Clay Tutorial: Build Your First Enrichment Table in 30 Minutes

This Clay tutorial shows you how to build a working Clay table in 30 minutes: source a list of companies, find the right people, get verified work emails through a waterfall, add one AI research column, and export. Every step below comes from Clay's own documentation, with the credit cost noted where Clay publishes it. Use the free plan to learn, but expect its 200-row limit per table to cap you quickly [1].

What do you need before you start?

You need a Clay account, a clear target, and about 100 Data Credits to spend. The free plan includes 100 Data Credits, 500 Actions a month and up to 200 rows per table; Launch raises that to 50,000 rows per table [1]. Prices are as of October 2026; see our Clay pricing breakdown for the full plan comparison.

Write your target down in one line before you open Clay. For this tutorial: "Heads of marketing at US B2B software companies with 50 to 200 employees." A vague target is the most expensive mistake in Clay, because every extra row costs credits.

Learn the two meters first. Actions measure platform work: each enrichment, AI run or exported record costs 1 Action. Data Credits pay for third-party data and AI, and Clay says a fully enriched record typically costs 6 to 20 Data Credits [2]. Sourcing lists, CRM imports, formulas and filters cost no Actions [2].

Step 1: How do you get companies into a table?

Start with the Find Companies source, or import a CSV if you already have an account list. In a workbook, click + Add, search for Find Companies, and set your filters [3].

The useful filters for our target are:

1. Industries: software.
2. Location: United States.
3. Estimated employee count: min 50, max 200.
4. Semantic company description: "B2B software company selling to mid-market teams," which ranks results by similarity [3].
5. Technographics, if you only want companies running a given tool, such as HubSpot [3].

Set Limit results low for your first run. The default and maximum is 10,000 [3]. Then click Save, choose Save to new table, and turn off auto-update for now [3].

If you have your own list, click + Create new, search for CSV, upload, and map columns. Choose "Save and don't run" so no enrichments fire before you are ready [4]. Webhooks and CRM imports such as HubSpot also work as sources [2][9].

Step 2: How do you find the right people at each company?

Add a Find People step scoped to your companies. The Find People source filters by job title, function, seniority, location and experience, and can search inside companies from an existing Clay table [5].

Pick the job title match mode on purpose. "Is similar to" includes synonyms and translated titles, so "Software Developer" returns "Frontend Engineer." "Contains" requires one of your terms. "Is exactly" is case-sensitive [5]. For a marketing leader, "contains" with "marketing" plus a seniority filter is usually tighter than "is similar to."

Set Limit per company to 1 or 2. Clay allows up to 100 people per company and up to 50,000 records per search [5], which is far more than a first table needs. Sourcing people costs no Actions [2].

Step 3: How does a waterfall find work emails?

A waterfall asks several email providers in order and stops at the first valid result. Clay's Work Email waterfall charges credits only for the provider that finds a match [6]. Click Add enrichment, search for Work Email, choose Quick setup, map first name, last name and company domain, and save [6].

Turn on two settings before you run it:

1. Infer Email. This free first step guesses first.last@domain.com and stops if it validates. Clay's internal test on a software-industry dataset found that pattern valid about 31% of the time [6].
2. Validation strategy. Pick Conservative for cold outreach, since it accepts only verified emails. Balanced adds catch-alls; Aggressive casts a wider net [6].

Clay's own example shows the cost. Enriching 100 contacts with profiles and emails used about 95 Data Credits on the marketplace, or 50 if you bring your own email provider's API key [2]. Catch-all results are deliverable on paper but may have no real inbox behind them, so treat them as lower confidence [7].

If you want the argument for and against waterfalls as a model, read our piece on waterfall enrichment versus browser intelligence.

Step 4: How do you add an AI research column?

Add a Use AI column and describe the job in plain English. Clay's Generate tab writes the prompt, recommends a model and sets output fields for you; the Configure tab lets you change each setting [8]. Choose the "Web research (Claygent)" use case when the answer lives on a website, and "Create or modify content" for writing or classifying data you already have [8].

A good first prompt: "Visit the company website. Return yes or no: does the company sell to other businesses? Return the URL you used as evidence." Ask for a source URL in every research prompt, so you can audit the answer later. This tutorial adds the AI column after the waterfall to teach one feature at a time; in a real table, place it before the waterfall, as Step 5 explains.

Pick the model by cost. Clay's own Helium model costs 1 Data Credit per web research run and Argon costs 3. GPT-5 Mini costs 0.4 for content generation and 1 for web research. Advanced models such as Claude 4.5 Sonnet use variable pricing based on real token cost, with no markup [10]. Every AI prompt also counts as 1 Action [10].

Step 5: How do you keep a table from burning credits?

Turn off Auto-run while you build, test on 10 rows, and gate expensive columns with conditions. These three habits come straight from Clay's credit-saving guide [11].

1. Auto-run. Toggle it off in Run settings or the table toolbar; the button reads Manual when it is off [11].
2. Test runs. Run a new column on 10 rows, check the output, then run the rest [11].
3. Only run if. In Run settings, add a run condition such as {{email}} is not empty, or {{lead_score}} > 80 [12]. Run the email waterfall only on rows where the AI column said "yes."

Put cheap filters before expensive enrichments. Your AI qualifier should run before the email waterfall, not after it.

Step 6: How do you export the results?

Send rows to your CRM or sequencer, or download a CSV. Each exported record costs 1 Action, including records sent to an external sequencer [2]. Clay's HubSpot integration can create, look up and update objects, with fields pulled from your own portal [9]. For a file, open the Actions menu, choose Export, then Download CSV; Clay suggests saving a dedicated view for exports so you only download the columns you need [14].

Add a run condition to the export step as well. Clay's own example is "add a contact only if it has a valid email" [12]. For CRM pushes, Clay's course pattern is a lookup by email first, then Create or Update Contact only where no match exists, tested on 10 rows [15].

What does the finished table cost to run?

A 100-row version of this table costs roughly 150 Data Credits and up to 300 Actions, using Clay's published rates. That is more than the free plan's 100 Data Credits, so run your first pass on 30 to 50 rows. The estimate below assumes every row reaches every column; conditions and failed lookups lower it, because Clay does not charge for empty results [1].

Column	Source or enrichment	Actions per row	Data Credits per row
Company, domain	Find Companies	0	0
Person, title	Find People	0	0
AI qualifier	Use AI, Claygent with Clay Helium	1	1
Work email	Work Email waterfall	1	about 0.5 per email found
Email status	Validation step	0	0 with select free validators
Export	CRM, sequencer or CSV	1 per record	0

The figures come from Clay's docs: sourcing is free, each enrichment or export costs 1 Action, a found email costs about 0.5 Data Credits in Clay's own example, select validators are free, and Helium costs 1 Data Credit per web research run [2][10]. Swap Helium for a variable-priced model and the AI column becomes the biggest line on your bill.

Put the AI qualifier before the waterfall and gate the waterfall with "Only run if" the qualifier says yes. If half your companies fail the qualifier, you halve your email spend.

What mistakes do first-time Clay users make?

Most first tables waste credits in the same five ways:

1. Leaving Auto-run on while editing, so every new row enriches immediately [11].
2. Running enrichments on all rows before testing on 10 [11].
3. Using "is similar to" title matching and pulling in the wrong roles [5].
4. Accepting catch-all emails for cold outreach and taking the bounce hit [6][7].
5. Re-enriching contacts that already exist in your CRM instead of using a lookup column [11].

A sixth is choosing Clay for a job it is not built for. If your workflow is mostly logic and API calls, compare it with code-first options in our Clay vs n8n vs Claude Code guide. If you want a head start, copy a table from our list of Clay templates.

Where does browser research fill the gaps in a Clay table?

Clay is strongest at structured fields: emails, headcount, tech stack. It is weaker at context that lives behind a login or in a conversation, such as a Reddit thread describing your problem or a Sales Navigator list built from a plain-English ICP. Database snapshots also age between refreshes, which is why we argue for live sources in the waterfall piece linked above.

Drevon covers that gap from your own browser session. It reads LinkedIn, Reddit, Sales Navigator, job boards, Crunchbase and company sites, returns a brief where every claim links to its source, and exports to Clay in about 10 minutes. Drevon finds the accounts and the reasons; Clay then enriches and routes them.

Frequently asked questions

Is Clay free to learn?

Yes. The free plan includes 100 Data Credits, 500 Actions a month and up to 200 rows per table, which is enough for this tutorial [1].

How many credits does one enriched contact cost in Clay?

Clay estimates 6 to 20 Data Credits per fully enriched record, depending on data types, waterfalls and your own API keys [2]. Phone numbers cost more than emails.

What is the difference between Use AI and Claygent?

Use AI is the column; Claygent is its web research mode. Clay describes Use AI for writing and classification, and Claygent for "last-mile" research with web access [8][13].

Can I use my own API keys in Clay?

Yes, on paid plans. Your own keys skip Data Credits for that provider, but each run still uses 1 Action [2][11].

Does Clay charge if an enrichment finds nothing?

No. Clay says that if an enrichment returns no result, you are not charged Data Credits or Actions [1].

Download Drevon to research accounts in your own browser and send them to your new Clay table, or book a demo to see the export.

Sources

1. Clay, "Pricing and FAQ", accessed 2026-10-10.
2. Clay University, "Actions and Data Credits", accessed 2026-10-10.
3. Clay University, "Find Companies", accessed 2026-10-10.
4. Clay University, "CSV import", accessed 2026-10-10.
5. Clay University, "Find People", accessed 2026-10-10.
6. Clay University, "Work Email waterfall", accessed 2026-10-10.
7. Clay University, "Email verification statuses", accessed 2026-10-10.
8. Clay University, "Use AI integration", accessed 2026-10-10.
9. Clay University, "HubSpot integration", accessed 2026-10-10.
10. Clay University, "AI pricing", accessed 2026-10-10.
11. Clay University, "Ways to save Clay credits", accessed 2026-10-10.
12. Clay University, "Conditional runs", accessed 2026-10-10.
13. Clay, "Claygent", accessed 2026-10-10.
14. Clay University, "Download as a CSV (Clay 101)", accessed 2026-10-10.
15. Clay University, "Push data to your CRM (Clay 101)", accessed 2026-10-10.