If you only check one highlight this week:
B2B CMO? How to easily build agents that save your team a ton of time, for instance, to generate a periodic product newsletter
Everyone else: what matters today in prompting
-François

The three winners of our AI workflow fest
Before the holiday break, James Raybould and I hosted an AI workflow demo fest.
We did a “call for workflows” for Product Marketing or Product Management.
I handpicked six presenters. The ~70 attendees of our virtual event voted.
Every attendee was impressed and stayed on for an extra 20 minutes to go deeper. It was super fun.
Below are the winners and their demos.
Each is only 6 minutes long. Highly recommend as inspiration.
🏅Winner: Jacob Bank - How to build your product newsletter agent with Relay.app.
In only 6 minutes, Jacob Bank, CEO and founder at Relay, built an agent - live! - that creates a monthly newsletter.
His agent will:
Trigger on the first of every month.
Retrieve all recently launched integrations from Linear, filtering for issues where the completion date is after the current date minus one month.
Use AI to turn that information into an email newsletter draft.
Output the draft as a Google Document in the email marketing folder.
What a time saver.
🥈 Silver medalist: James Raybould — Tailored Content Feeds with Claude Code
James built custom, curated content feeds for Spotify, YouTube, and news.
Uses Claude Code to connect APIs (Spotify, YouTube, news) and custom prompts to define content. A two-prompt process for the Spotify example took only one minute after the initial setup.
The result is a "completely custom" and "tailored" content diet optimized for user interests (e.g., product management, AI, tennis) rather than for ads or general algorithms.
🥉 Bronze medalist: Matias - Sales Call Intelligence Agent
Matias built a sales intelligence agent that analyzes sales call recordings, turning hours of them into instant, actionable insights.
He used: Cursor, n8n for the backend workflows, and Supabase.
The insights are presented in a consumable format and can be exported to tools like Linear to create specific tickets for the product team. The AI can detect things that humans miss
My selection of tips, news and workflows
✨ Vibecoding Inspiration: a CMO built this site himself with AI
Check out the latest version of the Augment Code website.
From copy to design to subtle animations: all done by Sylvain, the CMO.
In two days.
By himself.
Using Claude Code (including skills and subagents), Augment Code, Sanity (CMS), and Vercel (hosting).
Yes, that means no designer, no webdev, no copywriter. But advanced AI coding skills, great context, very strong product marketing work, and a deep understanding of programming.

Augment Code’s Homepage
📺 How to tap creators and influencers? Give them autonomy
Good case study by Tanay, the CEO at Wispr Flow, on how they unleashed creators.
Their guiding principles:
Give creators real autonomy (or they'll leave)
Build a viral replication system (see below)
Get extremely specific with your hooks
Be ruthlessly selective
Their viral replication system:
Step 1: Monitor 70 creators making content daily in real-time
Step 2: Identify videos hitting 1M+ views in the first day
Step 3: Extract that exact script
Step 4: Send script to every creator simultaneously
Step 5: One viral video becomes 70 viral videos
📄 Some prompting skills that still matter today
As Kieran explains, the models are now so good at crafting effective prompts that what matters more is providing the LLM with the right context and examples, rather than writing elaborate prompts.
He shares these good tips in his newsletter. I recommend reading the full thing.
Summary:
“1. Show, Don't Tell - Always find examples that work, paste them in, and ask for more like that.
2. Feed Real Data, Get Real Insights - Spend more time gathering your data vs. constructing your prompt.
3. Ask Open-Ended, Not Yes/No - You can also do external research to gather that context.
4. Make It Edit Itself - I also ask the AI to tell me what I haven't thought of, how it would make something better. Gold. The Self-Critique Loop:
Step 1: Generate first draft (write cold email, copy, or content)
Step 2: Ask the model to critique its own work
Step 3: Revise based on critique
Step 4: Critique again
Step 5: Final version
5. Store Your Context in Memory - You'll find that the output of your prompts becomes increasingly tailored for you.” - Kieran Flanagan
🤖 From gems and GPTs to skills and sub-agents
I fully agree with Professor Mollick. I've even started to transfer my GPTs (OpenAI) into skills (in Claude).
“A lot of companies invested substantial effort into GPTs. But GPTs are also an obvious dead end, likely to be replaced with subagents & skills, which have a lot of new advantages and are actually easier for non-technical people to create.
Instead of being individual prompts that users invoke, agents can use skills directly as needed, the user doesn't need to decide when to use GPTs, or chain GPTs together. Plus, skills can be more easily be centralized & managed than GPTs, which is much more useful at the organizational level.
Fortunately, it isn't that hard to make GPTs into skills (they are all text prompts after all), OpenAI could help with tools to make the conversion easier, but seems to have mostly ignored GPTs for a while.”
🕺Speaking of skills: marketing skills to add to your Claude account - FREE!!
This is for premium Claude users, knowing that it's coming soon to Gemini and ChatGPT (you can already add skills to the Gemini CLI today):
I thought he would charge for that, but he's giving them away!
Guillaume shares 43 skills you can download to supercharge Claude:
Professor Mollick, above 👆, and I recently wrote about skills.
As Guillaume reminds us:
“Claude Skills are persistent instruction files (.md) that you upload to Claude's memory. Unlike one-off prompts that you have to paste every time, Skills stay loaded across all your conversations.
When you upload a Skill, Claude gains permanent access to:
Expert-level frameworks and methodologies
Specific output structures that match industry standards
Quality rules that eliminate fluff and filler
Context auto-detection (B2B vs. B2C vs. D2C)
Think of it as giving Claude a new area of expertise; it can draw on anytime you need it.”
Below is a selection. I have not tested them yet, but I have seen the output of his prompts, and it's good :-)
Strategy (12 Skills) - Examples:
Buyer Persona: 400+ line personas with 10 psychological sections
Market Positioning - Strategic canvas with competitive alternatives, SWOT, and differentiation framework
Website Messaging - 1- 3-section messaging framework covering every page type
Advertising (5 Skills) - Example:
Ad Copy 15+ copy variations using psychological triggers
Content & SEO (8 Skills) - Examples:
SEO Strategy: Complete SEO plan with keyword research and link building
LLM Optimization Visibility strategies for AI search (ChatGPT, Perplexity)
Copy (5 Skills) - Example:
Headlines 20+ variations using PAS, AIDA, How-To, and Number formulas
Website (3 Skills) - Example:
Landing Page Complete page copy using StoryBrand, PAS, and urgency frameworks
Sales & Email (4 Skills) - Example:
Sales Script Discovery, demo, objection handling, and closing scripts
Video & PR (4 Skills) - Example:
Media Angles 5-10 story angles for journalist outreach
Research (2 Skills)
Competitor Analysis Deep-dive with SWOT, positioning audit, and battle cards
Competitive Intel Landscape comparison with differentiation matrix

Thanks, Guillaume!! 🙏
Final Words
The reality is, in a world of code AGI, a world of functional AGI, the org chart is broken. Bottlenecks shift from who can code to who has good ideas. The role of management shifts from resource allocation to taste and judgment.
Competitive advantage shifts from execution capability to speed of iteration. And the gap increasingly isn’t linear. It’s compounding.
Thanks for sharing these highlights with busy marketing execs around you.🙏
Someone forwarded you this email? You can subscribe here.
François | LinkedIn
I'm a CMO, advisor, and "CMO Wingman". Yes, that's a thing :-). Ask my clients: in this AI era, CMOs need a strategic proactive advisor more than ever. I’m former CMO at Twilio, Augment Code, Apollo GraphQL, Decibel, Udacity and Head of Marketing for LinkedIn Talent Solutions.
