Friday, April 10, 2026
FRIDAY – AI FOR THE C SUITE
Read time: 9–10 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
This past Monday, I was joined by Chris Happ, CEO of Virtuous AI and a serial operator who previously scaled a marketplace to $10 billion in gross merchandise volume. We explore something every CFO should hear: how one retail client used connected intelligence to predict inventory with 95% accuracy and free up several million dollars in trapped cash. All without an ERP overhaul.
Apple · Spotify · iHeart · Amazon · YouTube
Subscribe for free today on your listening platform of choice to ensure you never miss a beat.
2. Algorithmic Musings. Big Trouble in Midmarket: Calling Jack Burton
|
TL;DR If you’re wondering why I’m not leading with this week’s Anthropic Glasswing announcement, it’s because I already walked you through that scenario last Friday. Go back and read it if you missed it. This week we’re doing something different. |
Every rewatch of Big Trouble in Little China eventually teaches you the same thing: Jack Burton isn’t the hero. Wang Chi is. Jack’s the guy with the tank top and the catchphrases and the unshakable belief that he’s the main character, but if you actually track who drives the plot, who has the skills that matter, who saves whom, it’s Wang the whole way down. Jack is the sidekick who talks like a protagonist. He just hasn’t figured it out yet.
If you’ve never seen the movie you need to correct that this weekend. In related news, I think most mid-market CEOs are about to discover they’re Jack Burton.
Stay with me on this one.
The dominant AI narrative aimed at mid-market leaders is some flavor of efficiency. Save hours. Automate workflows. Clean up your data. It’s all true and it’s all table stakes. Four different newsletters told you the same thing this week. I don’t want to add to that pile. The more interesting question, the one almost no one’s writing about, is what AI does to the invisible architecture that actually runs your company.
Mid-market organizations don’t run on their org charts. They run on something underneath the org chart. A quiet, unwritten contract about who gets listened to. The VP of Sales who’s been there since 2011 and whose gut call carries more weight than a spreadsheet. The operations lead everyone routes around the formal process to talk to. The founder who’s still the person in the room who knows the most about the product, the customers, the market, and the history. That informal authority is load-bearing. It’s how decisions actually get made in a company of three hundred people where nobody has time for enterprise governance theater.
Now watch what happens when AI shows up.
A twenty-six-year-old marketing coordinator, on a regular Tuesday, with no particular permission, produces a competitive analysis that would have taken your strategy consultant two weeks and forty thousand dollars. A finance analyst spends a long weekend and builds a working scenario-modeling tool that contradicts the forecast your CFO has been defending for six months. A customer service rep starts handling inbound tickets in Portuguese. An operations lead automates a workflow that your most tenured employee has “owned” for a decade.
None of these people got promoted. None of them got a new title. Nothing on the org chart moved. But the thing that gave the VP of Sales and the CFO and the tenured ops veteran their real authority, the scarcity of what they knew, just quietly collapsed in the background. The price of expertise is falling, and informal authority in mid-market companies is priced in expertise.
That idea should resonate. Because if you’re running a middle-market business, your entire decision-making apparatus is about to face a stress test it was never designed for.
Who proposes a change when anyone can run the analysis? Who gets to challenge the VP of Sales when a coordinator has the receipts? What does the founder-CEO do when being the smartest person in the room stops being the default job description no one calls out and instead starts being a liability? Your authority was built on a scarcity that’s evaporating, and nobody tells you it’s evaporating because the early signs show up in places you don’t usually look. A meeting where a junior person pushed back harder than they used to. An analysis that came from the wrong department. A decision that got made without you.
Mid-market leaders are more exposed to this than enterprise leaders are. I know that sounds backwards. Enterprises have compliance departments and formal governance and decision-rights matrices that, whatever else you think of them, at least answers the question of who gets a voice. You almost certainly don’t. Your governance is tribal. Your decision rights are vibes. And tribal governance works beautifully right up until the tribe’s hierarchy of knowledge gets scrambled by a tool anyone can download on a Tuesday afternoon.
So what do you actually do about it?
Start by naming the informal authority structure out loud. Who actually gets listened to in your company, and why? What’s the real reason, not the title-on-the-door reason? Then ask whether the “why” is something AI is about to erode. If the answer is yes, you have a choice to make, and you have it now rather than in eighteen months when someone else makes it for you.
You can redesign the authority structure deliberately, which means hard conversations about what seniority means, what expertise means, and what it takes to be heard in a room where the old rules no longer hold. Or you can let the redesign happen on its own, which it will, on a timeline you don’t control, driven by whichever of your people figures out the tools first and stops waiting for permission.
One of those paths ends with you still running the company. The other ends with you playing Jack Burton in someone else’s movie.
If you want to go deeper on why this hierarchy disruption is structural rather than cyclical, my conversation with Brent Orrell published on February 23, 2026, Senior Fellow at the American Enterprise Institute and former Acting Assistant Secretary of Labor, is the one to pull up. Orrell’s line is that AI “rips hierarchies apart,” and he backs it with a P&G study where AI-equipped teams produced better work but trusted it less, and teams without AI produced worse work and felt great about it. That trust gap is the exact mechanism this piece is describing, just seen from the workforce-research side of the glass.
The reflexes line from the movie still applies, by the way. Jack gets it half right. It is all in the reflexes. He just has the wrong ones for the movie he’s actually in.
Which of your informal authority structures is most exposed right now? Drop me a line at chad@chadharvey.com and tell me where you’re seeing it crack. I read every one.
3. Research Roundup: What the Data Tells Us
The AI Metric Your Vendor Won’t Show You: Team Readiness
Your AI vendor’s accuracy score is measuring the wrong thing. New research out of Singapore Management University, presented at CHI 2026, argues that the real risk in AI deployment isn’t the model at all. It’s the gap between what the model can do and what your people actually do with it.
The numbers that matter: The research documents three failure modes hiding inside standard AI scorecards. High-accuracy models routinely cause users to flip correct decisions to incorrect ones after seeing AI output. Survey-reported trust barely predicts actual reliance behavior under deadline pressure. And strong evaluation-phase performance often reflects copying, not judgment, meaning the wheels come off once the training wheels do.
What this means for your Monday morning: If you bought an AI tool based on vendor benchmarks, you bought half the picture. The other half, whether your team overrides the AI when it’s wrong and trusts it when it’s right, is sitting in your interaction logs right now, unmeasured.
The catch: Tracking this requires logging initial human decisions alongside AI recommendations and final outcomes. Most deployments skip this on day one and can’t reconstruct it later.
Action item: Ask your AI vendor for team performance evidence, not just model accuracy. If they can only show you benchmark scores, you’re making a deployment decision on incomplete data.
The AI Disruption You’re Waiting For Isn’t Coming (Something Bigger Is)
Forget the “which jobs will AI replace” debate. MIT FutureTech just published the largest real-world study of AI workplace capability to date, and the finding flips the conventional narrative: AI isn’t a crashing wave hitting a few vulnerable industries. It’s a rising tide lifting across nearly every domain of text-based work at a remarkably steady pace.
The numbers that matter: Across 17,000+ expert evaluations of real workplace tasks, frontier AI already completes 50% to 75% of text-based work at a standard a real manager would accept, no edits required. Success rates are climbing 8 to 11 percentage points per year, and the amount of work AI can handle is roughly doubling every four months. If the trend holds, most text-based workplace tasks cross the reliability threshold by 2029.
What this means for your Monday morning: You have a three-year runway, not an indefinite one. And here’s the counterintuitive procurement lesson buried in the data: newer model generations beat larger models on complex, multi-step work. Stop chasing the biggest model and start staying current with generations.
The catch: High task-level success doesn’t equal workforce displacement. Last-mile implementation, governance, and oversight still mediate real outcomes… especially in legal, clinical, and compliance work.
Action item: Before your next leadership meeting, inventory which workflows in your company are primarily text-based and self-contained. That list is your 2026-2029 roadmap.
Read our full analysis of this and all other analyzed research papers at AI for the C Suite.
4. Radar Hits: What’s Worth Your Attention
Anthropic says its newest model is too powerful to release after it broke containment in testing. Claude Mythos Preview autonomously found a 27-year-old OpenBSD vulnerability (and thousands of other zero-days across every major OS and browser), so rather than ship it publicly, Anthropic is routing it to 11 founding partners (AWS, Apple, Google, Microsoft, JPMorgan, Cisco, Broadcom, CrowdStrike, Nvidia, Palo Alto Networks, and the Linux Foundation) through a defensive program called Project Glasswing. As I flagged last week, offensive AI has arrived ahead of your patch cycle. Ask your security lead how fast you can actually act on a zero-day your vendors haven’t disclosed yet.
Anthropic launches Claude Managed Agents to run production agents for you. The new service takes over the infrastructure layer (sandboxed execution, credentials, scoped permissions, session tracing) that has kept most companies stuck in agent pilot purgatory. Pricing is standard token rates plus $0.08 per active session-hour. Notion, Rakuten, and Sentry shipped production agents in under a week. If your team has been building agent plumbing from scratch, that work just became a commodity. Revisit your build-versus-buy math before the next sprint.
5. Elevate Your Leadership with AI for the C Suite
If reading this made you mentally flag one person in your company whose authority is about to get stress-tested, that’s the conversation to have next… before it happens on someone else’s timeline. While my Q2 is booked, I’ve got two intro call slots open in May for middle-market leaders who want to map their informal authority structure before AI maps it for them. Reply to this email to grab time with me. And, if this piece named something you’ve been feeling but couldn’t articulate, forward it to the peer who needs to read it. That’s how this newsletter grows, and I’m grateful every time.
Jack Burton’s still my favorite. But I’d rather you be Wang Chi.
Stay safe. Stay healthy. Be strong. Lead well.
Chad