Scrape Conference Speakers & Personalize Outreach
How to turn a public conference agenda into a qualified, message-ready prospect list in minutes — and what makes "personalized at scale" actually work versus mail-merge spam.
How do you scrape a conference speaker list and personalize outreach at scale?
Paste the public speaker page URL into SayIntel. The platform extracts every speaker, their talk title, abstract, company, and headshot, then enriches each one with verified job title, work history, and public LinkedIn. From that dataset it scores each speaker against your ICP, drafts a message that quotes the actual talk and ties it to a problem you solve, and hands you an editable queue. End-to-end on 800 speakers takes about 5–10 minutes — versus 30+ hours of manual copy-paste, browser tabs, and rewriting the same message 800 times.
What data can I actually pull from a conference speaker page?
Anything the conference makes public: full name, talk title, talk abstract or session description, company name, job title at time of speaking, headshot, social handles, and the linked personal or company website. SayIntel parses all of it in one pass — including JavaScript-rendered agendas — and normalizes inconsistent layouts (modal popups, tab carousels, multi-day tracks) into a single structured roster you can filter and act on.
How does enrichment turn a speaker name into something you can email?
After scraping, SayIntel matches each speaker against B2B data providers (Apollo, work-email enrichment, public LinkedIn) to confirm their current company, current role, and a verified work email. If the speaker changed jobs since the talk was announced, the current employer wins — that's almost always who you want to reach. Each enriched record carries a confidence score so you know which leads are safe to send and which need a manual second look.
What does 'personalized at scale' actually mean — isn't it just mail merge?
Mail merge swaps {first_name} and {company} into a template. Real personalization references something only that specific speaker would recognize — usually the title of their talk, the angle they took, or a recent post. SayIntel reads the talk abstract and the speaker's recent activity, then writes a 60–90 word opener that quotes the talk specifically, frames why you're reaching out about it, and ends with a soft ask. Every draft is unique to the person; you skim and approve rather than write from scratch.
How do I qualify speakers before spending enrichment credits on all of them?
SayIntel runs a pre-qualification pass on the raw scraped data first — title, company, talk topic, and bio against your ICP definition. Clear non-fits (wrong industry, wrong seniority, obvious mismatch) are filtered out before any enrichment runs. Only the qualified and ambiguous-but-promising leads get the full Apollo lookup. On a typical 800-speaker conference this cuts enrichment volume by 60–70% and lets you reach the actually-relevant prospects in minutes.
Can I use this for non-sales outreach — recruiting, podcast booking, partnerships?
Yes. Conference speakers self-select as decision-makers and category experts, which makes them strong targets for recruiting senior hires, booking podcast guests, sourcing investors, and opening partnership conversations. Define the ICP for whatever you're hiring for, scoring for, or pitching, and SayIntel will produce drafts in that voice — recruiter outreach reads very differently from sales outreach, and the prompt adapts.
What's the personalization-to-reply rate compared to a generic cold sequence?
Public benchmarks for cold sales email sit around 1–3% reply rate. Speaker-specific outreach that quotes the actual talk consistently runs 8–15% in our customers' data because the opener is impossible to ignore — the speaker just gave the talk you're referencing, often last week. The lift comes from relevance, not volume; sending fewer, sharper messages outperforms blasting a list.
Is scraping a public conference website legal?
Reading and processing publicly published data — speaker names, talks, public LinkedIn — is permitted in the US and EU; the foundational case law is hiQ Labs v. LinkedIn. Where the legal risk lives is in the outreach itself: GDPR (lawful basis for processing EU residents' data), CAN-SPAM (US commercial email rules), and CASL (Canada). SayIntel is the processor; the customer is the controller and is responsible for the lawful basis. See our Privacy Policy for the full subprocessor list and data-handling terms.
Why not just use Apollo or Sales Navigator filters to find the same people?
Apollo and Sales Navigator search a static database. They can tell you who works at a company in a role — they can't tell you who just spoke at SaaStr last week about the exact problem you solve. Conference attendance is a real-time buying signal: someone who flew across the country to talk about RevOps tooling is meaningfully more reachable than the same title pulled from a contact list. SayIntel layers that signal on top of the database, so you reach the right person at the right moment.
How do you keep messages from sounding AI-generated?
Three things: the model is grounded in the speaker's actual talk text (not invented), drafts default to short and direct (60–90 words, one ask), and every draft is human-edited before send — SayIntel is a queue, not an autosender. You scan, tweak the one line that doesn't sound like you, approve. That keeps reply rates high and your sender reputation clean.