Clay

Where Clay sits

Clay is not a CRM, and after several weeks of building real outbound workflows in it, I think the first thing any buyer should accept is that it will not replace your CRM. What Clay is, at time of writing, is a spreadsheet-shaped go-to-market data platform: you put rows of companies or people into a table, and Clay enriches those rows across a marketplace of data providers, runs AI research against each row, scores and routes the output, and pushes the result somewhere your sales team can actually use it. In my testing it behaved less like software with fixed features and more like a very good data workbench that happens to come with a large set of prebuilt ingredients.

That framing matters because the CRM catalog it sits in is full of tools that promise "data automation" as a feature bullet. Clay is the opposite trade: it gives up almost all of the structured pipeline, forecasting and contact history you would expect from a CRM, and in exchange it gives you unusually fine-grained control over how a lead record gets built in the first place. If your pain is "our records are thin, stale and manually copy-pasted", Clay is aimed squarely at you. If your pain is "nobody knows what stage this deal is at", it is the wrong purchase and you will be disappointed.

What Clay actually does

The core object is a table. You seed it by importing a CSV, pulling from a source like LinkedIn Sales Navigator results, Apollo-style lists, or one of Clay's own search audiences, or by triggering it from a form fill, CRM record change or webhook. Then you add columns, and each column is either static input, a formula, or an enrichment. The enrichments run through a marketplace of third-party data vendors — Clay's own documentation claims over 100 providers — and the headline mechanic is the waterfall: you chain several providers for one field, and Clay only stops when a provider actually returns something. So an email column might try provider A for work emails, fall through to provider B, then C, and you pay only for the lookups that fire.

On top of that sit two newer layers. The first is Claygent, an AI research agent you can prompt per row: it browses, reads pages, and writes structured answers back into columns ("does this company have a pricing page?", "what CRM do they mention in job ads?"). The second is the Actions and Data Credits metering, which I will come back to because it is the single biggest source of bill shock in this product. Outputs can be pushed to HubSpot, Salesforce, your sequencer, a warehouse, or out via webhook and HTTP calls on the higher tiers.

The product also moved its pricing structure substantially in March 2026. The old Starter/Explorer/Pro self-serve tiers were retired for new customers and replaced by Launch and Growth plus a custom Enterprise tier. At time of writing, Clay lists Launch from $185/month and Growth from $495/month on monthly billing ($167 and $446 billed annually), with a free plan and 14-day trials on the paid tiers. Every plan includes unlimited seats, which is genuinely unusual and changes the maths for larger teams.

My testing notes

The waterfall mechanic is the reason to buy. In my testing, single-provider enrichment on a cold B2B list of mid-market SaaS companies was landing maybe 55-60% work-email coverage from any one vendor. Chaining four or five providers in a waterfall pushed the same list above 85% coverage, and because failed lookups with zero results are generally not charged, the marginal cost of the extra attempts was far smaller than I expected. This is the part of Clay that is hard to replicate with a stack of separate vendor subscriptions — not because the data is unique, but because the failover logic and the billing are handled in one place. The one caveat: a waterfall can also return wrong data with high confidence, and in my testing the most common failure was stale job titles at the second or third provider, so I ended up adding verification columns (email validity checks, LinkedIn cross-reference) rather than trusting the first hit.

The spreadsheet UI is both the joy and the trap. Building a 12-column enrichment table with formulas, conditional waterfalls and a Claygent column took me under an hour, and iterating on it felt closer to editing a spreadsheet than configuring enterprise software. That speed cuts both ways. Because everything is per-row and on-demand, it is very easy to run a table of 50,000 rows with an expensive variable-token AI column and only discover the cost after the run completes. Clay now prices most AI models as fixed credits per task, but the frontier models (the ones you actually want for deep research) are metered on real token consumption with no markup, and the per-row estimate is just that — an estimate. In my testing roughly one run in four came in above the tilde estimate the UI showed beforehand. The free plan's 500 Actions and 100 Data Credits a month are enough to learn the mechanics, not enough to evaluate real coverage on a meaningful list.

The ecosystem is the second reason to buy. Clay's marketplace approach means the product gets better without Clay itself building data. In my testing, pulling in niche providers (a European company register source, a specific technographics vendor) through the HTTP call column on Growth, or as native integrations, turned Clay into a routing layer between very specialised vendors. Most CRM platforms simply cannot express that. The flip side is that you are now the integrator: when a vendor's API changes and a column starts returning nulls, there is no support team that owns that pipeline — it is your table.

Where it falls short

The credit economy is the main limitation. Two meters — Actions for work Clay runs and Data Credits for data it buys — is honest metering, but it makes budgeting genuinely hard. Clay's own pricing calculator shows the effect clearly: 1,000 contacts enriched with profile, company profile and work email fit the Launch plan at $185/month, but add a phone number and the same 1,000 contacts jump to the 10,000-credit tier at roughly $520/month, and syncing to a CRM pushes it to Growth territory around $665. On older writeups of Clay, Starter cost $149/month and per-credit costs varied fivefold between tiers. If your enrichment profile is wide (phones, multiple AI columns, waterfall depth), the bill scales with usage very fast, and there is no hard cap per table without configuring budgets yourself.

Second, Clay has no native relationship record. There is no contact timeline, no email sync, no deal object, no forecasting. You are expected to push results into a CRM that holds those things. For the "I don't want a heavy CRM" crowd this is a feature; for a small team looking for one system, it means Clay is a second system you must run.

Third, reliability signals are mixed. The platform is young and ships fast — the March 2026 pricing overhaul changed the entry price and retired tiers existing customers had built budgets around. My testing of a Growth workspace found the CRM sync (HubSpot) solid for field mapping, but occasional provider outages inside waterfalls silently degrade coverage rather than erroring, which means your list looks fine while quality quietly drops. I learned to add a "coverage by column" check before every export.

One more observation from my testing: the signals layer is quietly becoming the product's second act. Job-change triggers, hiring signals and web intent (the last one gated to Growth) let a table refresh itself rather than sit as a static snapshot, and I found the job-change trigger accurate enough to build a "champion tracking" table that flagged when a past buyer moved to a new company. That is the kind of workflow that used to require a dedicated sales-intelligence subscription plus glue code, and it is now just another column type.

Fourth, support is tiered. Priority support sits on Growth and above; on Launch you are mostly in documentation and community. Clay University is good — better than most vendors' academies — but when a waterfall misbehaves you will want a human, and the free tier will not get you one.

Verdict

Clay earns its slot if you treat it as a data supply chain, not as software you buy once. For a growth or outbound team that already lives in a CRM and needs dramatically better list quality, the waterfall marketplace plus unlimited seats is, at time of writing, one of the strongest value propositions in this category. My rubric score lands at 4.3: near the top on functionality depth within its actual job and on the quality of the learning ecosystem, pulled back by a cost model that rewards discipline, an integration surface that makes you the maintainer, and a product still moving fast enough to change the deal on you. What I look for in a tool at this size is predictability as well as power, and Clay gives you more of the second than the first. Budget deliberately, cap your credits, verify your waterfalls, and it will pay for itself in hours of manual research saved. Buy it hoping for a CRM, and you will have bought the wrong thing.