Hello Reader!

Something shifted recently. And I'm feeling it too.

Not the usual "oh isn't it all so fast-moving in this AI space" kind of thing, but something more specific than that. A kind of low-grade overwhelm. And it's not very comfortable.

We're not talking about the good kind of overwhelm here - that only comes when you download and start building with tools like Claude's Cowork which are genuinely powerful for non-technical people. For the first time, someone without a line of code to their name can build systems that do things that are worthwhile. Spin up dashboards that automatically update because they are connected to the sources. Workflows that move data between tools without anyone touching a spreadsheet. Create reports that write themselves from your project files.

The initial reaction for most people is to go all in. Hours disappear as you set your mind to solving problems you didn't even know you had. Tidy your files anyone? Done. In a jiffy.

Every week there's more announcements. Cowork in Copilot (if you're in a Microsoft shop, that's genuinely significant). Copilot in everything. All the OpenClaw lookalikes. New models everywhere.

And it's not just overwhelming for people starting out. In the forums I'm in, experienced practitioners are complaining that their multi-agent pipelines aren't talking to each other, or that their implementation of Karpathy's testing framework isn't performing the way the thread promised. These aren't beginners. And even they are struggling.

So I want to be straight with you: we are genuinely in the messy middle, or as Dario Amodei put it 'the adolescence of technology'.
​

The tools are real, the potential is real, but so is the complexity of adapting the way we've always done things to be an adopter. But now more than ever, we need to do that because agentic AI isn't just "better ChatGPT." It's an entirely different category of a thing. It takes a little bit more learning to get on board with it because you don't just show up and type like you did with ChatGPT. But trust me - it is so worth the effort.

Thoughts on my mind mid-March 2026:

• Will companies actually let their staff build agents?
• Is giving a non-technical employee the ability to build autonomous systems inside a business a step too far?
• Who owns what gets built?
• Who's responsible when it breaks?

I don't have clean answers to any of these. But I think they're the right questions to be asking right now.

These were some of the things on my mind when I spoke to Sean Blanchfield, CEO of Jentic, for this week's episode of ChattingGPT. Their whole approach is built around the idea that agents should prove themselves in a safe environment before they touch anything real - and it's one of the most grounded conversations I've had about where this is all heading.

Have a listen. Not to get ahead of the wave, but to understand the shape of it a bit better.

And finally... relax, because our team is working on a brand new course on Claude Code and Cowork, and we'll be releasing the details soon.

Sean Blanchfield Part 2: The Agents Edition

Most companies run on dozens of tools that barely talk to each other. Project management, ERP, payroll, documents - all siloed, all held together by human co-ordination.

In this episode of ChattingGPT, Sean Blanchfield, CEO of Jentic, explains how his platform builds a digital twin of a company's entire software stack. Inside that sandbox, AI agents can explore workflows, connect systems, and figure out how work actually gets done - without touching live data or breaking anything in production. Only once an automation proves itself do humans approve it for real deployment.

It's a genuinely different model for AI adoption: experimentation first, governance built in from the start.

Sean also shares his thinking on OpenClaw, and why the infrastructure layer for agentic AI matters more than most people realise right now.

If you've been watching the agentic AI space and wondering how it actually lands inside a real business, this one's worth your time.

🎧 Apple Podcasts ​

🎧 Spotify ​

Claude Code/Cowork Course coming soon...

Yours in overwhelm,
AI Institute