Give an AI writing tool a prospect's name, company, and job title, and it will produce an email in seconds. Give it a real signal – the event they're attending, a specific reason to reach out this week – and reply rates change. Personalized outreach at scale is the same tool either way; what changes is the data behind the prompt.
Why AI drafts default to generic
Salesforce's 2026 State of Sales report – a survey of more than 4,000 sales professionals – found that 87% of sales organizations now use some form of AI, and sellers expect fully implemented AI agents to cut prospect research time by 34% and email drafting time by 36%. That's a real gain, but the same report flags where it turns hollow: top-performing teams are 1.7x more likely to use AI agents than underperforming ones, which means the tool itself isn't the differentiator – it's the discipline behind it.
Lavender's Cold Email Benchmark Report, built from 231,818 cold emails across roughly 50,000 inboxes, shows the same pattern from the recipient's side. Emails that score well on Lavender's grading – built around specificity and a human voice, not keyword density – post meaningfully higher reply rates than average sends in the same segment; operations leaders reply 58% more often to a well-graded email than to the department's typical send. The lift comes from what's said, not from whether AI helped say it. Ask an AI tool to "personalize" with only a name and a title, and it fills the gap with the same safe, referential phrasing every other AI-drafted email uses – and that sameness is exactly what a busy inbox filters out.
Feed the model an event, not a mail-merge field
The fix for SDRs isn't writing every email by hand again – it's changing what goes into the draft. A prospect's attendance at an upcoming event is a concrete, dated, specific detail an AI model can build a real opening line around. "Saw you're on the list for [event] next month" reads as researched because it is, while "I noticed you're in the [industry] space" is the same sentence every vendor sends. The event doesn't need to be dramatic – a confirmed attendee list, a speaking slot, or a booth assignment are all specific enough to produce a first line an AI isn't also handing to five other prospects off the same template.
The same logic applies to any dated trigger – a funding round, a leadership hire, a product launch – but events are unusually reliable inputs because the date and reason are already public and shared with the recipient, so referencing one doesn't read as surveillance the way a scraped social post sometimes can. If you want ready-to-use starting points built around exactly this kind of trigger, our event outreach email templates post has three you can adapt with real signals instead of placeholders.
Split research from drafting
SDR teams get more out of AI drafting when they separate the two jobs: a research step that surfaces the specific, current signal for each account, and a drafting step where AI turns that signal into a short, on-brand email. Asking one AI pass to do both – research and write – is where outreach slides back into generic territory, because the model fills gaps in its own research with plausible-sounding but non-specific language.
Scryon's event and account data is built for that first step: attendee lists, exhibitor rosters, and fit scoring feed the drafting step with something real to reference instead of leaving the model to guess. Once that data is flowing, a rep can run the same sequence across a hundred accounts and have each opening line be genuinely different – not because a human wrote a hundred emails, but because the input to each draft was actually different.
Keep a human in the loop before sending
Even with good data behind the draft, a 30-second read before sending catches the two failure modes that erode trust fastest: a wrong or outdated detail (an event date that moved, a role someone no longer holds) and a tone that reads as AI-generated even when the facts check out. Teams running this hybrid model – AI drafting from real signals, a human reviewing before send – consistently outperform both fully manual outreach, which is too slow to cover a full pre-event list, and fully automated sending, which tends to drift generic the moment volume goes up.
Personalizing at scale was never really about writing faster. It's about having something true and specific to say to more people, and letting AI handle the mechanical part of turning that into a well-formed email. Get the account and event data in front of your reps before the next show, and the "at scale" part stops being a trade-off against sounding real.
For the product side of this motion, see For Sales. Related reading: event outreach vs cold outreach.