Agent AI: Your Agency is Calling
The AI evolution, nefarious AI use cases, and lots and lots of weekend reading – your Friday Briefing is here!
Dear Friend,
As we are baking in temperatures reaching 100°F here in Boulder, CO, the AI wars continue to rage. The latest target? Your browser – the very piece of technology Jane and I worked on over 15 years ago when we were at Mozilla (home of the venerable Firefox web browser). Arguably, not much has happened since we launched Firefox 1.0 to today. Browsers have become faster, capable of rendering and running more complex, interactive apps – but by and large, it is still a URL bar and some rendered HTML. Now Perplexity (and rumored OpenAI) are integrating their AIs deeply into the browsing experience – and, of course, you have Johnny Ive’s fabled AI-native device. And Zuck’s AI-enabled Ray-Bans… It’s going to be interesting to see which user experiences will truly dominate our AI-powered world – it surely won’t be the prompt/text box of today’s LLMs.
And now, this…
Headlines from the Future
From AI to Agents to Agencies: The Next Evolution of Artificial Intelligence ↗
Nishant Soni on the evolution of AI agents (or agentic AI) toward “agencies”:
What I’m witnessing is the birth of what I believe should be called Agencies - systems that tackle individual tasks by dynamically orchestrating different types of intelligence, each optimized for specific subtasks, all working toward completing a single overarching objective. An Agency is fundamentally different from an Agent. While an Agent is a single intelligence (an LLM) enhanced with tool-calling capabilities working on a task, an Agency is a coordination system that can access and deploy multiple specialized intelligences (LLMs) to complete different parts of the same task.
Think of an agency as a “boss AI” (a term I learned from Scot Wingo, founder and CEO of ReFiBuy.ai – he brought this up on our recent podcast conversation), consisting of three parts:
Task Context Management: The Agency maintains unified context about the specific task at hand - requirements, constraints, progress, and accumulated decisions. This ensures continuity as different intelligences contribute to different subtasks.
Intelligence Allocation System: Rather than using one model for everything, the Agency has access to multiple specialized intelligences and dynamically selects the most appropriate one for each subtask within the larger task.
Orchestration Logic: A coordination system that breaks down the main task into subtasks, determines which intelligence to use for each part, and ensures all contributions integrate coherently toward task completion.
In summary: “Agencies are not multiple Agents collaborating on a project. They are single unified systems that can access multiple types of intelligence to complete individual tasks more effectively. […] We’re moving beyond asking ‘What’s the best model for this task?’ to ‘What’s the best combination of intelligences for different aspects of this task?’”
In many ways, this is the evolution of “mixture of experts,” where a single AI (LLM) has access to multiple, specially trained, typically smaller models and routes requests to the most capable model for a specific task (e.g., a model which is optimized for coding tasks).
—//—
Hertz AI Scanner Charges $350 for Tiny ‘Dings’ on Rental and This Is Going Off the Rails ↗
Blockbuster had late fees. Apparently, Hertz (the car rental company) has damage fees – turbo-charged by AI.
Hertz, the “we try harder” folks, started to deploy an AI-powered scanner for vehicle inspections (the thing you do when you return your rental, and, in the old days, a human walks around the car to see if you have any dings on it). The scanners reportedly identify slight dings (the stuff a human would just ignore) and immediately send you a bill:
“If you followed our last story involving the wheel scuff, you know that UVeye—the firm that produces and operates the scanners—and Hertz like to secure payment of these fees as quickly as possible. They do this by discounting the charge if the customer admits fault and pays within seven days. Foley said that Hertz offered to knock $65 off the bill if he paid immediately. Furthermore, we’ve heard that contacting a human agent at the company to discuss or contest the charges is very difficult, and not possible within the web portal where customers can view and pay for damages. You have to call a separate support line instead, though Hertz doesn’t seem to make that very clear.”
There might be a good reason for Hertz to act like this:
“I suspect the math of investing in such expensive technology indicated they needed to go to extortive levels to get a [return on investment].”
—//—
Adding a Feature Because ChatGPT Incorrectly Thinks It Exists ↗
Sign of the times: We are now adding software features not because we need or want them – but because ChatGPT hallucinated them and gets its users to request them.
Turns out ChatGPT is telling people to go to Soundslice, create an account and import ASCII tab in order to hear the audio playback. So that explains it! […] Problem is, we didn’t actually have that feature. We’ve never supported ASCII tab; ChatGPT was outright lying to people. And making us look bad in the process, setting false expectations about our service. […] We ended up deciding: what the heck, we might as well meet the market demand. So we put together a bespoke ASCII tab importer.[…] To my knowledge, this is the first case of a company developing a feature because ChatGPT is incorrectly telling people it exists.
—//—
Techno-Feudalism and the Rise of AGI: A Future Without Economic Rights? ↗
More food for thought:
The rise of Artificial General Intelligence (AGI) marks an existential rupture in economic and political order, dissolving the historic boundaries between labor and capital. Unlike past technological advancements, AGI is both a worker and an owner, producing economic value while concentrating power in those who control its infrastructure. Left unchecked, this shift risks exacerbating inequality, eroding democratic agency, and entrenching techno-feudalism. The classical Social Contract-rooted in human labor as the foundation of economic participation-must be renegotiated to prevent mass disenfranchisement.
—//—
The Force-Feeding of AI on an Unwilling Public ↗
Here is an interesting weekend comment on the current reality of AI being integrated into… everything:
Most people won’t pay for AI voluntarily—just 8% according to a recent survey. So they need to bundle it with some other essential product.
You never get to decide.
Before proceeding let me ask a simple question: Has there ever been a major innovation that helped society, but only 8% of the public would pay for it?
That’s never happened before in human history. Everybody wanted electricity in their homes. Everybody wanted a radio. Everybody wanted a phone. Everybody wanted a refrigerator. Everybody wanted a TV set. Everybody wanted the Internet.
They wanted it. They paid for it. They enjoyed it.
What We Are Reading
🎭 Influencing Without the Influencers Brands are increasingly replacing traditional influencers with in-house content creators who produce casual, behind-the-scenes videos that blur the line between marketing and everyday social media content. @Mafe
🎯 Trump, Mamdani, and Cluely: Attention and Speculation as Primary Economic Drivers Don’t be put off by the political names in the title: this is a fascinating argument about how attention is leveraged to create value across domains. @Jeffrey
🚀 The Year of Quantum: From Concept to Reality in 2025 Is quantum finally having its moment? McKinsey reports surging breakthroughs and investment doubling that point to a $100B+ market in a decade. @Kacee
🎓 How Steve Jobs Wrote the Greatest Commencement Speech Ever A great account of what we’ve all probably come across, heard, or read about in some form. The usually stage-dominant Steve Jobs was apparently quite taken aback by the preparation and delivery of his Stanford commencement speech. In his popular finishing touches, “Stay Hungry. Stay Foolish,” he transported the work of Stewart Brand to a new generation. @Julian
🎭 My Instagram Reel Just Went Viral (1.3M Views). Here’s Why That Sucks. A comedian sees his joke go viral – only to realize that it’s not the humor but the fighting that makes it so. A sober lesson in what works (and doesn’t) in today’s media landscape. @Pascal
Rabbit Hole Recommendations
Microsoft shares $500M in AI savings internally days after cutting 9,000 jobs
Your’s truly quoted in EY’s report on “What if disruption isn’t the challenge, but the chance?”
‘Positive review only’: Researchers hide AI prompts in papers
Springer Nature book on machine learning is full of made-up citations
Happy Distractions
🧶 Behold the “Wind Knitting Factory”
🥗 The story behind Cesar salad
😯 Hundreds of robots move Shanghai city block (video)
😎 What makes someone cool? A new study offers clues.
🤑 Wondering how much money Meta spends on their AI engineering hiring spree? Wonder no more: Zuck’s Haul: Meta AI Talent Tracker
Re: “What if disruption isn’t the challenge, but the chance?”, it feels like strategic consulting is experiencing its own terminology race condition. Much as dating trends get a weekly new phrase and, apparently, vacations are now "micro-retirements".
We had VUCA, but not content we now have NAVI but also:
-BANI: https://stephangrabmeier.de/bani-versus-vuca/
-RUPT https://www.ccl.org/articles/leading-effectively-articles/navigating-disruption-vuca-alternative/
-TUNA: https://www.forbes.com/sites/adamgordon/2016/04/06/oxford/
I'm not clear on the purpose of all of these other than either a perceived need for a shock & awe campaign or just plain old boredom. VUCA still covers complex adaptive systems in my book and all their interconnectedness. The environment is the same, just matured.
A better target for replacement would be a term like "moonshots", though not to diminish the achievement. When JFK said we'd go to the moon at Rice University on September 12, 1962, we knew precisely where the moon would be within centimeters at the time of Apollo 11's landing at 20:17:40 UTC on July 20, 1969.
The difference with complex adaptive systems is more like the moon changing course because it sensed we were coming... and it swung several meteors in Apollo 11's path.
Thinking about the Hertz story, businesses jump the shark when they put the onus of loss prevention on the honest customer. A good rule of thumb is to never put the burden of LP on the honest customer. Even honest customers know that very little of shrinkage costs can be passed on to the consumer if the market is the least bit competitive.
Oddly, of all companies, I see Ticketmaster going in the opposite direction. Long reviled due to their "convenience charges" appearing to be just a mere "rake" or "vig" on ticket purchases, they are now offering a better CX and added value for what they charge, such as a reliable resale platform, easy selection of seats, reminders to buy tickets, seat-views of the stage and the like, etc. Sure beats camping out outside the record store or the department store!