How a Zap replaced my morning AI news routine
Hi all,
Confession: I haven't read an AI news article on its own website in months. Not the WSJ piece on the energy crunch. Not Ethan Mollick's latest. Not the OpenAI launch posts.
And somehow I'm catching more, not less.
Here's how.
Three Bots. Three Channels. One Brain.
About six months ago, I built three Zapier workflows that each feed a different Slack channel:
- #ai-news pulls headlines from Google News and a handful of feeds I trust
- #ai-research scrapes new academic papers and industry reports
- #ai-products watches for new AI tool launches and feature drops — so I always know what to demo on stage
Each one runs on the exact same three-step pattern: Zapier catches the article, hands it to an LLM with a prompt I wrote, and the result lands in Slack — formatted, skimmable, searchable.
The prompt is the secret sauce. I tell the model: summarize the piece, score how relevant it is to my business on a scale of 1–10, and surface the angle I'd take if I decided to write about it. Same prompt, every article, every time. The relevance score is what makes the whole thing actually usable — I skim the 10s, dip into the 7s, and ignore the rest.
Important caveat: the drafts are prompts for me to think with, not text I ship. Almost nothing makes it to LinkedIn the way the bot wrote it.
The Move That Changed Everything
Here's the thing I didn't expect. The first version of these bots worked — but the output was generic. Useful summaries, but they sounded like summaries. They didn't sound like me, and they didn't filter for the things I actually care about.
So I tried something. I fed the LLM a copy of my book, Future Proof, as context.
The difference was striking. Suddenly the bot wasn't just summarizing — it was filtering through the lens of my own frameworks.
It started flagging articles about "AI fluency" and "frontier professionals" and "the toolset-mindset-skillset triangle" because those are the ideas in the book. It surfaced angles in my voice — short sentences, a little dry, no hype. It became less like a research assistant and more like a colleague who'd read everything I've ever written and knew exactly what I'd find useful.
This is the bigger pattern I keep seeing: the AI in your stack is only as good as the context you give it. Your IP, your frameworks, your voice, your client history — that's what turns a generic LLM into something that actually works for you.
Why I'm Telling You This
I'm sharing this for two reasons.
First, because if "I can't keep up with AI" is the thing keeping you up at night, this is a solvable problem. You don't need a data team. You don't need a six-figure project. You need a Zapier subscription and a clear prompt. 20 minutes, no code, runs while you sleep.
Second, because this is exactly the architecture I'm spending more and more of my time building with enterprise clients — only at scale, and wired into their actual tools (CRM, ticketing, comms, whatever). Zapier's MCP layer makes that wiring radically easier than it used to be, and it's the closest thing I've seen to a "default plumbing layer" for the agentic stack.
If you want the Zap template I'm using, hit reply — happy to send it over. And if you want to talk about what this looks like for your team at scale, that's what the AI Performance Lab is for.
Either way, you don't have to read every article. You just have to build the bot that does.
Best,
Dr. Michael "House" Housman
P.S. Full disclosure: I'm collaborating with Zapier on a sponsored LinkedIn post about this, which is what kicked off this newsletter. They're not paying for this email — but they did build the rails this whole thing runs on, and I built it well before that conversation started.