The pitch for AI in outbound used to be "write your emails for you." That was always the wrong use case — generic AI copy is easy to spot and converts worse than a human-written template. The real leverage is upstream of copywriting.
Research, not writing, is the bottleneck
A skilled SDR spends most of their week not writing emails but figuring out which 200 accounts are worth emailing this week, and what's true about each one right now. That's the part AI agents are actually good at: pulling signal from job changes, funding announcements, hiring pages, and tech stack data, then summarizing it into something a human can act on in seconds.
We run this as a standing pipeline for every client account:
- Daily signal scraping across firmographic and intent data sources
- An AI layer that scores and ranks accounts against the ICP, not just matches keywords
- A structured brief per account that a copywriter uses as raw material — not a finished email
Where AI-written copy still fails
Fully AI-generated first-touch emails read fine in isolation and convert poorly in aggregate, because every prospect has now seen a dozen emails with the same generic structure. The accounts that reply are the ones where the email references something specific and verifiably true about them — and that's still a step a human writer does better than a model, even with good inputs.
The honest framing
AI compresses the research and qualification stage from hours to minutes per account. That time gets reinvested into better targeting and sharper, more specific copy — not into sending more volume. Teams that use the time savings to just send more emails see deliverability drop and reply rates fall with it.
A concrete before/after
Before: an SDR manually researches 15 accounts a day and personalizes lightly. After: an AI agent triages 150 accounts a day, surfaces the 15 with the strongest signal, and the SDR spends the same hour writing sharper copy for fewer, better-fit accounts.
