What doesn't work anymore is sending the same message to everyone and hoping it lands.
The inbox is packed with noise, irrelevant promos, lazy personalization, and content that doesn't match what the reader actually cares about. So when your emails don't get opened or clicked, it's usually not because email is dead. It's because the message wasn't relevant.
I recently watched an episode of HubSpot's Field Notes, "How AI-Powered Email Skyrockets Conversion," and it clicked. They're not using AI to write better subject lines. They're using AI to make email more useful and more specific to the person receiving it.
Lucy shared they're seeing 50% average conversion lifts, and in some placements, 200%, 300% spikes, by automating the full flow from creation to send, with real guardrails and real QA.
As an agency owner who's spent 20+ years living at the intersection of storytelling + performance (newsroom days taught me to respect attention; agency life taught me to respect attribution), this validated the direction the industry is heading:
AI doesn't replace strategy. It forces you to finally get serious about relevance.
Here are my key takeaways and what they mean for agencies aiming to scale without turning into a generic content factory.
One of the most important points Lucy made: personalization for the sake of personalization doesn't work.
They tested AI-generated, personalized copy promoting a generic marketing certification. The words changed. The results didn't. The lift came when they used AI to semantically match a contact's job title, industry, and company size to a different resource, one that actually solved that specific segment's pain point.
If your client has one hero eBook and you're just swapping intros and subject lines, you're leaving money on the table.
AI makes it realistic to build dynamic resource-matching engines:
We're not just writing emails anymore. We're building systems that route the right value to the right person at the right time.
Here's the part nobody wants to hear: this entire system collapses with bad data.
HubSpot's team uses a custom workflow that calls out to OpenAI, processes user data, and stamps it back into CRM properties before sending.
That's not AI magic. That's data architecture + automation discipline.
Most clients want the AI layer first. But their CRM is usually:
So the first win is rarely writing better emails. The first win is: clean the inputs so the machine stops guessing.
And yes, small businesses with fresh CRMs can actually have an advantage here. No legacy baggage. No Frankenstein workflows from 2017.
If you want an LLM to sound like your client's brand, a two-sentence prompt won't cut it.
Lucy shared that their prompts are literally pages long. They include:
And then the part I respect most: human QA is non-negotiable.
They'll send 50-300 test emails to a Google Sheet or Slack channel for manual review before fully automating.
One-click AI is a myth. Meaningful conversion lifts usually don't happen until the 5th, 6th, or 7th prompt iteration, and that's before you even talk about segmentation logic.
So agencies need to price and position this correctly:
The work didn't disappear.
It shifted from drafting emails to engineering instructions and protecting brand trust while you scale.
I've seen this pattern across every channel: video, web, inbound, social.
When you have:
AI becomes a multiplier. When you don't, AI becomes a faster way to ship meh at scale.
If you want to test this without blowing up your client's brand (or your own reputation), start here:
This episode is a must-watch if you're ready to move beyond batch and blast and into automation that actually respects the customer.
Link to watch the full video: [How AI-Powered Email Skyrockets Conversion]
Have you started testing fully automated AI workflows in your email marketing (for you or your clients)?
What's working, and what's breaking?