Anthropic's GTM hire is an actual engineer

Hi there. Ramp engineer Sandra Yang gave coding agent Devin a company card, named it BugBasher, and told it to make money.

Four weeks later: about $3,000 spent, $75 earned, and thousands of calls. She called it the first public cold-start B2B sales case for an agent.

The agent was great at chasing a concrete goal. It was useless when the destination was vague. Peak 2026.

In today’s Circuit:

🛠️ Operator Tip of the Week
📖 Anthropic's GTM Claudification hire
🗳️ Poll: what is your next GTM hire?
📡 Across The Wire
🧪 Worth Trying This Week

Before you ask anyone to "build an agent for outbound," write the human layer first. Who approves what the agent sends, where it hands off to a person, and who owns the miss when it gets something wrong.

Put those three lines in a doc, then scope the agent. Most GTM agent projects fail after the excitement wears off. A great demo, and nobody accountable the moment it touches a real CRM record.

Everyone in GTM Slack is passing around the same Anthropic posting this week. The live Greenhouse title is Staff Software Engineer, GTM AI Engineering, though the same link keeps getting shared under an earlier name, AI Engineer, GTM Claudification.

The comp band on the page is $320,000 to $405,000, and the team is literally called GTM Claudification. So the seat that owns Anthropic's go-to-market is an engineering seat, and they're paying staff-engineer money for it.

Read the responsibilities and it's clear what they actually want.

Build autonomous agents that run GTM motions end to end, across inbound, outbound, pipeline, and customer engagement.

Design the human oversight for each one, the approval gates and handoffs and escalation paths that keep sellers in control.

Ship eval frameworks in development and in production, instrument every model and tool call, and tie what the agent does back to pipeline and revenue.

Wire up MCP servers and skills that connect the CRM, the comms tools, and the warehouse.

The preferred profile is blunt about it: eight-plus years as a software, ML, or forward-deployed engineer, fluent in production LLMs and agents, ideally with real GTM systems behind them. This is an engineer whose product happens to be revenue.

Adam Schoenfeld broke the same posting down on LinkedIn and landed on three beats: an agent builder with revenue goals, real engineering sitting inside GTM, and a heavy dose of human enablement.

He points out the same shape popping up elsewhere too, at Gusto, Customer.io, and Vanta. The titles bounce around (AI Engineer, Applied AI, GTM Engineer, Marketing Engineer) but the job underneath is the same: builders embedded in sales and marketing orgs instead of a separate tooling team.

The feed is already treating this as the template for every revenue org. For most teams it won't be.

If you're a big company with the budget for a half-million-dollar hire to build these systems in-house, that seat can pay for itself. But if what you really need is someone semi-technical who can connect the tools, work alongside the team, and ship the ops and content glue, you're still hiring a GTM engineer, not a staff AI engineer.

Different companies, different jobs. One Anthropic posting doesn't mean every revenue team needs a staff engineer on the GTM payroll by Monday.

n8n is a workflow automation tool you can self-host, and it's a sane place to try the entry-level version of what that JD describes without hiring for it.

It fits the sotry because you don't have to reproduce the whole staff-engineer job to get value; you can wire one narrow motion, connect the CRM, call a model, drop a human approval step in the middle, and log what happened.

There are limits of course. This gets you the simple, well-scoped version, not the evals, observability, and production judgment a role like Anthropic's is actually paying for. When a motion gets genuinely complex, the plumbing stops being the hard part and the engineering starts.

🔗 Anthropic GTM Claudification JD - the $320k-$405k posting

🔗 What everyone is missing about Claudeforce - The Signal Club

🔗 Vanta Sr. AI GTM Engineer - Linkedin

🔗 How Anthropic uses Claude for GTM Engineering - Linkedin

- Pick one motion (inbound triage, outbound research, pipeline hygiene) and write its approval gate and escalation owner before you write a single agent prompt.

- Build one CRM-connected skill read-only first, and run it against the same segment as a plain chat research loop to see if the connection earns its keep.

- Write a short eval checklist for one outbound research agent: seed scenarios, fail cases, and how often a human spot-checks it.

- Audit your own "GTM engineer" job post: does it ask for production LLM and agent work, or just tool admin?