Illingworth rebuilds his cold email process in Claude Code
An eight-step build that moves list pull, enrichment, validation, writing and QC into one agent, with no results attached yet.
Richard Illingworth posted an eight-step breakdown of what he says is his entire cold email process rebuilt inside Claude Code. He feeds Claude his ICP and his offer, and Claude pulls a targeted list from Apollo, enriches every contact in parallel by scraping their site and pulling their tech stack, then runs that list through a verifier before a single line of copy gets written.
The writing step uses what he calls the PPQ framework, built off the enrichment so every email references the prospect's real setup. A QC pass runs on every variant and caps each email at 80 words. From there it is a conventional send setup. He loads into Instantly, maps merge fields, caps daily volume at 15 to 20 per inbox, kills open tracking, and sends from warmed inboxes split across Google and Outlook. Step eight is the part most people skip, letting Claude read the campaign data and say which segments to scale and which to cut.
What is actually new here
Not much of the send discipline. Low per-inbox volume, no open tracking and split providers are standard hygiene. What is new is that list pull, enrichment, validation, copy and QC all sit inside one agent loop rather than across a stack of point tools. Illingworth is the second operator this week to describe running campaigns this way, after Lieben described doing the same without Clay.
The verification step before writing is the one worth reading twice. Illingworth himself has said Apollo-only lists come back 50 to 65 percent valid, which is exactly why an agent that pulls from Apollo needs a verifier bolted in front of the copy step rather than after it. Writing personalised sequences for contacts that do not exist burns tokens and, worse, burns inboxes on bounces.
What is missing
Results. There are no reply rates, no send volumes, no campaign in the post at all. This is a process description, not evidence that an agent-written sequence outperforms whatever you send today. The 80-word cap and the PPQ framework are assertions about what he does, not tested variants.
For people sending, treat this as a build to copy at the plumbing level and nothing more. The parallel enrichment and the verify-before-write order are cheap to adopt and hard to argue with. The claim that an agent should decide which segments to scale is where you want your own numbers before you hand over the wheel.
it runs a QC pass on every variant, capping each email at 80 words.