Hey, curious people.
Yesterday I was at Sculpt, Clay’s GTM conference in San Francisco. I spoke, and I linked to this newsletter from my talk, so a lot of you are likely new here. Welcome.
This is a slightly special edition, mostly about Clay and what happened at the event.
Disclosure: I spoke at Sculpt, and my team uses Clay. What I checked: I build Workflows through Clay’s CLI myself and have been for a few months. I feel behind on using Clay’s agents inside the product, so those launch details are Clay’s description, not my testing.
On the menu this week:
🖥️ Clay announced three big things at Sculpt. All three point at agents.
🛠️ The play: build Clay Workflows through the CLI. An hour of table building became minutes for me.
🎤 The best talk I saw, from Jen Igartua at Go Nimbly.
📡 On the radar: a year-end deadline for legacy Clay plans.
01 / THE BIG MOVE
🖥️ Clay wants an agent to run your plays
Clay used Sculpt to preview what it calls a self-learning revenue engine, with three launches behind it. Clay’s recap.
The number that matters: 2,000. That’s how many people Clay says came, from 30 countries. For a conference about building go-to-market workflows, that’s a lot of people.
What shipped:
An agent-first Clay: you give it an outcome, and an agent plans the play, runs it and suggests changes. Sculpt attendees have it now, with early access rolling out in the coming weeks.
Workflows: a canvas where a play can branch, loop, run code or call an agent. It’s in beta for all Clay users and we love this feature.
An Inbound SDR Agent: it qualifies inbound leads in a voice conversation, then demos your product in a sandbox. This is a preview with a waitlist, so there’s nothing to buy yet. Matthew (the PM) sold us Clay, so I’m always excited about what he’s working on.
Why it matters for GTM: We’re particularly excited about two things. One is the Inbound SDR Agent. The other is what Clay’s CEO, Kareem Amin, said at the start: a go-to-market motion that’s continuously learning and self-improving.
Our vision is to build at least the rep experience for this internally. I can totally see that vision including Clay plus some internal build for us. For teams without ML engineering resources, it could be fully Clay.
02 / PUT IT TO WORK
🛠️ The play: build Clay Workflows through the CLI
Two Clay things are easy to blur, so here’s the split. The CLI is a way to work with Clay: a command-line tool that lets a coding agent like Codex build for you. Workflows is a separate product: the canvas where a play branches, loops and runs.
Out of the recent announcements, building Workflows through the CLI is what we’ve been leaning on the most.
A traditional table build used to take me upwards of an hour. Now I scope it in Codex in a couple of minutes, let it run, and test it nearly autonomously. That’s my experience, not a benchmark, but it’s saving me a ton of time.
THE PLAY AT A GLANCE
Good for: anyone who builds Clay tables and already works in a coding agent.
You’ll need: a Clay account with Workflows, Clay’s CLI, and a coding agent such as Codex, Claude Code or Cursor.
Setup: Clay’s CLI and API page is the place to start.
Status: I use this.
How it runs:
Scope the build in your coding agent. Say what goes in, what gets enriched and where the output lands.
Let the agent build the Workflow through Clay’s CLI.
Run it on a few records before the full list.
Have the agent test the output, then check a sample yourself.
TRY THIS ON ONE TASK
Rebuild one table as a Workflow w/ Clay’s CI.
Pick a table you already run and trust. Rebuild it in Workflows using Clay’s CLI and Codex/Claude Code, then compare what comes back with the original.
Pick something small first, like those portion of accounts that are missing some key enrichment details. Then watch yourself vibe code the details into existence.
03 / FROM THE STAGE
🎤 The best talk wasn’t mine
I spoke about the GTM AI adoption curve: four stages a go-to-market team moves through, from individual use to custom tools. I’m hoping everyone took something from it.
The curve from my talk. The long version is coming as its own post.
Beyond my own session, the Queen of RevOps gave by far and away my favorite presentation. That’s Jen Igartua, CEO of Go Nimbly, with The Changing DNA of RevOps. Her notes and deck are here.
Her argument is that the teams getting this right are changing three things at once:
The roadmap: RevOps takes on more GTM workflows and ships real product experiences.
How they build: centralized, agentic and code-driven, with a person approving every change.
The team: new technical and product roles, and a product mindset to match.
One number she cites, from Retool: 96% of organizations use agents, and only 12% have a central way to manage them (which should resonate with my talk on AI Adoption)
Where I land: I really align with Jen’s thinking. People consistently ask me whether they should have their own GTM AI team inside RevOps. My answer: maybe, but definitely only temporarily.
Everyone in RevOps is going to need to be AI-native quickly. That means field ops, systems and enablement. It’s where I expect my own career to go too: leading global RevOps organizations where everyone is an AI expert.
Borrow it: Jen’s homework is to count the agents your team runs and note where each one lives. If any exist only on one person’s laptop, move those into a shared repo first.
04 / PIPELINE SIGNALS
📡 On the radar
Clay is letting legacy Starter, Explorer and Pro customers use Workflows, the CLI, Audiences, Account Agents and Sequencer through the end of the year. So what: if you’re on a legacy plan, test the workflow play above before the window closes.
Clay’s Skills Marketplace opened with community-built skills that are free to install in Claude Code, Codex or Cursor. So what: look there before you write a Clay skill from scratch.
Clay inbound had its agents fill out demo forms at 6,346 companies, and says most never wrote back. So what: fill out your own demo form this week and time the reply.
OVER TO YOU
💬 Should GTM AI be its own team?
You have my answer: maybe, but only temporarily. What’s yours?
How was this week’s issue?
Robert
P.S. Know someone trying to put AI to work in go-to-market? Forward this to them. They can subscribe at bizgrowth.ai.