Hey There,
I decided withAgents still hasn’t been something I’d want to read. So here's a new format.
The ‘card’ format used recently just, honestly, wasn’t that interesting. I found that it wasn't super memorable for me as I went about the rest of my life. And I write this thing! Maybe the newsletter has been to frequent too... not sure people want something daily on AI.
So we iterate! I want to lean into bundling something extremely valuable for you.
Call this a Love Bomb. A pack, stealing a bit from James Clear’s 3-2-1 format, but focused on giving you more confidence in using and talking AI.
I give you [drumroll] the GAIN format.
Each week, something to “Go Do” with AI, “Alerts” covering top news, “Ideas” to chew on and “Network” tips (people to follow, share, etc.).
LMK what you think,
Clay
Guess the number
According to a new report by Wolters Kluwer, what % of financial institutions are now exploring AI use-cases (or have at least one proof of concept?)
(answer in signature)
Go (do this)
Today, we look at creating your own personal newsetter, with AI.
other things that launched this week,
for you to check out:
YC Alum, Klavis AI launched “Strata”, an MCP server to rule them all - Go see it
Gamma (deck creation AI) launched their developer API this week - Go see it
Notion 3.0 arrives, deeply integrated AI agents (great video here) - Go see it
Update your Chrome for the new AI features (okay video here) - Go see it
Alerts
Spotify removed 75 million fake or AI slop-tracks and is tightening its content filters. Your favorite music app is adding stronger spam detection and clearer labeling for AI-”assisted” music to protect legitimate artists and keep the platform usable for listeners. Read more here.
Books warning of superintelligent AI taking control a’la Skynet are climbing the bestseller lists. Two new releases with spicy titles like If Anyone Builds It, Everyone Dies argue that tomorrow’s AI could escape human control. Given the news buzz around these books (written by highly credentialed AI researchers, btw) expect this to bring existential risk back into mainstream conversation. Read more here.
McKinsey reports that AI is forcing software companies to change how they make money, not just what they sell.
Instead of charging for access to tools, leading firms are shifting to usage- or outcome-based pricing as AI systems begin to perform work on behalf of customers. Read more here.Start saying “WORKSLOP”. It stands for “AI generated work content that masquerades as good work, but lacks the substance to meaningfully advance a given task.” HBR finds that rushed AI adoption is creating “workslop” that lowers productivity. We’ve all seen it in our feeds; teams are making loads of AI content without fixing broken processes (see the ideas section below, I fully agree), which adds busywork and erodes the value that automation was supposed to deliver. Read more here.
Disney and other studios are suing a Chinese AI company Minimax for aggressive IP theft of characters / stories. Similar to suits happening against Midjourney and others. The outcome could set new legal standards for how intellectual property is protected when AI models are trained on creative works. Read more here.
Ideas
“Can this be done with AI?” is mostly determined by your current workflow and processes … And your current workflow and processes probably somewhat suck.
It’s a truth we’re finding more and more, both in personal and professional usage of AI.
There is so much in our regular work — administrative, creative, collaborative, etc. — that is ad hoc, toil-heavy and inconsistently executed. Smart, experienced people often work a bit on instinct. They manage their chaos and selectively filter workloads to deal with whatever pressures they’re under in the here/now.
This makes it super hard to accurately figure out where AI (and most specifically, AI-enabled automation) can really be put to use.
When it comes to applying AI to work, the #1 most important thing to be clear on is the outcome you’re after — and usually that means the output you’re after. The actual asset, or document, or template. How standardized and regular is that output, and how consistently is it allowed to be in this form?
The same goes for the workflow. How well is it defined, end-to-end, from the start? If you’re not actually all that clear on this, you’ll probably miss the AI improvement opportunity.
Take this forward:
Document your steps. Take an hour to sketch your current process. Inputs, steps and outputs. Even this rough map will reveal bottlenecks and ad‑hoc tasks that could be automated.
Standardize the output. Decide on a “canonical” format for the deliverable (template, style guide, required fields). The more uniform the output, the easier it is to train or prompt AI to produce it.
Run small experiments. Pick a low‑risk, early step and automate it with an AI tool or script. Measure time saved and quality; iterate before scaling. Build on top of that iteratively. Don’t hand over the whole workflow until each piece works reliably.
(☝️ This might annoy AI‑agent maximalists who think a general‑purpose agent will solve everything. Oh well.)
Network
This week’s list was pretty easy. Some giants.
Reply to their next post: Andrew Ng
Andrew Ng is a computer scientist and entrepreneur best known for his work in artificial intelligence and deep learning. He is recognized for founding the Google Brain project and for co-founding the massive learning platform, Coursera. He is an active, compelling voice and posts frequently. A good follow.
Learn from their thinking: Mustafa Suleyman
Career-wise, a titan, but Mustafa first caught my attention with his recent post on “Seemingly Conscious AI (SCAI)”. Mustafa is a Brit, CEO of Microsoft AI and long entrepreneur/major figure in the world of AI.
thanks
So how many financial orgs are using AI? The Wolters Kluwer survey found that 86% of North‑American finance leaders said their organizations were either beginning to explore AI use cases (53%) or piloting AI in select areas (33%)
Have a great week,
- Clay G [LinkedIn]
Script for n8n asset creation workflow here! [Google Doc link]
