If you only check one highlight this week:

For a glimpse of our tools’ future capabilities: the first section.

To accelerate the production of case studies, see the AI workflow.

-François

The agents are coming here

How will our favorite tools’ agentic capabilities evolve?

How will that change our jobs?

What just happened in Coding AI is telling, as Coding AI is paving the way.

👉 In the last few weeks, most coding AI vendors introduced background agents and tools access. And that changes how developers work.

How coding AI has evolved in only 2 years (a simplified view)

1️⃣ Stage 1: Smart autocomplete 

→ Coding AI started as a smart autocomplete in the Integrated Development Environment (IDE), i.e. where devs write and edit code. Think Gmail's suggested replies, but for code.

2️⃣ Stage 2: Multi-file transforms

→ Next, AI began suggesting entire sequences of code changes across multiple files with cascading suggestions. Still in the IDE.

3️⃣ Stage 3: IDE Task agents

→ Then came in IDE agents that handle specific tasks ("write me a test for XYZ") following chat prompts in the IDE and central instructions. Sometimes using reasoning models, working from a few seconds to minutes.

4️⃣ Stage 4: Remote/Background agents and tool use

→ Now we have background agents that can work independently. They use advanced reasoning models and connect to other tools via Model Context Protocol (MCP - think universal API access). They work outside of the IDE and can be triggered from there or other tools (ticketing systems, terminal, etc). They work simultaneously from a few to dozens of minutes.

📔 Throughout: more and more context

→ Coding AIs initially indexed local files with code, then entire codebases, and now ingest more and more context, also tapping unstructured but essential data (chats in Slack, Linear tickets, Notion docs, etc).

🔜 What’s next?

→ Expect agents to move upstream into planning (turn Jira tickets into code tasks) and downstream into deployment and optimizations to assist developers with more tasks in their workflow.

📍 The result:

→ Developers are becoming orchestrators. They spend less time writing code. They manage agents, set context, review results. Understanding architecture and code is still required to architect all of this correctly. Humans are STILL responsible for the outcome.

This shift took less than 18 months.

What does that mean for non devs?

Marketing, sales, ops, and design tools will likely follow the same curve.

With advances in reasoning models, giving full access to our “codebases” (i.e. our foundational documents, style guides and body of work), and connecting our AI to most of our tools (integrations via MCP), we should benefit from more agentive workflows that can:

1. Work in the background for us (already the case when we use deep research in ChatGPT or Gemini)

2. Propose cascading changes when we update something to our central foundations or instructions. For instance, we could see automatic updates to some of our marketing assets after updating a color in our brand style guide.

3. Help us go from planning all the way to execution - with the right check points

4. Notify us when we need to review a plan or some work before the AI gets to the next stage, bringing us “humans in the loop”.

So let’s stay tuned. These tools will be a significant time saver.

My selection of tips, news and workflows

🔍 The must-watch influencer talk for B2B marketers

I shared the key takeaways from Jessica's influencer talk two weeks ago, but I recommend watching her talk. She shares examples of successful - and not successful - TikTok and Instagram videos made by influencers for B2B.

🛠️ Claude can now analyze your Gmail game

Did you know Claude can run quantitative and qualitative analyses of your email activity, habits and style? It gave me new insights about my email game. 

First, connect Gmail with your Claude account (“connect apps” under the chat window)

Second, select one of the pre-loaded prompts (screenshot below)

It starts with a qualitative analysis, because the Claude team pre-loaded this prompt for the first one:

“​​Hey Claude, could you analyze my email communication style? Use whichever of my integrations that make sense. When you think you’ve gathered sufficient information, please do the task. Use an artifact if you think that’s helpful. If using an artifact, consider what kind of artifact (interactive, visual, checklist, etc.) might be most helpful to this specific task. Do not use analysis tool. Please do it right away.”

Third, ask it to do a quantitative analysis with the data analysis tool. You can iterate on my prompt, depending on what insights you need:

“Now, use the analysis tool and run an analysis over the last X months. 
Create a series of charts - I let you select the best formats - that will provide me with insights on:

• when I read emails

• when I write emails

• the top 10 people I email most with

• the top 10 companies I email most with

• the distribution of emails I sent (work, personal, admin, etc.)

• anything else that will give me quantitative insights about my email game

🛠️ An AI workflow to craft and promote case studies

Luke Harries (Head of Growth at ElevenLabs) shared his workflow to streamline the time consuming process of writing and promoting case studies.

It turns a casual Zoom chat into a blog post, tweet thread, and even a LinkedIn post—all in their brand voice.

Step-by-step workflow:

1. Use Zapier and Calendly to automate interview scheduling after each closed deal picked up in your CRM.

2. Interview your customer on Zoom (10–20 min) while Granola transcribes*.

3. Grab the Granola summary and full transcript.

4. Paste them into a Custom GPT, trained with tone, examples, and detailed formatting rules.

5. Get back a blog post + a tweet thread—with image placeholders and hooks.

6. Optional: Clone for LinkedIn, or founder-voice with another GPT.

*Notion and ChatGPT this week for Team accounts have also introduced their background recorders.

His best tip: don’t like the outcome? Instead of editing the output, further work on the prompts.

You can also ask the GPT to craft emails promoting the case studies, copy for slides, FAQs. You can also attach all case studies in a Notebook LM or ChatGPT Project and ask them to create summaries or compilations by vertical, geo or company size.

💡A CEO shares what helps make his team more fluent with AI

Matt, CEO at Concord, shared what’s been working for them.

“Here’s what we found that works:

1) The “before/after” method

Teams write out how they solve problems pre-AI. Then they plan out how to use AI to solve those same problems, faster. Then they implement it.

2) Peer teaching Fridays

Every Friday, several team members explain how they used AI to solve a specific problem. This celebrates AI achievements that have real impact on the company, and encourages others to ask questions. The best sign is when they go, “Oh cool!”

3) Innovation tracking

We don’t count “usage,” we count new use cases that measurably save time. Last quarter we tracked 5 new applications just from the marketing team (well done Dean!).

4) Problem-first thinking

Teams need to identify problems that need AI, vs. ones that need human judgment. Correct identification of an AI-suited problem predicts successful adoption more than any other factor.

🛠️ (MCP) Our Chatbots can work with our favorite tools but what can non-devs do so far?

“If you are not a developer, you are probably confused with this MCP thing.”

Yet, it unlocks so much:

So Greg Baugues kindly introduced MCP for non-developers to the PLL community, explaining what is MCP and sharing what non-devs can and can’t do with it yet.

It’s changing fast! Just this week: ChatGPT rolled out MCP for Deep Research and Claude introduced remote MCP server access to the $20/month Pro plan (was only in the $100/mth Max when Greg recorded his talk).

Final Words

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.