Friday, February 27, 2026
FRIDAY — AI FOR THE C SUITE
Read time: 10-11 min · Read online
Hi, it’s Chad. Every Friday, I serve as your AI guide to help you navigate a rapidly evolving landscape, discern signals from noise and transform cutting-edge insights into practical leadership wisdom. Here’s what you need to know:
1. Sound Waves: Podcast Highlights
“It rips hierarchies apart.” That’s how one of the most-cited workforce policy experts in the country describes what AI does to traditional org structures. If you’re still running a top-down AI rollout, my most recent episode with Brent Orrell, Senior Fellow at the American Enterprise Institute, explains why your initiative is already failing. Catch up on my conversation with Brent, plus all the others, wherever you press play for your pod.
Apple · Spotify · iHeart · Amazon · YouTube
Subscribe for free today on your listening platform of choice to ensure you never miss a beat. New episodes release every two weeks.
2. Algorithmic Musings: “You’re Not Special, Babe.”
This title (and hook) from Orla Gartland’s song echoes in my ears every time I read or hear someone state that AI is incapable of doing this, that, or some other thing. These points of view boil down to one core tenet: that humans are special. And the sum total of the counter-point? You guessed it. You’re not special, babe.
Humans are special in a great many ways. And yet… our specialness does not in any way mean that many of the things we take pride in cannot and will not be done by AI in a superior fashion. We’ve seen this story before and it’s time to stop pretending that John Henry didn’t die from challenging the steam engine. Instead, it’s time to start acting with the knowledge that AI will keep surpassing human abilities, and the pace isn’t slowing down.
Now, before you accuse me of being an AI cheerleader, let me be clear: AI isn’t special either. Not yet, anyway. It has real and measurable shortcomings that we’ll explore in this piece. But its limitations are a moving target while ours tend to be more… persistent. And that distinction matters more than most leaders realize.
We come by our reactionary nature quite honestly. Dylan Thomas famously advised us to “not go gentle into that good night” and to “rage, rage against the dying of the light.” It’s stirring poetry. It’s also terrible business strategy.
Stated another way, it’s time to stop raging (against the machine) and begin embracing our humanity.
This week, five papers crossed my desk that put hard numbers behind all of this. Their findings not only challenge the “humans are special” narrative but also reinforce what we’ve been talking about in this space for the past several years. Let’s dig in.
3. Research Roundup: What the Data Tells Us
The Doctor Will Not See You Now
If there’s a profession we hold up as the pinnacle of specialized human expertise, it’s medicine. Radiologists in particular spend years training their eyes to detect abnormalities in imaging. Surely AI can’t touch that, right?
Wrong. Research from a 2023 study involving 180 professional radiologists found that AI predictions outperformed two-thirds of participating physicians on diagnostic tasks using historical chest X-ray cases. But what should really concern leaders is this: providing radiologists with AI assistance didn’t actually improve their diagnostic performance. The study identified specific cognitive biases that kicked in when experts tried to incorporate AI’s predictions alongside their own judgment. In many cases, the doctors were worse off trying to collaborate with AI than they would have been letting the AI work independently.
Read that again. The AI was better alone than the humans were with AI help.
If you’re sitting in a C-suite thinking, “Well, our people are different,” I’d gently remind you of the title of this article. The radiology study’s findings extend well beyond medicine into financial analysis, quality control, and every other domain where expert classification is being augmented with AI. The researchers make a provocative recommendation: optimal task allocation should treat human-AI collaboration as the exception rather than the rule.
The number that matters: AI outperformed 67% of trained radiologists on diagnostic tasks. And the study used a 2023-era deep learning model for chest X-ray interpretation. Think about what 2026 models might accomplish.
What this means for your Monday morning: Audit where your teams are “collaborating” with AI versus where you should be letting AI lead. The answer will probably surprise you.
The Middle Is Disappearing (And Your Firm Might Be Standing on It)
A paper called “The Headless Firm” describes how agentic AI is fundamentally reshaping the boundaries of the enterprise itself. This aligns with Professor Anton Korinek’s research on the rapid de-escalation of knowledge work costs that we’ve discussed before in this space.
Traditional business structures required integration costs that scaled quadratically. The more components you added, the more expensive coordination became. Think of it like a dinner party. Six people? Manageable. Sixty? You need a whole events team. Agentic AI, operating through standardized protocols, reduces that scaling from quadratic to linear. That’s not an incremental improvement. It’s a structural revolution.
The result is what the researchers call the “Headless Firm,” an hourglass-shaped architecture. Personalized generative interfaces at the top. Standardized protocols in the middle. Competitive, micro-specialized execution agents at the bottom. For middle-market firms, this points to a bifurcation in firm size where organizations either specialize deeply or scale massively, with the broad middle increasingly unable to compete.
Your next move: If your organization sits in that broad middle today, this should be the most important research you read all year.
When AI Writes, Whose Voice Remains?
A study analyzing over 22,000 outputs from five large language models found something that should give every global organization pause. AI doesn’t just help you write. It systematically erases cultural identity markers from the text it touches.
The researchers call it “cultural ghosting.” When AI processed texts from Indian, Singaporean, and Nigerian English speakers, it stripped out linguistic markers unique to those varieties at a rate of about 10%. And the paradox that should keep every leader awake? The AI maintained high semantic similarity while doing it. Meaning was preserved. Identity was not.
The study calls this the Semantic Preservation Paradox. The models aren’t accidentally homogenizing voice. It’s a built-in consequence of how they’re aligned. The good news? Explicit cultural-preservation prompts reduced erasure by 29% without sacrificing quality.
Action item: If you’re deploying AI across global teams, this is an organizational governance issue. When your AI communication tools subtly flatten the voices of your international workforce into a uniform, culturally-neutral paste, you’re not just losing linguistic diversity. You’re undermining the very inclusion efforts you’ve probably got plastered on your values wall.
Your Mental Model of AI Is Probably Wrong (And That’s Dangerous)
Two more papers round out this week’s research, and they converge on a theme that should worry every leader deploying AI decision support.
The first, “2-Step Agent,” introduces a framework for understanding how decision makers interact with AI systems. The core finding: the effectiveness of AI-assisted decision making depends critically on whether you have the correct mental model of what the AI was trained on and what it can actually do. Even in idealized scenarios with well-calibrated AI, incorrect user expectations lead to worse outcomes than not using AI at all.
Sit with that for a second. You can have a perfectly good AI system and still make worse decisions with it if you don’t understand its limitations.
The second paper, on Theory of Mind in large language models (how well AI understands human mental states), reinforces this concern from the opposite direction. Researchers found that while top models achieve human-level accuracy on classic reasoning tasks, their performance falls apart when you introduce even modest variations. The models’ understanding is brittle. They fail not because they get lucky and then guess wrong but because their underlying reasoning is genuinely flawed in specific, identifiable ways.
What these two papers tell us together: The “human-AI collaboration” story being sold to boardrooms is far messier than vendor slide decks would have you believe. Leaders who assume AI “understands” context, culture, or nuance the way humans do are building strategy on bad assumptions.
So Where Does This Leave Us?
Back where we started. You’re not special, babe, and neither is AI.
Humans aren’t special in the way we’d like to believe. AI will continue surpassing us at tasks we once considered exclusively ours. The radiology study proves it. The Headless Firm research shows the structural consequences. But AI isn’t special either. Not yet. It erases cultural identity. It breaks under pressure. Leaders who misunderstand it make worse decisions than leaders who don’t use it at all. Yet every single one of those shortcomings is a snapshot of a moving target. The capabilities that feel “limited” today are advancing at a pace that makes year-over-year comparisons almost meaningless. The AI that stumbles on Theory of Mind perturbations this quarter will likely handle them by the next.
The path forward isn’t rage or denial. It’s clear-eyed understanding of what AI does well, what it doesn’t, and where human judgment still matters. Not because we’re special, but because we’re different. And different, when properly deployed alongside AI capability, creates something neither can achieve alone.
Three things to consider this week:
1. Where in your organization are experts trying to “collaborate” with AI when you should be letting AI lead? The radiology study suggests the answer might surprise you.
2. Does your leadership team have an accurate mental model of the AI tools you’ve deployed? If not, the 2-Step Agent research says you’re likely making worse decisions than you would without those tools.
3. Are your AI-powered communication and content tools silently homogenizing the voices of your global team? Cultural ghosting isn’t a future risk. It’s happening right now.
Research Sources:
2-Step Agent: A Framework for the Interaction of a Decision Maker with AI Decision Support — arXiv
Understanding Artificial Theory of Mind: Perturbed Tasks and Reasoning in Large Language Models — arXiv
The Headless Firm: How AI Reshapes Enterprise Boundaries — arXiv
Combining Human Expertise with Artificial Intelligence: Experimental Evidence from Radiology — MIT Economics Publications
4. Radar Hits: What’s Worth Your Attention
IBM stock drops 13% after Anthropic publishes AI-powered COBOL modernization playbook. Wall Street is finally pricing in what IT leaders already know: AI is making legacy code migration affordable. If you’re sitting on COBOL or other legacy systems and your vendor keeps quoting seven-figure modernization projects, the leverage just shifted in your favor. Get competitive bids that include AI-assisted approaches.
AEI: Companies are replacing labor hours with AI tokens as their core unit of cost. Enterprise AI spending jumped 36% last year even as per-token costs dropped 80-90%. The takeaway: your competitors aren’t just experimenting with AI anymore, they’re budgeting for it like electricity. If AI spend isn’t a line item in your 2026 operating budget, you’re already behind on cost structure.
AI systems now solving 40%+ of PhD-level math problems, up from 2% eighteen months ago. This isn’t about math. It’s about the speed of AI capability gains outrunning every forecast. If your AI strategy is built on assumptions about what these tools can’t do, pick the one thing your team says AI “can’t do yet” and search for benchmarks from the last 90 days. You might be planning around a limitation that no longer exists.
5. Elevate Your Leadership with AI for the C Suite
This week’s research paints a clear picture: the gap between leaders who understand AI’s real capabilities and those running on outdated assumptions is widening fast. If any of those three questions at the end of the Research Roundup made you uncomfortable, that’s a signal worth paying attention to.
I work with middle-market leadership teams to build AI strategies grounded in what the technology actually does today, not what the vendor pitch deck promises. My Q2 calendar is filling up, but I’ve got a few slots open for executive strategy sessions. If you want to pressure-test your AI assumptions before they become expensive mistakes, hit reply to this email or ring me at 717.868.8735 and let’s talk.
Stay safe. Stay healthy. Be strong. Lead well.
Chad