Summary: The article explains how 11x uses Firecrawl’s crawl and search endpoints to turn websites and public sources into structured company research briefs for personalized AI outbound. It cites more than 11 million requests across 175,000 domains, while noting limits around public data, vendor dependence, stale or noisy pages, cost, and rate limits.
Why AI Outbound Stalls Without Real Company Context
Put AI in charge of sales outreach and the same question shows up almost immediately: what does it take for the model to genuinely “understand” a company? Scraping the homepage is not enough. It has to know what the company sells, which problems it solves, what customer proof it can point to, and which tools sit in its tech stack. Skip that depth, and any personalized message is nothing more than a template with a company name dropped in.
11x is an AI growth platform, and revenue teams at Checkr, Rho, and Xerox are among those using it. The company has built a system that handles this problem at scale, automating GTM workflows across the entire outbound cycle—prospect identification and research on one end, personalized outreach and booked meetings on the other. Sitting at the center of that research layer is Firecrawl, a web scraping and search API that converts raw web pages into structured markdown and data. To date, 11x has pushed more than 11 million requests through Firecrawl, spanning 175,000 unique domains.
Below, we walk through how 11x turns Firecrawl into company research briefs, why that research is the foundation of effective AI outbound, and what product builders can take from the architecture. We also cover the trade-offs and limitations that come with relying on public web data.
Inside 11x’s Research Briefs, Powered by Firecrawl
Every prospect has to feel that the message was written for them. Getting there means knowing what each company sells, the problems it solves, the proof it offers customers, and how it presents its brand. Firecrawl collects that material and structures it into research briefs, which 11x then draws on to tailor its outreach.
Two endpoints account for the bulk of 11x’s Firecrawl usage:
/crawl: Works an identified company’s website in depth, following its sitemap and internal links to assemble a complete profile of the site and pick up the pages covering products, solutions, pricing, and more./search: Picks up where the site-level research stops, finding sources elsewhere on the web—customer case studies, partnership announcements, and company facts such as headcount or headquarters location.
Depth comes from /crawl, breadth from /search, and the two are meant to run together. The heavy lifting belongs to Firecrawl: finding relevant sources, following sitemaps, rendering JavaScript, extracting page content. What comes back is clean markdown and structured data, which gives 11x one research workflow that covers any site on the web—no custom parsers to write.
Jeson Patel, CTO at 11x, describes it this way: “Firecrawl gathers and structures information from across the web to give us a detailed picture of any company, from its position in the market down to the tools in its tech stack. That depth of research is what makes our outbound so context-rich.”
Research Depth Is What Sets Outbound Quality
How rich the input context is determines, almost directly, how good AI-generated outreach turns out to be. Hand the model nothing but a homepage URL, and the message stays shallow. Hand it a structured brief covering the company’s market position, customer proof, and tech stack, and the model can produce copy that reads as genuinely tailored.
What 11x’s approach shows is that research is not a nice-to-have extra—it is the foundation. The platform uses Firecrawl to build those briefs automatically, so deep research reaches every prospect without anyone doing it by hand. That, as Patel notes, is what makes outbound “context-rich.”
The lesson for product builders is plain: when designing an AI workflow that generates personalized content, put the data pipeline first. No sophisticated language model makes up for thin input. Structured research briefs of the kind 11x builds feed straight into more relevant output.
Scaling Up: 11 Million Requests So Far
Look at the raw volume—11 million requests spanning 175,000 unique domains—and you see how labor-intensive prospect research is in sales. Attempt that by hand, and a few thousand companies are already enough to overwhelm a team. Automating repetitive GTM work such as prospect research is not a play to replace salespeople; it is a way to hand back time for the human side of sales: earning trust and closing deals.
That scale also carries a message for product builders about infrastructure reliability. Moving 11 million requests means Firecrawl’s uptime, speed, and cost control all have to hold up. When you pick a tool for your own AI workflow, features alone are not the question—whether it can absorb your growth is.
Building a Research Brief Pipeline in Practice
If you’re building an AI outbound tool or a similar product, start by replicating the pattern 11x uses. Here’s a high-level workflow:
- Pick a target company out of your CRM or lead list.
- Run
/crawlagainst that company’s website to assemble a site profile. Firecrawl walks the sitemap and internal links, renders JavaScript, and hands back clean markdown. - Run
/searchfor the supporting facts: case studies, partnerships, headcount, headquarters, and any mention of the tech stack. - Merge both outputs into one structured research brief (e.g., JSON with fields like
products,customers,tech_stack). - Pass the brief to your AI model together with a prompt, and have it generate personalized outreach messages.
What follows is a simplified example of how you might call the Firecrawl API (using a hypothetical SDK):
import FirecrawlApp from '@firecrawl/sdk';
const app = new FirecrawlApp({ apiKey: 'your-api-key' });
// Crawl the company website
const crawlResult = await app.crawlUrl('https://example.com', {
limit: 50,
scrapeOptions: { formats: ['markdown'] }
});
// Search for supplementary facts
const searchResult = await app.search('Example Corp case study', {
limit: 5
});
// Combine into a research brief
const researchBrief = {
website: crawlResult.data.map(page => page.markdown).join('\n'),
externalSources: searchResult.data.map(item => item.markdown)
};It is only an illustration, yet the core idea comes through clearly: automate the research, give the output structure, and leave the personalization to the AI.
Limitations and Trade-offs
Firecrawl is a capable way to gather public web data, but there are real limitations to keep in view.
First, the data stops at what’s publicly available on the web. Internal decisions, unpublished product plans, private customer data—Firecrawl cannot reach any of it. If your outbound strategy leans on insights like those, you’ll need to supplement with other data sources.
Second, building on a single vendor like Firecrawl brings dependency risk. A spell of downtime, or a change to its pricing or policies, can ripple into your workflow. Keeping a fallback or an abstraction layer is the wise move.
Third, web data can be noisy or out of date. Not every page is accurate, and websites change. You may need validation steps or periodic re-crawls to keep your research briefs fresh.
Finally, cost and rate limits come into it. Processing millions of requests isn’t free, and usage has to be monitored to hold costs down. Firecrawl offers various plans, but you should evaluate whether the pricing scales with your expected volume.
What Product Builders Should Take Away
If you’re building AI-powered outbound, or any product that relies on personalized content, three lessons from 11x’s approach matter most:
- Research quality determines output quality. Put the effort into a data pipeline that gathers rich, structured context on each prospect. Better input means better AI output.
- Balance depth and breadth. A crawl of the company website supplies depth, but customer case studies or partnership news will slip past it. A web search supplies breadth, yet site-specific details may be thin. Draw on both to build a complete picture.
- Plan for scale from day one. Should your product succeed, millions of requests lie ahead. Choose infrastructure that’s reliable, fast, and cost-effective, and design your workflow to handle growth.
Begin on a small scale: take a handful of prospects, stand up a research brief pipeline with Firecrawl or a comparable tool, and measure how much your AI’s output improves. The results might surprise you.
As 11x has shown, automating repetitive research tasks isn’t only a question of efficiency—it enables a level of personalization that manual work could never reach. Give your AI the right context, and every outbound message can feel like it was written for that specific company.