.

If that sounds like work your team is stuck doing, book a free call and we will scope what we can take off their plate.