At Sculpt on 8 October, Clay launched three things: Workflows (beta, open to all users), an agent-first way to build (early access) and an Inbound SDR Agent (preview). Only Workflows is something you can open today. Everything below comes from Clay's own recap, and I have kept to what it says.
I sorted each launch by one question: does it change what I do on Monday?
What did Clay actually launch?
Three items, each at a different stage. This is the part that matters most, because "announced" and "usable" are far apart.
| Launch | Status per Clay | What Clay says it does | Changes my week? |
|---|---|---|---|
| Clay Workflows | Beta, available to all users | A visual canvas where steps can branch, loop, run code or call an agent; you can trace individual record runs and re-run fixes in bulk | Yes, test it now |
| Agent-first builder | Available to Sculpt attendees; early access to more users in the coming weeks | An agent takes an outcome-based request and works across Clay, guided by a Knowledge Hub holding your ICP, personas and positioning | Not yet, unless you have access |
| Inbound SDR Agent | Preview, waitlist for pre-beta partners | Qualifies inbound leads by voice, then demos the product in a sandbox | No, it is inbound |
Clay also described a "self-learning revenue engine" as the direction it is building toward. The recap presents that as a preview of direction, not a product you can switch on.
Why do Workflows matter more than the other two?
Because it is the only one you can test this week, and it targets the part of Clay that gets messy. A big table with branching logic, conditional columns and agent steps becomes hard to read. A canvas with per-record traces is a different way to see the same logic.
Two details in the recap stood out to me. You can trace what happened to an individual record, and you can fix a play and re-run it in bulk. If you have ever hunted through a table to work out why one row got skipped, you know why that is useful. The recap also says workflows can call existing Clay functions, so work you have already built is not thrown away.
I am not going to tell you it is better than tables. I have not run my client work through it, so I would be guessing. I wrote a separate decision guide on Clay Workflows versus tables with the method I would use to decide.
What should you do about the agent builder?
Nothing, until you can get access and test it on a copy. Clay says it learns from campaign results and recommends changes to the Knowledge Hub. That is a claim about a system I cannot see yet.
The principle I would hold to regardless: an agent that can act across the product can also spend credits across the product. Read how to stop burning Clay credits first, and test on three rows before you widen anything. Clay's changelog also lists Workflow Budgets (5 October, beta), which lets you assign credit budgets to workflows. That is the kind of guardrail I would switch on before running anything unattended.
Does the Inbound SDR Agent change outbound?
No, and it is worth saying plainly. Clay describes it as handling inbound leads: a voice conversation for qualification, then a sandbox demo. Outbound operators send first. Deliverability, list quality and signals are still the work, and none of that moved on 8 October.
What it may change is the other end of your funnel, the replies and form fills you get after a campaign works. I wrote about that side in qualify inbound leads with AI. I will judge the agent when I can see it run, not from a launch page.
Where this advice stops working
If your outbound is already running on a stable table and hitting its numbers, there is no urgency to rebuild it. Beta tools reward curiosity, not migration. If you are starting a new build from scratch this month, trying Workflows first is more reasonable.
Also, this is a read of a vendor's own announcement. It tells you what Clay says it built, not how it performs on a messy list.
Related reading
The Clay Cup final was held at the same event. I covered it in Clay Cup 2026 winner: Ben Broadley, and the thinking behind building fast is in what is a GTM engineer.
I'm Malik, a GTM engineer in Islamabad. More about my work is on my about page.