Pricey subscribers,
Right this moment, I wish to share a brand new episode with Geoff Charles.
Geoff is the CPO of Ramp and runs probably the most AI-native orgs I’ve seen. He confirmed me an incredible Claude Code PM talent that his group makes use of to go from thought to product and shared stay demos of Ramp’s inside AI brokers for buyer analysis, information evaluation, and extra. Geoff additionally shared 5 ways Ramp adopted to go from barely anybody utilizing AI to non-engineers delivery manufacturing code.
Watch now on YouTube, Apple, and Spotify.
Geoff and I talked about:
-
(00:00) “When you’re not utilizing Claude Code, you are most likely underperforming”
-
(04:45) Voice of buyer agent (demo): 8 days of analysis in 8 minutes
-
(07:57) Analyst agent (demo): Get insights and pull information utilizing English
-
(13:08) The perfect Claude Code PM talent I’ve ever seen
-
(15:20) “50% of Ramp’s code is written by AI. It’s going to most likely be 80% quickly.”
-
(16:38) Examine agent (demo): Constructing a manufacturing characteristic in 5 min
-
(28:21) The 2 instructions the PM function is splitting into
-
(34:05) Ramp’s L0-L3 framework for getting each worker to construct with AI
-
(36:36) The AI interview query Ramp asks each candidate
-
(39:15) “Administration might be lifeless…be the perfect builder on the earth.”
I’m proud to associate with Granola
Assembly notes are essentially the most helpful context for AI in any firm. With Granola MCP, now you can entry this context from Claude, ChatGPT, and any AI app:
-
Ask Cursor to create tickets for bugs mentioned in standup
-
Have Claude replace your Linear board immediately from a planning assembly
-
Get ChatGPT to draft a follow-up from a gross sales name and push it to your CRM
I actually couldn’t work with out Granola, and the MCP simply makes it extra helpful.

-
Ramp’s product shaping Claude Code talent is without doubt one of the finest PM abilities I’ve seen. This talent turns Claude right into a product pondering associate that pushes again in your concepts, runs parallel analysis, and outputs a clear spec. It really works in 3 phases:
-
Section 1: Body the issue. Claude pushes the PM with 7 questions together with “What is the job to be performed?”, “Why now?”, and “If we construct this, what does it unlock?” If you cannot defend your solutions, Claude pushes again.
-
Section 2: Analysis. Claude launches 6-10 parallel brokers to scan opponents, Gong calls, Zendesk tickets, and the codebase concurrently. Every agent writes a markdown file that Claude then synthesizes into key patterns.
-
Section 3: Form the spec. Claude converges on a design by way of dialog. The ultimate spec is a 2-min learn with context, design rules, necessities grounded in proof, options thought of, and open questions.
-
I’ve up to date the talent to take away Ramp-specific data in order that any particular person or firm can undertake it. Simply copy and paste the 5,000+ character talent within the part under immediately into Claude to set it up.
Elevate your perspective with NextTech Information, the place innovation meets perception.
Uncover the newest breakthroughs, get unique updates, and join with a world community of future-focused thinkers.
Unlock tomorrow’s tendencies in the present day: learn extra, subscribe to our e-newsletter, and turn into a part of the NextTech group at NextTech-news.com

