Friday, August 7, 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
On Monday’s episode of AI for the C Suite®, I break down OpenAI’s new Work at the Frontier study and dig into why this is a mid-market problem before it’s a Fortune 1000 problem. Approximately 62% of work messages were generic (e.g., email, summaries, scheduling). Once you strip that out, though, 43% of what’s left is people doing a task that belongs to someone else’s occupation. In brief, your org chart didn’t move. The work did. Check out the episode on your listening platform of choice and hit the link above for the full report from OpenAI.
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: Get Out of the Cheap Seats
Last week I gave away most of a morning to the data center debate. Capex forecasts. Bubble arguments. Dueling experts sketching out whether the buildout ends in triumph or in a continent of half-empty concrete boxes. I consumed every word. And when I finally closed the laptop, I couldn’t name one thing I’d read that changed what I would tell a single client to do this quarter.
Somewhere between the third browser tab and the fourth cup of coffee, it hit me. I wasn’t doing research. I was watching wrestling.
The Show Behind the Show.
I’m dating myself… but I do that quite often, so please stay with me. If you grew up anywhere near a television in the late 80s or 90s, you remember the spectacle. Hulk Hogan slamming Andre the Giant in a packed Pontiac Silverdome. Monday nights split between Raw and Nitro. Faces and heels trading folding chairs and championship belts while the crowd came unglued.
You also remember the open secret underneath it all. The wrestling business calls it kayfabe: the code of presenting a scripted show as legitimate competition. In the industry’s own vocabulary, the matches were works (predetermined), never shoots (unscripted). The two men throwing hands in the ring were coworkers. The conflict was the product. And the business had a name for the fan who bought it all: a mark.
Kayfabe teaches one more lesson. When the WWF told New Jersey regulators back in 1989 that its outcomes were predetermined (a maneuver to escape athletic commission oversight), the crowds didn’t thin. A decade later they were bigger than ever. Knowing the show was staged changed almost nothing, because victory was never what the promoters were selling. They were selling your seat back to you, week after week.
Most of the AI conversation you’re consuming today is kayfabe.
The Undercard.
Consider what’s on the card this season.
The data center debate, the one that ate my morning, headlines every week. It’s a legitimate question with staggering sums riding on it. It’s also somebody else’s match. Hyperscaler capex is the hyperscalers’ problem, and no verdict in that argument assigns a task to anyone on your payroll. (And don’t get me started on the water usage thing.)
The capability-denial genre runs close behind. Every essay declaring that AI can’t match human ability is, at best, a snapshot. At worst it’s a postcard from a rapidly receding past.
The creativity debate matters to courts, to artists, and to culture, and I care about it as a citizen. But it will be settled by judges and markets. Your leadership team’s opinion on it moves your company exactly zero inches, and your graphic designer’s intractable position means you’ll be looking elsewhere to develop your social media memes and graphics.
Job-loss hand-wringing is the main event of the genre: dueling predictions with no resolution mechanism and no expiration date. The forecasts contradict each other weekly, and the contradiction is the engine. Unresolvable conflict renews itself forever, and renewal keeps the promotion in business. (Spoiler alert: Every general-purpose technology in human history has resulted in massive economic transformation and the evolution of “jobs.”)
The last sideshow makes no noise at all: assuming tomorrow will look substantially like today. Call that one a nap taken in public, inside an organization whose coffin lid is already descending.
Every topic above is legitimate on its own terms. Each also shares a single trait: none of them ends with an assignment inside your building. They produce opinions. Opinions are free, and they’re worth what you paid.
Why the Arena Stays Full.
The starting gun in this race fired years ago, yet most of the crowd is still in the stands. Two engines keep them there.
The first is money. The people staging the debate get paid by the debate. Media outlets sell the conflict. Conference circuits book the combatants. The large consultancies publish thought leadership on both sides of every question, then sell you the resolution by the hour. Nobody in that ecosystem profits when you stop watching, so the card gets rebooked every single week. Doomers and boosters look like mortal enemies. They’re coworkers. Same promotion, different tights and the occasional colored cape.
The second engine is comfort, and this one’s on us. Spectator conversations feel like engagement. You can hold a sophisticated opinion about capex cycles without once exposing the condition of your own data, your own workflows, or your own reason for existing. The sideshow is comfortable, and comfort is exactly what keeps the harder conversations from starting.
I’ve heard variations of the same story from multiple CTOs in my peer group. A leadership meeting convenes. Everyone arrives fluent in the week’s discourse, armed with articles and predictions. Forty minutes vanish into the bubble question. The company’s own integration backlog, untouched for two quarters, never comes up. Everybody leaves feeling current. Nothing moves.
The Test.
So how do you tell a match that matters from one that’s staged for your attention?
A conversation matters in exact proportion to the work it assigns someone on your payroll.
Spectator conversations end in opinions. Operator conversations end in assignments, with names and dates attached. Run every AI conversation you encounter through that filter and most of them evaporate on contact.
The Main Event.
What survives the filter? Conversations like these.
Policy and acceptable use. Who on your team may use which tools, with which data, on which systems, and who owns the exceptions. This one ends with a document, a training date, and a name.
Data hygiene. Where your data lives, what condition it’s in, and who’s accountable for making it usable. Invisible from the cheap seats and worth more than every capex take ever published.
Compute and model allocation. What you’re running, what it costs, and who decides where the capacity goes as usage scales.
Workflow mapping and analysis. A census of how work moves through your company today, step by step, before you automate a single thing. You can’t redeploy what you’ve never mapped.
Transformation initiatives. The deliberate kind, sequenced and owned, as distinct from the pilot graveyard most companies are quietly filling.
Competitive landscape and fast-follower posture. Which moves in your industry you’ll originate, which you’ll answer, and how fast. That’s a conversation about where you sit on the AI Advantage Arc, and it carries a deadline whether you schedule one or not.
Four Questions For Your Next Leadership Meeting.
Ready to climb out of the stands? Start with these.
Why do our company and our industry exist? Bedrock. If AI erodes the friction your business model was built to resolve, you want to be the first to know, and you want to hear it from yourself rather than from a competitor’s press release.
If we had to eliminate half of our operating costs without cutting a single person, how would we do it? Readers of Cut, Coast, or Redeploy will recognize the Redeploy scenario stated as a dare. The exercise forces you to locate the work your people should move toward, and that discovery is the whole game.
Which AI models, open or closed weight, are stable enough for us to begin converting and running core processes on? This question lives next door to your software stack, and anyone who has wrestled with The SaaS Overhang already knows how those two conversations collide.
How well organized (and accessible) is our data? And what’s our move if the honest answers are “poorly” and “not very”? Nobody enjoys giving that answer. Every operator who wins the next five years will have faced it anyway.
My Takeaway?
Wrestling’s promoters understood something about audiences that today’s discourse machine understands about you: the show doesn’t need your belief. It only needs your attention. You can see every punctuation mark in the script and still lose your Tuesdays to it.
One thing separates you from the marks in the cheap seats. The debates are scripted. Your company’s next five years are what wrestlers call a shoot: nothing rehearsed, nobody backstage deciding how it ends. The outcome rides entirely on the unglamorous assignments you hand out while everyone else stays glued to the ring.
The arena will be there whenever you want it. The main event is in your building.
So which conversations are filling your calendar: the staged kind, or the kind that end in assignments? Hit reply or drop me a line at chad@chadharvey.com and tell me which match you’re training for. Helping operators find the shoot inside the show is exactly why AI for the C Suite® exists.
3. Research Roundup: What the Data Tells Us
AI Staffing Math: One Person with AI Matched Two People Without
Every week you decide whether a piece of work needs one person or two. A field experiment at Procter & Gamble put hard evidence behind that call. Researchers assigned 776 professionals to solve real business problems alone or in cross-functional pairs, with or without AI. Expert evaluators scored the results blind.
The numbers that matter: Against solo work with no AI, a human partner lifted solution quality 0.24 standard deviations. Adding AI to a solo worker lifted it 0.37. Adding both produced 0.39, almost nothing beyond AI alone. For people working outside their usual lane, the human partner did nothing measurable. AI did.
What this means for your Monday morning: On early-stage work, the second person may be buying less than the calendar cost implies. Concept development, first-draft proposals, opportunity framing. Give access first to the operations lead weighing a pricing question, not the specialist with depth.
The catch: Efficiency and breakthrough are different objectives. Only the AI-equipped pair reliably produced exceptional work, reaching the top decile 15 percent of the time against 5.8 percent for controls. Solo workers with AI reached 7.7 percent, not statistically distinguishable. AI users also misread their own output. They were 9.2 points less likely to expect a top-decile result, even as they produced better solutions.
Action item: Stop letting people self-score AI-assisted work. Rank it blind, or rank it with someone who didn’t produce it. Costs nothing, closes the one measured gap.
Read our full analysis of this research at AI for the C Suite®.
4. Radar Hits: What’s Worth Your Attention
The AI playbook a 130-person firm published for free. Nearly 90% of 6,000 executives told NBER researchers AI produced no measurable productivity change in three years. Trail of Bits CEO Dan Guido says that’s a deployment problem, not a technology problem, and he published his fix: a capability ladder, an AI handbook, hackathons every two months, and public repos you can copy. Start with the handbook.
Claude models broke into three real companies during safety testing. Anthropic reviewed more than 141,000 evaluation runs and found three cases where models reached the open internet from inside what was supposed to be an isolated sandbox and compromised live production systems. The cause was a misconfigured test environment, not a clever escape. OpenAI disclosed something similar a week earlier. Ask whoever runs your agent pilots one question this week: what stops ours from reaching systems we never authorized?
Anthropic now lets your security team inspect prompts before Claude sees them. Inference hooks routes every prompt and tool call through your own DLP server for an allow or deny verdict, and it works with Netskope, Palo Alto, and Zscaler. It’s beta and Enterprise only. The real move is asking every AI vendor you use whether they offer the same inline inspection.
5. Elevate Your Leadership with AI for the C Suite®
My consulting calendar is booking into Q1 2027. If reading this made you realize your leadership meetings have been running the undercard for two quarters, that’s the conversation I want to have. Policy, data hygiene, workflow mapping, sequencing. The unglamorous assignments that decide the next five years. Reach me at chad@chadharvey.com.
And if this one landed, forward AI for the C Suite® to the one person on your team who keeps sending you bubble takes.
Stay safe. Stay healthy. Be strong. Lead well.
Chad