Alright, let’s talk AI.
This week we’re calling out three things: the editors getting exposed by their own AI tools, the SaaS era that never delivered, and how the companies actually using AI are finally getting real results. And yes, we hit the AI news you actually need to know.
Let’s get into it.
No, But Seriously…
Did Everyone Forget How to Proofread?
A national newspaper literally printed ChatGPT’s editing note in an article.
“Would you like me to extend this section to flow directly into the next part…”
Unreal, but it happens. I’ve done it myself a few times. Copy too fast, forget to trim the suggestion, boom. AI speaks for you in public.
I even did it once in Slack to my boss. I deleted it instantly. I wonder if he saw, I guess we’ll never know.
Somewhere, an editor is staring at their screen like it is a Monday horror show. And honestly, with AI policing everything these days, you need to be deliberate about protecting yourself.
The lesson? AI is helpful until it isn’t. Pay attention, or it pays for you.
The Drop
The Death of SaaS Platforms and the Rise of Workers
Listen up, all you stupid SaaS companies. Nobody wants your "platform" anymore. They want a digital workforce.
Every few years, SaaS sheds another layer of BS. Here’s the trail of bodies.
On-Prem (1990s to 2000s)
Buy, Install, Maintain, Suffer
Software used to live on physical machines. You bought a license, installed it manually, and prayed the update didn't crash your system.
IT ran the show, not the user. Innovation was slow, expensive, and limited to whoever could afford hardware and infrastructure.
Cloud-Based SaaS: 2000s - 2010s
The Great Migration
Software used to live on physical machines. You bought a license, installed it manually, and prayed the update didn't crash your system.
IT ran the show, not the user. Innovation was slow, expensive, and limited to whoever could afford hardware and infrastructure.
The Platform Era: 2010s - 2020s
The Rise of Ecosystems
SaaS companies realized integration was power. Platforms like Salesforce, HubSpot, and Shopify built ecosystems that connected hundreds of tools. The focus shifted from selling features to building marketplaces. The new advantage was "extensibility." The more apps plugged in, the more valuable the platform became.
AI & Automation in SaaS: 2020s - Present
Software Starts Thinking
AI entered the chat. We added automation, copilots, and chatbots. SaaS finally started to predict, assist, and learn. But it still wasn’t autonomous. These tools could suggest or summarize, but they couldn’t own outcomes. They made humans faster, not freer.
Digital Workers: Now - Future
SaaS Becomes a Workforce
Now, SaaS is crossing the final frontier. AI agents can actually perform the work, research, outreach, support, recruiting, without needing constant human oversight. They don’t just automate tasks; they manage workflows, collaborate, and deliver measurable results. The software itself becomes the workforce.
Here’s the full breakdown.
The Shortcut
The 74% Are Winning. The 95% Are Just Talking
MIT says 95% of AI projects fail. Wharton says 74% of enterprises are succeeding. Somebody’s lying. And it’s not the ones in production.
Here’s what the 74% winners are actually doing:
They stopped dicking around with pilots: Eighty-two percent of leaders now use AI weekly, forty-six percent daily. Not pilots. Not prototypes. Production.
They treat AI like money, not magic: Seventy-two percent track ROI tied to profit, productivity, and throughput. If it doesn’t hit the P&L, it doesn’t get funded.
They hired AI agents, real cloud employees: Fifty-eight percent are already using or testing autonomous agents. Not dashboards. Not prompts. Actual digital coworkers that execute, not experiment.
They gave AI a seat at the table: Sixty percent of enterprises now have a Chief AI Officer. Smart orgs didn’t add another tool, they added authority.
They killed the excuses: Seventy-four percent report positive ROI. That’s not luck. That’s execution.
The Weekly Wire
1. Humans and agents aren’t competing, they’re combining
CMU and Stanford found agents were far faster and cheaper, but humans still won on quality. Agents also made up data when stuck. The best results came from pairing agent speed with human judgment.
2. Data centers are moving underwater
China is scaling ocean-cooled data centers as cooling becomes compute’s biggest bottleneck. The ocean solves land, water, and energy constraints in one move. Infrastructure is shifting because traditional sites can’t keep up.
3. The MIT doom story is breaking
Wharton’s new survey shows 72 percent of companies get real AI ROI, unlike MIT’s 95 percent failure claim. MIT relied on 52 interviews; Wharton used 800 respondents. AI use is now mainstream across core workflows.
4. Yann LeCun just split the AI world
LeCun left Meta to build models that learn from the real world instead of predicting text. He believes LLMs are the wrong path entirely. If he’s right, today’s models are already outdated.
5. Cursor’s growth curve is not a bubble
Cursor went from 200M to over 1B ARR in six months. Investors wrote multi-billion-dollar checks because the revenue supports it. This is structural demand, not speculation.
6. Microsoft quietly outlined the agent economy
Satya made it clear the moat won’t be models but the environment agents work in. Identity, storage, workflows, and compute become the new stack. Microsoft already owns the pipes those agents will run on.
That’s it for today. Connect with me on Linkedin to hear more about how to build an Autonomous Organization.
