Agentic Layer Check
A rare move: Clay priced its new Growth plan nearly 40 percent below the legacy Pro plan it replaces, published the internal memo explaining why rather than burying the change in a footnote, and split what a customer pays for into two separate, legible meters. Here's how the new pricing reads against the nine layers.
What they sell
A GTM automation platform for building outbound and enrichment workflows: dozens of data provider waterfalls, an AI research agent called Claygent, and automations that write out to a CRM, ad platform, or outbound sequencer.
Who they sell to
Sales, marketing and RevOps teams running their own outbound and lead scoring, self serve from Free through Growth, sales assisted for Enterprise, plus agencies running the platform on behalf of clients.
How they price
Two decoupled meters, Actions (workflow orchestration) and Data Credits (enrichment data and AI usage), across Free, Launch (from $185/mo billed monthly), Growth (from $495/mo, replacing an $800/mo legacy Pro plan), and custom Enterprise, under the pricing model Clay introduced in March 2026. Existing customers keep legacy plans by default unless they switch.
Two moves here are worth other vendors studying, not just noting.
The new model splits Actions (what it costs to run a workflow step: an enrichment call, an AI research step, a webhook) from Data Credits (what the underlying data or AI provider actually charges). Clay's own externally published memo states the logic plainly: "Customers don't come to Clay to buy inputs. They come to run outbound, score leads, and automate their GTM." That's Layer 3 done properly: the meter now maps to two real, separable things a customer is paying for, instead of one blended "credit" that hid whether a given month's bill moved because of more workflow volume or because a workspace happened to call a pricier AI model. It also lowered the price of admission: HTTP API access, CRM auto sync, web intent signals and a new Ads product, all previously gated behind the $800/mo Pro plan, now sit on the $495/mo Growth plan.
For the token heavy reasoning models Clay prices variably (its pricing page names GPT-5.1 and Claude 4.6 Sonnet), the company now bills "exactly for the tokens consumed with no markup", and the model picker shows those choices as an estimate rather than a flat per row cost, so a customer can see which selections will vary before they run them. Most AI native tools blend model cost into one opaque credit rate; Clay chose to separate its own margin from a pure cost passthrough and say so in writing, alongside cutting marketplace data costs by 50 to 90% and lowering the one off top up premium from 50% to 30%. That's an unusually legible metering choice for a category where "what does one credit actually buy me" is rarely answerable.
The opportunity isn't in the split itself: it's in what one balance now has to represent.
Around 80% of AI models in Clay still cost a flat number of Data Credits per task, a rate Clay sets; only the variably priced reasoning models pass through at cost, and both draw from the same Data Credits balance. That means "Data Credits remaining," the one number a customer watches, no longer maps to a stable dollar figure of Clay's own revenue: a credit spent on a flat priced task carries whatever margin Clay builds into that rate, while a credit spent on a variably priced reasoning model carries none on the tokens, and which one a workspace draws down depends entirely on which models its workflows happen to call that month. Clay's own memo is candid about the trade off, describing the change as "intentionally revenue- and profit-negative in the near term." That candour is worth crediting, but it also means blended gross margin per workspace is now a function of AI model mix rather than of anything Clay sets directly.
That margin blending doesn't stay a modelling footnote. It lands on Commercial Systems (Layer 9): whatever system produces the invoice and the internal margin report now has to itemise, inside a single "Data Credits" line, which portion of a given month was a marked up enrichment call and which was a 0% margin token passthrough, for every workspace, every month, just to report an accurate gross margin figure at all. Get that itemisation wrong, or simply not build it, and Clay is left estimating its own margin rather than knowing it, at exactly the moment it has told the market a real number about accepting a revenue hit for the change. It's the same pattern seen elsewhere in this series: the finding that reads as a pricing story on the page shows up as a systems and forecasting problem one or two layers over, not on the page where the change was announced.
Part of the Agentic Layer Check series: named companies, read against the nine layers. Want your own read? Request a teardown →
This is a public signal read: Clay's own pricing page, its plans and billing documentation, its own announcement of the new pricing model, and its externally published internal pricing memo, not a private briefing or company input. It reflects what's publicly documented as of the date above; pricing pages and credit terms change, and this piece will note it if Clay's does. Clay was not consulted on or paid for inclusion; see the disclosure approach this work is run under.
Sources: Clay: pricing → · Clay: plans & billing docs → · Clay: new pricing model announcement → · Clay: pricing memo →