Why I'm Done Chasing AI News (And Where Smart Executives Really Focus)

🧭 THIS WEEK AT AI SECOND ACT

Howdy, well, vacation is over. Summer feels like it stopped almost immediately here in Chicago. The only upside of that is (maybe) fewer weather complaints from my Australian family! Kids are back in school, with one now in college and so it feels like everyone is back to the grind!

This week:

• AI Second Act pivot/positioning

• Organizing chats and 10x time savings tips with chatbot tools

• Training / Onboarding - a MASSIVE first win possible through AI

If you’re enjoying this newsletter, please share it! Forward this to a friend/colleague and have them subscribe at aisecondact.com/subscribe. Thanks!

👉 You can also just hit "Reply" and let me know what you want more (or less) of, or use the poll to give feedback.

My goal is to make this as valuable and practical as possible as we navigate the new AI era. 🚀

🧰 AI NEWS + LEARNING

Here are a few things I found recently:

• Claude Code as Personal Assistant - definitely a bit ‘low level’ for some, but others will like this and get into it as a way of getting complete organization.

  • Reminder, Claude Code is a terminal-based ‘vibe coding’ tool from Anthropic - it’s what all the cool kids are using. I suspect Gemini CLI is as good or ‘nearly there’, but Claude Code is what seems to get all the press.

    • Hold the press - Claude Code wins, but Gemini CLI is free currently with huge limits, so great for learning.

• Wired article on ‘vibe coding’ - actually, in all likelihood, I’ll be launching another newsletter on this topic of AI leveraged coding - basically, coding through the use of natural language. It is fascinating and amazing. Read this just to get the briefest ideas as a journalist contributes code to a huge startup, Notion. Out with Pascal, in with natural language!

• Adobe is adding AI to PDFs.

• Anthropic Claude usage limits are extreme, and turning me away from Claude, even though I do like it. Not my fault [kind is] that it’s 8:58 pm, the night before newsletter publishing, and I need to use Claude but limited. Dang it.

🗺️ FEATURED INSIGHT

AI Second Act going where?

I can’t / don’t want to compete (and drown) in (somewhat useless) AI news. There is little to no impact on who got what funding, or who acquired whom, or even a new AI model being released. What matters is the value we derive from AI, the strategic steps we put in place to make our lives & business better through AI.

So, AI Second Act has a new position!:

Where Operating Executives Master AI Strategy

Less ‘news/tips’ and more on how to drive value with AI.

Brand new content calendar and plan coming up.

Chat Organization and Super Powers

After a lot of chatbot use, here are my latest conclusions:

• Custom GPTs are limited, really - They are only useful if you want to create and customize a chatbot that you can then share with others. A common set of instructions and context that your team can benefit from. Key limitations:

  • You can not see the history of chats for that GPT - they are lost in the UI of ChatGPT

  • Each chat is a silo - no context/memory sharing between chats - this is a huge disadvantage

  • (Note, I’ll be experimenting with these a lot more at work soon, and so will have some further thoughts then)

• For a lot of cases, both business and personal, create a project in your tool of choice:

  • Hobbies

  • Work

  • Side Gig

  • Travel

  • etc.

• Within the project, upload custom context via text or documents with instructions and ‘background’ for the project.

  • These files should give context to the project - basic instructions, type of responses/voice/formatting you’re looking for, background information for the project, goals you have, and any timeline you’re working on, etc.

  • Make sure each project file has a meaningful name

  • Don’t go too crazy - 10 to 20 to 50 files (if the tool allows it), maybe OK - not more! [Claude and Gemini have higher file limits vs ChatGPT]

• Finally, consider what connections are possible, = Super Powers.

  • Claude is the clear leader here. Connections make it possible to ‘do extra stuff’. Read your calendar or email and connect to automation tools like Zapier, among many other efficiency opportunities.

Training 10x

You hear it almost weekly at work, right? Training takes 2x, 3x the time estimated. Team ramp-up causes inefficiencies left, right, and centre. AI to the rescue!

Building upon the above discussion of projects and custom GPTs, and Projects.:

• Figure out which is better and, importantly, which can be shared with your team - potentially, this is only a form of custom GPT

• Keep each chat with a relatively narrow focus. Examples:

  • Basic project introduction

  • Feature xyz expert

  • Customer introduction and ‘handy hints’

  • Project management & schedule

  • etc

• Upload custom context documents and instructions for each chatbot. Examples:

  • Requirements documents, even complete code bases for a feature

  • Customer org. chart & key contacts, escalation paths

  • Team org. chart and meeting calendar

  • High-level schedule and key critical path information

Then, simply roll out each chatbot to the team. Let them know where and how to use each, and fundamentally, these become the central source of truth for the project.

The models are truly amazing, and the ability to ask a model to show you a sequence diagram based on a certain code base, show you the architecture, etc, should massively improve ramp-up and training efficiency.

What do you think of this training angle? Do you use projects in the chat tools? Let me know what you think of my proposed AI Second Act position/niche!

Weekly AI strategies for operating executives
— Brett

👉 Hit “Reply” and share your experience — I read every one!

_{Picture by Mark Konig on Unsplash.}