#fail: I wrote that this issue wouldn't focus solely on AI

#not-comprehensive: I share what matters most to me as a CMO. I don't cover everything AI-related (others already do this well). If you don’t want to miss a beat, happy to share my daily digest with you (I love it). The buzz this week was about more people impressed by Manus and the latest image editing capabilities of Google Gemini.

#feedback? Simply hit reply, text me. I love it all!

POV: When should our teams use and not use AI for customer-facing materials?

Some of you asked about CMO guidelines for AI content our teams send to customers.

In general, with exceptions, most CMOs tell their teams to edit heavily first.

Why:

• it often shows - unless you mastered your style guides and prompts

• it feels less special - if I suspect an AI wrote the content, I feel the sender didn't care enough about me to craft it themselves

• taste matters - I hope you hired your team because they are great marketers

• we are and love humans - there's a reason no one watches chess games between two AIs

• you are still responsible for the output

• AI still hallucinates

• AI is wonderful at research, idea generation, execution, and refinement, but rarely at all of it at once. For now.

So when is it OK to send something without editing first?

• when we disclose it - e.g. “You’re now chatting with our virtual assistant”

• when it’s hyper-personalized and/or the customer can feel the care that went into it - e.g., “I ran this deep research for you. Here’s my take on it {XYZ}”

• when it's obviously not mission or brand reputation-critical: e.g. an extra cold outbound email to a non-critical prospect (even then, monitor them frequently)

• when there is no legal or compliance risk

• after we have tested the flow and vetted the content long enough

• after we trust it provides value AND is on brand

Despite my custom GPTs (post summarizers), automations (digests) and Claude projects (newsletter editor), I still write 90% of the words.

1. It's my choice

2. It's out of respect for you

3. maybe I’m still not that great at a) getting Claude to write exactly like I want b) letting go of my ego ;-)

What rules did you give your team? We can crowdsource our guidance in this post’s comments. Already a great pointer from Sheila:

My selection of tips, news, and workflows

⚙️ How I built my own daily automated newsletter AI digest

I can’t keep up with all my newsletters but I don’t want to miss key news or analyses. So I built my own AI newsletter summarizer with Relay, following their CEO’s online tutorial.

It saves me a ton of time. I feel like I have a better handle on news. I've even subscribed to more newsletters. (But please don't run my newsletter through an AI summarizer! 😉)

The tutorial helps navigate Relay's occasionally clunky UX.

He shows how to:

1. flag newsletters in Gmail (I added this shortcut: when subscribing to newsletters I add “+NL” before @gmail.com - and created a filter that marks emails sent to that email address as “newsletters”)

2. create a daily summary with your favorite AI model (I keep switching to see the difference, it’s super telling - Claude Sonnet 3.7 is still my favorite)

3. email that to you back to you

4. I added a step that archives these emails once scanned

Relay could use some UX love in a few spots, as he admits, so the tutorial was helpful.

Email me if you want to receive that daily digest (I'll mention what model I used to summarize - important to understand the differences given cost variations).

💡 Consider doing that for your teams or VIP customers.

💡15 Prompting Tips from Andrej Karpathy

Some good ones in there. I didn’t practice or know these three:

1. When uploading documents for analysis, first ask the LLM to transcribe or summarize the content to verify it's correctly processing the information.

2. For language learning, create custom GPTs with few-shot examples (showing input and desired output) for better accuracy in translation and language tasks.

3. For academic or technical reading, break documents into digestible chunks and ask questions as you read to improve comprehension.

💡 Analyzing Manus’ launch playbook with ChatGPT

Emma generated this good and succinct analysis with ChatGPT. It identified not only Manus’ launch playbook but also the influencers it tapped + press pickup.

💡 The better the model, the more you pay, and the more you can get great results with very little prompting.

💡 Is your brand showing up in AI-generated answers? SEO tips for LLMs

Scan today, save this for when you’re ready to tackle LLM optimization. There is a lot in there.

AI is becoming a key influencer, not a major traffic source yet. but you want to be well-positioned when it does. This two-part guide offers concrete tips (thanks, Jessica Gilmartin, for sharing). Part I - Part II.

Their main tips:

1. Structure content with clear headings, FAQs, and proper HTML tables

2. Focus on author credibility with detailed bio pages for all content contributors

3. Link to high-authority sources and create clear internal linking structures

4. Highlight media mentions and third-party citations to build authority

5. Dominate a specific niche rather than covering everything superficially

6. Make valuable content public instead of hiding behind gates (well, we can debate that another day…)

7. Test by asking AI the same questions your prospects would ask"

They make great points, but remember:

• Focus first on creating genuinely valuable, novel insights that AI can't generate itself and that solve customer problems

• Test AI citation strategies with small experiments before full implementation

• Balance AI optimization with traditional SEO (most traffic still comes from Google)

Final Words

With appreciation for you reading this. François | LinkedIn