This guide walks through how to build that system in Clay, from defining who you are targeting through to having verified, enriched, personalized leads in your sequencer the same day.
What you need before you start
One thing: a Clay account. The free trial covers this full workflow. You do not need a separate data provider subscription, because Clay's waterfall enrichment pulls from 150+ providers. You will need a sequencer (Instantly, Smartlead, or Clay's native Sequencer) and sending domains separate from your primary domain.
Why automated outbound works, and why most teams do it wrong
The SDR at the top of the leaderboard is not working harder. They have stopped doing the things a computer can do.
How the same eight hours get used: manual SDR versus Clay automated
Manual SDR
Clay automated
Enriching, scoring, routing runs automatically, all day
Automated outbound does not replace the human judgment that closes deals. It replaces the manual research that precedes every send. The rep still picks up the phone. They just spend the day doing that, not building spreadsheets.
Step 1: Define your ICP before touching Clay
The most expensive mistake in automated outbound is enriching the wrong people. Every credit spent on a company that was never going to buy is wasted. Write down three things before opening Clay:
- Firmographic criteria: Not just industry and headcount, but tech stack signals, headcount by department, funding stage and recency, and growth signals. If your criteria describe more than 50,000 companies, narrow further.
- Contact profile: Seniority level and time in role matter more than job title alone. A VP of Sales in their first 90 days is actively evaluating tools. One who has been in the role four years probably is not. Define the seniority tier, function, and whether you want the decision-maker or the practitioner.
- One timing signal: What event makes a company likely to buy now rather than generally. Hiring for a specific role, recent funding, team growth above a threshold, a competitor tool appearing in job postings. Pick one and enrich for it specifically.
Move the dial: watch specificity unlock the opener
matching companies
Opening line
Hi [Name], I saw your company works in the sales technology space and wanted to reach out about how we help teams like yours.
Step 2: Build your source list in Clay
With a defined ICP, open a new table in Clay and choose Find Companies as your source. Set filters for industry, headcount range, revenue bracket, and funding amount. To target funding recency, add a news and fundraising signal rather than a static filter. Add technographic filters if tech stack is in your ICP criteria. Import from Apollo or Sales Navigator if you need supplementary coverage.
Start with 200 to 500 rows. Run the full workflow on this batch, review the outputs, confirm the ICP signals are catching the right companies, then scale. Fixing enrichment logic at 500 rows is fast. Fixing it at 5,000 is not.
Step 3: Run waterfall enrichment on company and contact
This is the most important structural decision in the workflow. Company enrichment and contact enrichment are both independent, and both start from the same email address. Run them simultaneously. Staggering them doubles latency for no reason.
For email, use a multi-provider waterfall. No single provider covers your full list. Clay runs each lookup through multiple providers in sequence, charging credits only for the provider that succeeds.
Email waterfall: find a verified match
The free check misses, the next provider hits, and the run stops there. One credit spent, the rest never touched. You only pay for the lookup that succeeds.
The coverage math makes the waterfall essential. No provider dominates all markets:
Work email provider benchmarks (Clay Data Tests, clay.com/data-tests)
Sorted by coverage; cost is per successful lookup.
Sort by coverage or quality and the ranking reshuffles. No single provider tops both.
Read it by job: Hunter and Findymail lead on quality, Findymail and Wiza on coverage, Enrow and Icypeas on cost. No single name wins all three, which is why the waterfall sequences them cheapest-viable-first.
Waterfall coverage typically reaches 85 to 95% versus 60 to 75% from a single provider.
Rootly automated its outbound workflows end-to-end with Clay: 10x team performance and zero manual research required. (clay.com/customers/rootly)
Step 4: Generate AI personalization snippets
AI writes one sentence per email, not the whole thing. A fully AI-written cold email is recognizable and gets deleted. A human-written template with one AI-generated first line referencing something specific from enrichment is indistinguishable from bespoke outreach.
From enrichment to opening line: generate, then swap the signal
Enrichment record
Northwind Labs
Helps enterprise revenue teams automate sales research and pipeline generation.
AI-generated opener
Click generate to draft an opener from the open role signal.
Here are the two prompts that run in Clay AI formula columns, credit-free:
Using the input, write one sentence summarizing what this company does.Input: {{company_description}}Keep under 12 words. Use specific keywords from the input.Complete with prefix: "Saw you're helping"Write casually. Do not start with "I."
Using these inputs, write one sentence inferring what problemthis company is solving by hiring for this role.Job title: {{open_role}}Company description: {{company_description}}Keep under 15 words.Complete with prefix: "Noticed you're hiring a {{open_role}},"
Test both on 20 rows before scaling. Read every output. If any look hallucinated or generic, tighten the prompt constraints.
“Clay enables insanely targeted prospecting at scale. Users can build their dream list with a fraction of the time, effort, and money.”
Step 5: Clean and normalize
Raw enriched data breaks personalization silently. "SALESFORCE, INC." in a scoring rule looking for "Salesforce" will not match. Normalize before any scoring runs.
Three normalizations, all credit-free:
Add enrichment, Tools, Normalize Company Names.No prompt needed. Strips Inc./LLC/Corp., normalizes casing.Runs automatically on every row.
Classify this job title into exactly one tier:C-Suite, VP, Director, Manager, IC (individual contributor), Unknown.Job title: {{raw_job_title}}Return only the tier label. Nothing else.
Convert this location to how a local would refer to it informally.Location: {{raw_location}}Return only the result. No explanation.Example: "Austin, Texas, United States" should become "Austin, TX"
Step 6: Set up sending infrastructure
No automated outbound works if emails land in spam. This step comes before launch, not after problems appear.
Register variant domains that redirect to your primary (tryclayapp.com redirects to clay.com). Configure SPF, DKIM, and DMARC on each. Add 2 to 3 inboxes per domain. Warm every new inbox for 3 to 4 weeks using Instantly or Smartlead warming. Cap sends at 20 to 30 per inbox per day. With 10 domains and 2 inboxes each, that is 400 to 600 sends per day while your primary domain stays untouched.
Non-negotiable
Never use your primary company domain for cold outbound. One blacklisting event on a sending domain is recoverable. On your primary, it affects every employee's email.
Step 7: Push to sequencer and launch
Once enrichment, normalization, and opener columns are complete, push to your sequencer. Clay maps enriched columns to campaign custom variables: validated email, AI subject line, personalized opener.
Before launching, check three things: validated email shows valid (not catch-all or unknown), sends are coming from a warmed variant domain, and you have manually reviewed 20 AI opener outputs. Launch only when all three pass.
Measure positive response rate, not open rate. Below 1%, something fundamental is wrong: ICP, message, or timing signal. Between 1 and 3%, you have a foundation to iterate from. Above 3%, optimize and scale. Change one variable at a time.
Results from teams running automated outbound
Outcomes teams report after automating outbound with Clay
| Company | Outcome | Story |
|---|---|---|
| Rootly | 100% automation end-to-end, 10x team performance, zero manual research | Read |
| Sendoso | $1M+ pipeline, 10x outbound productivity, 20% higher response rates | Read |
| Rippling | 2x cold email performance, 100% direct mail automation, zero engineering resources | Read |
| Terrapinn | +19% revenue per rep, -90% prospect acquisition cost, +28% email open rates on 35% more volume | Read |
| Harmonic | 2x SDR capacity, 3x response rate, 50 hours per month eliminated per rep | Read |
“Clay has allowed us to scale our outreach in ways we never thought possible. We are able to automate key processes, reach more leads, and stay incredibly efficient without sacrificing the quality of our engagement.”
The pattern repeats across very different businesses: when research stops being manual, rep capacity and response quality climb together rather than trading off.
Terrapinn lifted revenue per rep by 19% and cut prospect acquisition cost by 90%, with a 28% lift in email open rates on 35% more volume, producing 21.5 million additional opens year over year. (clay.com/customers/terrapinn)
Common failure modes
- Scoring before normalizing: Seniority routing rules fail silently on unnormalized titles. Normalize first, always.
- Single-provider enrichment: Coverage gaps on a single provider mean 25 to 40% of your list gets no outreach at all. Build the waterfall before scaling.
- Sending from your primary domain: Non-negotiable. Sending domains only.
- Not reviewing AI outputs: Run the opener prompt on 20 rows and read every output before scaling. An AI opener that reads like automation is worse than no personalization.