How to Turn Website Visitors Into Leads With AI

Website visitor identification is the process of determining who is visiting your website — not just how many people, but which companies and, in some cases, which specific individuals. Google Analytics tells you that 500 people visited your site last week. Visitor identification tells you that 12 of them were from companies you want as clients, what pages they looked at, and how to reach the decision-maker.

That distinction is the difference between a dashboard you glance at and a pipeline you sell from.

This guide breaks down a 5-stage system for turning anonymous website traffic into qualified leads — from initial visitor identification through enrichment, suppression, personalized outreach, and CRM integration. The through-line is a real case study: Eric Siu, Chairman of Single Grain and CEO of ClickFlow, used an AI agent to build this entire pipeline in 11 minutes on a live demo.

If you run a B2B company or an agency, the enterprise version of this system uses tools like RB2B, Instantly, and HubSpot. If you run a local service business — a dental practice, law firm, or restaurant — there is a simpler version that uses tools you may already have. Both are covered here.

This post is part of our AI automation series. If you are new to the concept, start there for the fundamentals.

The 5-Stage Visitor-to-Lead Pipeline

Most businesses treat their website like a brochure. Someone visits, they either call or they don't, and that is the end of the interaction. The 96% who leave without converting — roughly 96.4% in B2B — are gone forever (First Page Sage).

A visitor-to-lead pipeline changes that. Instead of hoping people fill out a form, you identify who they are, enrich their data, filter out existing contacts, reach out with relevant messaging, and track it all in a CRM.

Here are the five stages:

1. Visitor Identification — Know who is on your site
2. Data Enrichment — Fill in the contact details
3. Suppression Pipeline — Filter out people you should not contact
4. Personalized Outreach — Reach out with relevant sequences
5. CRM Integration — Track everything in one place

Each stage builds on the previous one. Skip a stage and the system breaks. Most lead generation content only covers stages 1 and 4 — identify and blast. That is why most outbound campaigns underperform.

Stage 1: Visitor Identification

The first question is simple: who is on your website right now?

Enterprise Approach: Visitor ID Pixels

Tools like RB2B, Leadfeeder, and Clearbit Reveal install a tracking pixel on your website. When someone visits, the tool matches their IP address against business databases to identify which company they work for — and in some cases, the specific person.

The match rates vary significantly. Person-level identification tools match 5-20% of anonymous traffic. Company-level tools match 30-65% (MarketBetter).

Eric Siu runs RB2B pixels across three properties: Single Grain (his agency), ClickFlow (his SaaS), and Carrot (a real estate marketing platform). Each pixel identifies visitors across those sites and feeds them into the same pipeline. The key insight is that he segments by property from the start — a visitor to an agency website has different intent than a visitor to a SaaS product page.

The major tools in this space:

Person-level identification is more actionable but less common. Company-level identification gives you higher volume but requires a second step to find the right contact within that company. Both have their place depending on your sales motion.

Local Business Approach: Behavioral Tracking

Most local service businesses do not need a visitor identification pixel. Their customers are consumers, not companies, and person-level consumer identification raises compliance concerns that are not worth navigating at small scale.

Instead, local businesses should focus on behavioral tracking:

• GA4 audience building — Create audiences based on page behavior (visited pricing page, viewed 3+ service pages, spent 2+ minutes on site) and use those audiences for retargeting
• Meta pixel — Retarget website visitors with Facebook and Instagram ads. You do not know their name, but you can show them relevant ads after they leave your site
• Call tracking — Tools like CallRail assign tracking numbers to different traffic sources so you know which visitors called, even if they did not fill out a form
• Form submissions — The most direct identification method. Every form fill gives you a name, email, and usually a phone number

The goal is the same: move from "500 anonymous visitors" to "12 people who showed buying intent." The tools are just different.

For a deeper look at how CRM systems connect to this, see our CRM automation services.

Stage 2: Data Enrichment

Identification tells you who visited. Enrichment fills in the details you need to actually reach them.

Enterprise Approach: Automated Enrichment

Once RB2B identifies a visitor, the next step is enriching that record with contact information — email, phone, job title, company size, LinkedIn profile. This happens automatically through integrations with data providers.

The common enrichment tools:

• RB2B's built-in enrichment — Provides email and basic company data for identified visitors
• Apollo.io — Database of 275M+ contacts with email verification. Strong for B2B outreach. Plans start at $49/month.
• ZoomInfo — Enterprise-grade B2B database. More accurate for large companies. Starts at ~$14,000/year.
• Clearbit — Company data enrichment that auto-fills CRM records. Now part of HubSpot.
• Clay — Workflow tool that chains multiple enrichment sources together (waterfall enrichment). If Apollo does not have the email, Clay checks the next source automatically.

Eric Siu's system enriches RB2B data automatically before it hits the suppression stage. By the time a visitor record reaches the outreach step, it already has a verified email, company name, job title, and the pages they visited. No manual research required.

The concept of waterfall enrichment is worth understanding: instead of relying on one data source, you chain multiple sources in sequence. If Source A does not have the email, check Source B. If Source B does not have the phone number, check Source C. Tools like Clay automate this entirely. The result is a significantly higher enrichment rate than any single tool provides.

Local Business Approach: Manual + CRM Enrichment

For local businesses, enrichment is simpler but equally important:

• Google Business Profile insights — Shows you search queries people used to find you, actions they took (calls, direction requests, website visits), and peak hours
• Call tracking metadata — CallRail and similar tools capture caller ID, call duration, and call recordings. A 45-second call is someone asking hours. A 6-minute call is a potential client.
• Form data — Add 2-3 qualifying fields to your contact form (service interested in, timeline, budget range) and you have enrichment built into your intake process
• CRM auto-enrichment — HubSpot's free CRM automatically enriches company records with public data. Enter an email domain and it pulls company name, size, industry, and social profiles.

The principle is the same at every scale: do not let a lead enter your system as just a name and email. The more context you have, the more relevant your follow-up can be.

Stage 3: Suppression Pipeline

This is the stage that separates a professional system from spam. And it is the stage that most lead generation content completely ignores.

A suppression pipeline filters out people you should not contact: existing customers, current email subscribers, people already in your CRM, paying users of your product, and anyone who has previously opted out.

Why Suppression Matters

The practical reason: deliverability. Senders with list hygiene scores above 95% achieve 97% inbox placement. Drop below 85% and inbox placement falls to 76% — a 21-point gap that means a quarter of your emails never reach anyone (Mailgun).

The strategic reason: reputation. If a paying customer gets a cold email from you pitching services they already buy, you have damaged trust. If a warm lead in your CRM gets a generic cold email because your outbound system did not check the CRM first, you have undermined your own sales process.

The legal reason: CAN-SPAM requires honoring opt-out requests. If someone unsubscribed from your newsletter and then gets a cold outreach email, you have a compliance problem.

Eric Siu's Suppression Stack

Eric's AI agent checks every identified visitor against five sources before adding them to an outreach campaign:

1. HubSpot CRM — Is this person already a contact, deal, or customer?
2. ActiveCampaign — Are they on any existing email marketing lists?
3. Mixpanel — Are they an active product user? (For ClickFlow and Carrot)
4. Existing email lists — Have they already been contacted through other campaigns?
5. Paying customer database — Are they currently paying for any product?

Only after passing all five checks does a visitor enter an outreach sequence. This is not optional. This is the infrastructure that makes the rest of the system work without destroying your sender reputation or annoying your existing customers.

Local Business Suppression (Simpler, Still Critical)

For a dental practice, law firm, or restaurant, the suppression logic is simpler but no less important:

• Check your patient/client management system — Do not send a "we miss you" email to someone who had an appointment yesterday
• Check your CRM — If someone is already in an active sales conversation, do not blast them with marketing emails
• Check opt-out lists — Anyone who unsubscribed, ever, stays off outbound lists
• Check recent communication — If someone received an email from you this week, they should not get another automated one tomorrow

Most local business CRMs handle this automatically if configured correctly. The problem is that many businesses run their email marketing, CRM, and patient management as separate systems that do not talk to each other. That is how a patient gets a "book your first appointment" email the day after their cleaning.

A connected system — where your CRM, email tool, and booking system share data — eliminates this. See our CRM automation services for how we set this up.

Stage 4: Personalized Outreach Sequences

After identification, enrichment, and suppression, you have a clean list of people who visited your website, showed buying intent, and are not existing customers or contacts. Now you reach out.

Enterprise Approach: Segmented Cold Email

Eric Siu built five separate Instantly campaigns in a single session, each targeting a different audience:

1. Single Grain Services — General ICP — Companies that visited Single Grain's agency pages
2. ClickFlow Agency — Agencies that visited ClickFlow's product pages
3. Carrot Clients — Real estate professionals who visited Carrot
4. Single Grain Services — Enterprise — Larger companies showing interest in agency services
5. ClickFlow Direct — Non-agency companies interested in the ClickFlow product

Each campaign had its own email sequences, custom variables, and sending accounts. The segmentation matters because a real estate agent who visited Carrot needs a completely different message than an enterprise VP who browsed Single Grain's case studies.

Key operational decisions Eric made (and that most people get wrong):

No calendar links in initial cold emails. The ask is "are you open to a call?" not "book a time on my calendar." Calendar links in a first cold email feel presumptuous. They assume the person wants to talk before you have earned that. A simple question gets higher reply rates.

No icebreaker variable. Many cold email templates start with a personalized line ("I saw you posted about X on LinkedIn..."). Eric deliberately skipped this. His reasoning: fewer variables mean fewer points of failure. A broken merge tag in an icebreaker line is worse than no icebreaker at all.

First-name fallback to "Hi there." If the enrichment data does not include a first name, the email opens with "Hi there" instead of "Hi ." Simple, but many campaigns break when a name field is empty.

1-2 email accounts per campaign. Not 10. Not 20. Spreading volume across too many accounts actually hurts deliverability for new accounts. Better to send fewer emails from established, warmed accounts than to blast from a bank of fresh domains.

The Numbers on Cold Email

The average B2B cold email reply rate is 5.8%, based on a study of 16.5 million cold emails (Belkins). That is the baseline for everyone.

But the distribution is not even. Targeting one person per company yields a 7.8% reply rate. Blasting 10+ people at the same company drops it to 3.8%. Personalized emails get roughly 2x the reply rate of generic templates.

That lines up with Eric's approach: segment by audience, customize by intent, and target the right person — not everyone at the company.

Local Business Approach: Automated Follow-Up Sequences

Local businesses rarely send cold email. Their version of Stage 4 is automated follow-up for people who have already raised their hand — form submissions, phone calls, appointment requests, and quote inquiries.

The principles are the same:

• Segment by intent. Someone who requested a quote for dental implants gets a different follow-up than someone who asked about teeth whitening. Someone who called about a personal injury case gets a different sequence than someone asking about a traffic ticket.
• Time your follow-up. Contacting a lead within 5 minutes makes you 21x more likely to qualify them compared to waiting 30 minutes, and 100x more likely to make contact at all. 78% of customers buy from the first business to respond (Harvard Business Review / MIT).
• Use sequences, not single emails. A follow-up is not one email. It is a sequence: immediate confirmation, value-add at 24 hours, check-in at 72 hours, final follow-up at 1 week. Most businesses send the first email and stop. The leads that convert are often the ones who needed the second or third touch.

Tools like Mailchimp, ActiveCampaign, and email automation platforms handle this without any coding. The setup takes a few hours. The ROI is measured in recovered leads that would have gone to a competitor.

Stage 5: CRM Integration

The final stage is where everything connects. Without a CRM as the single source of truth, stages 1-4 create chaos: duplicate outreach, missed follow-ups, and no way to measure what is working.

Enterprise Approach: Native Integrations

Eric connected Instantly directly to HubSpot using Instantly's native integration. Every reply, every email open, every campaign interaction syncs to the CRM automatically. His team manages replies from HubSpot's unified inbox — not from five different Instantly campaigns.

This is critical for two reasons:

1. No duplicate outreach. If a salesperson is already talking to a contact in HubSpot, the outbound system will not email them separately. The suppression pipeline (Stage 3) checks HubSpot before every send.
2. Attribution clarity. When a deal closes, you can trace it back to: which website they visited, which page they viewed, when they were identified, which email sequence they received, and when they replied. That data tells you which campaigns to scale and which to kill.

Eric's AI agent also sent a summary to Slack and created a Google Doc changelog documenting every campaign, every sequence, and every integration. Documentation is not glamorous, but it is what lets a team operate a system instead of depending on the person who built it.

Local Business Approach: Free CRM + Connected Tools

For local businesses, the CRM setup is simpler:

• HubSpot Free CRM — Unlimited contacts, email tracking, deal pipeline, task management. Free forever. It is the best free CRM for local service businesses and it is not close.
• GoHighLevel — Built specifically for agencies and local businesses. Combines CRM, email, SMS, booking, and pipeline management in one platform. Starts at $97/month.
• Jobber / ServiceTitan / Dentrix — Industry-specific tools that act as both CRM and practice management. If you already use one, connect it to your email tool rather than adding a second system.

The minimum viable setup: one CRM, one email tool, connected. Every form submission, phone call, and appointment request flows into the CRM. Every follow-up sequence triggers from the CRM. Every interaction is logged in one place.

If your current setup involves checking three different inboxes, a spreadsheet, and a sticky note to figure out who you need to call back — you need a CRM.

How Eric Siu Built This in 11 Minutes

The most striking part of Eric Siu's demonstration was not the system itself — it was the speed. He went from a written spec to five live outbound campaigns with full CRM integration in approximately 11 minutes.

Here is how it actually worked:

The Setup (Before the Clock Started)

Eric wrote a 16-page spec describing exactly what he wanted: which tools to connect, which campaigns to create, what the sequences should say, how suppression should work, and how everything should integrate with HubSpot.

He was transparent about this: "I workshopped this with AI quite a bit" before the recorded demo. The AI agent did not figure out the strategy. Eric did. The AI agent executed a well-defined plan.

This is an important distinction. The value of AI in this system is not "AI figures out your marketing strategy." The value is "AI builds the infrastructure in minutes that would take a human days."

The Execution (11 Minutes)

1. Eric pasted the 16-page spec into his AI agent (he calls it "Alfred")
2. The agent read the spec and asked clarifying questions: Which Instantly workspace? Which HubSpot properties? Which suppression sources?
3. Eric answered via Super Whisper (voice-to-text) — faster than typing
4. The agent built all five campaigns in parallel, not sequentially
5. Eric opened Instantly and spot-checked: refreshed the dashboard, previewed sequences, verified custom variables
6. The agent connected Instantly to HubSpot via native integration
7. The agent sent a summary to Slack and created a Google Doc changelog

What AI Built

• 5 separate email campaigns in Instantly
• Complete email sequences for each campaign (subject lines, body copy, follow-up timing)
• Custom variable logic (first name with fallback, company name, page visited)
• Suppression pipeline connected to 5 data sources
• HubSpot integration for unified reply management
• Slack notification with campaign summary
• Google Doc changelog documenting every decision

What AI Did Not Do

• Approve the email copy (Eric previewed every sequence)
• Decide which sending accounts to use (Eric assigned them)
• Verify brand voice compliance (Eric read the emails)
• Set the overall strategy (Eric wrote the 16-page spec)

This is the pattern we see repeatedly with AI automation: AI is exceptional at executing defined tasks at speed. It is not a substitute for knowing what to build or whether the output is correct.

The speed-to-lead data makes this relevant beyond convenience. If contacting a lead within 5 minutes makes you 21x more likely to qualify them (Harvard Business Review), then the speed of building the system directly impacts the speed of responding to leads. A pipeline that takes two weeks to set up manually is two weeks of leads slipping through the cracks. A pipeline that takes 11 minutes means you are live the same afternoon.

For more on how AI agents are changing marketing workflows, see our post on building an AI SEO agent with Claude Code.

What This Looks Like for a Local Business

You do not need RB2B, Instantly, or a 16-page spec. The underlying pattern — identify interest, filter noise, personalize follow-up, track in CRM — scales down to any local business with a website and a phone number.

Dental practice. A potential patient browses your dental implants page, reads the FAQ, looks at your financing page, and leaves without booking. With basic automation: they see a retargeting ad the next day featuring implant before-and-after photos. If they previously gave you their email (say, for a cleaning appointment), they get an automated email with your implant FAQ, financing options, and a direct booking link. All tracked in your CRM so the front desk knows the context when the patient calls. See our dental marketing services for how we set this up.

Law firm. A potential client visits your personal injury page, starts filling out your intake form, and abandons it halfway through. With automation: they receive a follow-up email within 1 hour ("We noticed you started a consultation request — no pressure, but we are here if you have questions"). At 48 hours, they get a case study showing a similar outcome. At 1 week, a final touchpoint offering a free consultation with a direct phone number. The entire sequence runs without your paralegal lifting a finger. See our law firm marketing services for the full approach.

Restaurant. A corporate client submits a catering inquiry through your website. You send a confirmation. No response for 24 hours. With automation: they receive your catering menu PDF at 24 hours, a prompt asking about their event details and guest count at 48 hours, and a seasonal special offer at 1 week. If they reply at any point, the sequence stops and your catering manager takes over. See our restaurant marketing services for more.

The principle is always the same: someone showed interest, so follow up in a way that is relevant, timely, and not annoying. You need a CRM, an email automation tool, and someone to set up the sequences. That is the minimum viable version of what Eric built for enterprise.

Red Flags and What Can Go Wrong

This system is powerful when built correctly and damaging when built carelessly. Here is what to watch for.

Compliance Is Not Optional

CAN-SPAM (US): Every commercial email must include a physical mailing address and a clear unsubscribe mechanism. Opt-out requests must be honored within 10 business days. This applies to cold email — there is no exemption for "they visited my website."

GDPR (EU/UK): If you serve European customers, you need explicit consent before any outreach. B2B visitor identification is a gray area under GDPR. Company-level identification is generally defensible under "legitimate interest," but person-level identification and outreach without consent is risky. Consult a lawyer before running this system for EU audiences.

B2B vs B2C: B2B visitor identification using company-level IP matching is generally legal and accepted in the US. B2C person-level identification for cold outreach is a different legal and ethical territory. The system described in this post is designed for B2B.

"AI Built It" Does Not Mean "AI Verified It"

Eric spot-checked every campaign. He refreshed Instantly, previewed sequences, verified variables, and read the email copy before activating anything. This is not a nice-to-have. It is the difference between a system that generates revenue and one that generates spam complaints.

AI is fast but not infallible. It can:

• Use the wrong merge tag and send "Hi " as literal text
• Write email copy that sounds robotic or off-brand
• Miss a suppression source and email existing customers
• Set incorrect sending schedules that overwhelm new email accounts

Human review before activation is non-negotiable.

Suppression Lists Are Not Optional

If you skip Stage 3 to "move faster," you will:

• Email existing customers with cold pitches (damages trust)
• Email people who already opted out (violates CAN-SPAM)
• Hit spam traps from unverified email lists (destroys deliverability)
• Duplicate outreach from sales and marketing simultaneously (confuses prospects)

Build suppression first. Everything else second.

Volume Discipline Protects Sender Reputation

New email accounts should not send 500 emails on day one. The standard warm-up protocol:

• Week 1: 10-20 emails per day
• Week 2: 30-50 per day
• Week 3-4: Scale gradually to your target volume
• Ongoing: Monitor bounce rates, spam complaints, and reply rates

Eric used 1-2 sending accounts per campaign deliberately. Fewer accounts, warmed properly, outperform a bank of fresh domains every time. If your bounce rate exceeds 3% or your spam complaint rate exceeds 0.1%, stop sending and fix the list.

No Calendar Links in First Touch

This is a small detail that Eric called out specifically: do not put a calendar booking link in your first cold email. The first email should ask a simple question — "Are you open to a quick call?" — not demand a 30-minute time commitment.

Calendar links in cold email signal "I assume you want to talk to me." A question signals "I am asking if this is relevant to you." The second approach gets more replies because it requires less commitment.

The Stack: Enterprise vs. Local Business

Here is a side-by-side comparison of the tools for each version of this system:

The enterprise stack costs $500-2,000+/month. The local business stack costs $0-150/month. The underlying logic is identical.

Where to Start

If you are an agency or B2B company, the full pipeline is worth building. Start with one visitor identification tool (RB2B or Leadfeeder), connect it to your CRM, set up suppression rules, and build your first outreach sequence. Do not try to build all five stages at once. Get Stage 1 and Stage 5 working first (identification + CRM), then layer in enrichment, suppression, and outreach.

If you are a local service business, skip visitor identification entirely for now. Your first priority is making sure every lead that already contacts you gets followed up within 5 minutes — automatically. Set up a CRM. Connect it to your email tool. Build three automated sequences: new inquiry, no-show, and quote follow-up. That alone will put you ahead of 90% of your competitors.

If you want to understand which parts of your marketing can be automated and which still need a human, take our AI readiness quiz — it takes 2 minutes and gives you a personalized automation roadmap.

For the full picture on AI automation for local businesses, including the 5 workflows that save 10+ hours per week, start with our pillar guide. And if you want to compare the tools that power these automations — Make.com vs Zapier vs n8n — we break down the real costs and tradeoffs.

When you are ready to build a system like this for your business, our AI automation team handles the full setup — visitor tracking, CRM integration, email sequences, and the suppression logic that keeps everything compliant and clean.