Sales, Lead Gen & Outreach

An AI-Powered Lead Generation and Email Outreach Platform With a Four-Stage Discovery Pipeline

How Pfactorial Technologies built a platform that searches, crawls, and scores prospective business leads, then runs compliant automated email outreach against them.

August 21, 2026

OUTLINE

ENGAGEMENT SNAPSHOT

Figure 1 - Key figures from this engagement, at a glance.

EXECUTIVE SUMMARY

Our client's sales and business development team needed to find prospective companies matching a market description, judge whether their own services were a genuine fit for each one, and follow up by email - a workflow they were running manually, one search and one message at a time.

Doing this well isn't a single search-and-send step: a market query returns companies, not qualified leads, so each website has to be crawled and summarized before any fit judgment is possible, and a real email campaign needs delivery tracking, an unsubscribe path, and a way to avoid double-scanning the same query while a scan is already running.

Pfactorial built Lead Radar: a four-stage pipeline - search, crawl and extract, service-match, and persist - feeding a campaign system with real-time progress over Server-Sent Events, template-based outreach, and a scheduler that runs compliant, interval-based sends without manual triggering.

Why this engagement is representative This engagement demonstrates Pfactorial's approach to turning a manual research-and-outreach workflow into a pipeline with real-time visibility at every stage, backed by a local LLM to keep per-lead evaluation cost low at scale.

THE CHALLENGE

Automating lead discovery and outreach meant solving problems that only show up once real websites, real inboxes, and real compliance requirements are involved.

1. A market query returns companies, not qualified leads

Search results only give a name and a website; deciding whether a company is actually a fit for the services on offer requires reading and summarizing the page first.

2. A user needs to see results as they arrive, not after the whole scan finishes

A market scan can take a meaningful stretch of time to crawl and evaluate every result, so the interface needed live, progressive results rather than a single response at the end.

3. The same scan can't run twice at once

Without protection, a page refresh or a second submission could kick off a duplicate scan against the same query, wasting crawl and LLM budget and producing duplicate leads.

4. Outreach at scale needs built-in compliance, not an afterthought

Every recipient needs a working unsubscribe path and the system needs to honor it permanently, or automated outreach becomes a liability rather than a growth channel.

The real brief Not "search for companies and email them" but "crawl and judge fit for each company in real time, and never send a message to someone who has opted out."

THE SOLUTION

Pfactorial built Lead Radar around a four-stage discovery pipeline that streams results to the frontend as they're found, feeding into a campaign system with template-based outreach and a scheduler that respects delivery intervals and unsubscribes.

Figure 1 - Each stage streams its result to the frontend as it completes, rather than waiting for the whole scan to finish.

Architectural principles

Stream progress, don't make the user wait for a batch response - Server-Sent Events push a notification after every successful extraction, so the results table fills in live instead of the frontend polling or waiting for the entire scan to finish.
Run the matching model locally, not against a metered API - Service-fit evaluation runs against a local LLM rather than a hosted API per call, keeping the cost of judging fit for hundreds of candidate companies predictable.
Prevent duplicate work at two layers - The backend rejects a second scan while one is already running, and the frontend independently checks its own session state before ever submitting - belt and suspenders against wasted crawl budget.
Make opting out permanent and load-bearing - Every lead gets a unique unsubscribe token at save time, and clicking it excludes that contact from every future batch send, not just the current campaign.

CAPABILITIES DELIVERED

Each capability covers one stage of the journey from a market description to a compliant, tracked outreach campaign.

CAPABILITY	WHAT IT DOES
Market-query lead discovery	Users describe a target market and the services they offer; the platform searches for matching companies and returns only those with a live website.
Automated company summarization	Each company's website is crawled and summarized by a local LLM, so a user never has to read the site themselves to know what the company does.
Evidence-based service-fit scoring	The platform recommends up to two services per company with supporting evidence and a confidence level, rather than a generic yes/no match.
Real-time scan monitoring	A live dashboard shows leads populating as they're found, and resumes automatically if the page is refreshed mid-scan.
Template-based email campaigns	Four preset HTML templates with a live preview let a user launch outreach without writing email markup by hand.
Scheduled, compliant sending	A background scheduler sends batches on a user-defined interval and permanently excludes anyone who unsubscribes from future sends.

Figure 2 - From a market description to compliant, scheduled outreach, in four stages.

Design note Duplicate-scan protection is enforced at two independent layers - a backend status check and a frontend session check - rather than trusting either one alone, because a page refresh mid-scan is exactly the moment a naive single-layer guard would fail.

ENGINEERING FOR SCALE AND RELIABILITY

Six decisions kept the pipeline fast to run and safe to leave unattended between scans and sends.

Server-Sent Events instead of frontend polling

The frontend subscribes to a per-campaign event stream and only re-fetches data in response to an actual state change, rather than polling on a timer regardless of whether anything changed.

A local LLM for extraction and matching

Web extraction and service-matching both run against a locally hosted model rather than a metered hosted API, keeping the marginal cost of evaluating each candidate company low.

A single-poll-loop scheduler instead of a task queue

One background loop checks all active campaigns every 60 seconds and dispatches due sends, avoiding the operational overhead of a message broker and worker fleet for a workload this size.

Email sends on a thread pool, off the event loop

SMTP delivery runs on a dedicated thread pool so a slow mail server can't block the async event loop that's also serving the live dashboard.

Upsert-by-website-and-campaign to prevent duplicate leads

A unique index on website and campaign ensures re-scanning or re-processing a company updates the existing lead record instead of creating a duplicate.

A generated, unique token per lead for unsubscribe

Each lead gets its own unsubscribe token at save time, so an opt-out is enforced per contact rather than relying on a shared or guessable link.

DELIVERY APPROACH

The engagement built outward from the discovery pipeline, so the campaign and outreach layers had real, scored leads to work against from the start.

1. Pipeline foundation - building the four-stage search, crawl, match, and persist pipeline with real-time event notification wired in from the start.

2. Real-time dashboard - building the Server-Sent Events channel and the scanner interface that populates results live as the pipeline runs.

3. Campaign & template system - adding campaign creation, lead selection, and a set of preset HTML email templates with live preview.

4. Outreach & compliance - wiring SMTP delivery, per-lead unsubscribe tokens, and delivery-status tracking into the sending path.

5. Scheduler - adding the background poll loop that dispatches interval-based batch sends without manual triggering.

RESULTS AND IMPACT

- Key outcomes from this engagement.

The four-stage pipeline, real-time dashboard, and scheduled outreach system are built and operational, taking a market query through search, crawl, service-fit scoring, and persistence with live progress visible throughout.

Because duplicate-scan protection, per-lead unsubscribe tokens, and delivery-status tracking are built into the pipeline rather than bolted on afterward, the outreach layer can run unattended on a schedule without a person checking compliance by hand.

What it enabled commercially

The sales and business development team can now point the platform at a defined market, get evidence-scored leads back automatically, and run interval-based outreach against them - turning what was a manual research-and-email workflow into something that runs on a schedule.

WHY PFACTORIAL

This engagement reflects Pfactorial's approach to building discovery-to-outreach pipelines with real-time visibility at every stage and compliance - unsubscribe handling, delivery tracking, duplicate protection - treated as core requirements rather than later additions.

- Service lines this engagement draws on.

Engagement enquiries Pfactorial Technologies works with sales and business development teams that need lead discovery and outreach turned into a pipeline with real visibility and built-in compliance, not a set of manual scripts. If you're evaluating a lead-generation or outreach platform, we're happy to give you an honest read on scope and risk before anyone commits to anything. · pfactorial.ai

APPENDIX A - TECHNOLOGY STACK

The technology stack underpinning the system, grouped by the layer it serves.

Result and Analysis

ENGAGEMENT SNAPSHOT

How Pfactorial Technologies built a platform that searches, crawls, and scores prospective business leads, then runs compliant automated email outreach against them.