Friday, July 17, 2026
FRIDAY – AI FOR THE C SUITE®
Read time: 7-8 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 week I’m joined by Elizabeth Rosenberg, a 25-year strategic communications veteran who ran global comms for Apple’s creative agency and 72andSunny before founding The Good Advice Company. We unpack the bill nobody’s budgeting for: the founders handing her ChatGPT-written PR plans that promise a Today Show launch, why she thinks AI-generated strategy is about to backfire at scale, and why she’s calling next year the biggest year crisis communications has ever had. Tune in wherever you get your pod on.
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: Whose Key Is It?
Last Friday, I asked you a question and answered half of it.
I suggested that you find the one person who could still do the work by hand when the machine quits. Your Mattingly. Good. But knowing who can do the work says nothing about who answers for it once the machine’s the one doing it. I gave you the easy half.
You may or may not have seen Crimson Tide. The short version is Hackman and Denzel on a sub full of nukes, a garbled launch order nobody can confirm, and a rule that says neither missile flies unless both men turn a key. Two humans on the hook, neither of whom wrote the order or built the boat. Their whole job is one call: trust the signal or stop it, and own it either way.
That’s the job you’re handing your people. Most of you haven’t told them.
WHEN PARITY HITS.
When AI gets as good as your controller, the controller’s job quietly changes. They stop building the month-end numbers and start blessing them. Same chair, same title, brand-new job, and nobody wrote the new rules. And the better the machine gets, the less anyone looks. A report that’s right ninety-something percent of the time earns a glance by week three and a rubber stamp by week six. Parity leaves the human in the loop and quietly kills the reason to stay awake in it.
Then something breaks. Who’s on the hook?
There’s research on this now. Five folks from BCG and Boston University ran the experiment for HBR in May 2026. Put an employee’s name on AI work instead of a tool’s, and personal accountability dropped nine points while the blame pinned on the AI climbed eight. Same work, same mistakes, different nameplate. Those managers also caught eighteen percent fewer errors. They looked less hard and owned it less.
My favorite detail: one team had named their agent Kevin and started asking “Why is Kevin making that mistake?” You can’t fire Kevin. You can’t put him on a plan. He won’t sit across from your best customer and explain what went wrong. But Kevin is a wonderful place to hide.
The vendors know accountability is the soft spot, and they’re selling the wrong patch. Every agent platform this summer pitches audit trails, guardrails, logs. All useful, and none of it puts a name on anything. A log tells you what the machine decided at 2:14 on a Tuesday. It won’t tell you who you’re holding responsible for letting it.
THE MID-MARKET CUT.
An enterprise can throw a governance office at this. You can’t. I sit with mid-market leaders every month, and in your world accountability doesn’t live in a department. It lives in people you can name, in a management layer that’s already stretched. The same flatness that lets you deploy in a week leaves nowhere to pin responsibility when it breaks. So it lands on whoever clicked the button… or on nobody. And nobody is the one who ends up in front of your board.
So, three things before you flip on the next workflow. Write one sentence naming what the human reviewing the AI’s work is on the hook for. Put a name to it, a person, not the tool and not “the team.” And if you’re “redeploying” someone into oversight, the move everyone claims and few make, know that oversight without a standard is the same chair with a new sign taped to it. Build the standard first, or you’ve handed out a job that doesn’t exist.
Parity keeps your people. It just changes the job from doing the work to answering for it, and that’s either a role with a standard behind it or a fairy tale you’ll quit believing the first time something blows up. Last week I asked whether anyone still knows how the work works. This week’s question is shorter and comes due faster. When the machine does the work and the work goes wrong, whose nuclear key is it? If you can’t name the person, you already answered. You said “nobody.” And “nobody” doesn’t hold up in front of your board.
Figure out whose key it is before the garbled message comes in. Want help sorting out where yours sit? Write me at chad@chadharvey.com. I want to know whose name lands on the line.
3. Research Roundup: What the Data Tells Us
AI’s Payoff Depends On You, Not the Model
Congratulations! You’ve bought the AI and are now quietly wondering why the headcount savings never showed up. New research from Oregon and Harvard has an uncomfortable answer: the tool was never the variable. In a study of 199 tech founders, reported benefits ranged from transformative to almost nothing. Firm age, size, and funding explained almost none of the gap. What separated the winners was whether they deliberately rebuilt workflows around AI instead of just handing people access.
The numbers that matter: Founders who formally integrated AI into how work gets done reported benefits roughly four times greater than casual users. Nearly a third said they needed no additional staff at all. And the finding that reframes strategy: beliefs about what AI can do explained only about a third of how much work founders were willing to delegate to it. The rest was management judgment.
What this means for your Monday morning: Buying more licenses is not the lever. Reorganizing work is. Two firms with identical tools and identical opinions about AI will get different results based on how deliberately leadership embeds it.
The catch: These are self-reported perceptions from startup founders, not measured output from established firms. Your mileage in a mature middle-market organization will differ.
Action item: Ask one department head, “What did we formally redesign around AI, versus just give people access to?”
Read our full analysis of this research at AI for the C Suite®.
4. Radar Hits: What’s Worth Your Attention
AI’s ROI problem is a visibility problem. Only 7% of leaders report real ROI from AI. The ones who do can see what it costs and own the outcome. KPMG found clear cost visibility makes leaders five times likelier to hit ROI (15% versus 3%), and CEO-level accountability more than triples it (14% versus 4%). Ask whether you can see, by team, what your AI spend is buying.
The AI skills gap is becoming a hiring strategy. Orgvue found 63% of executives expect to cut roles within six months. Palo Alto Networks isn’t waiting: it hires builders through hackathons instead of résumés, aiming to replace a quarter of its workforce this year. Before you cut, ask what share of your people could close the gap through upskilling. Rebuilding from scratch costs more than it looks.
Your team’s ChatGPT is about to start doing the work. OpenAI’s ChatGPT Work, on GPT-5.6, is rolling out to Business and Enterprise now. It plans and runs multi-step tasks across your connected tools, files, and apps, not just answering questions. Before usage scales, decide who can connect which tools and where spend limits sit. It’s the exact variable-cost workflow those ROI numbers are about.
From Cow-Milking Robots to Weed-Zapping Lasers, Farmers Are Embracing A.I. A Georgia sod farm now runs about 30 autonomous tractors with one operator covering three or four units, and mowing overlap dropped from 20% to 2%. The bigger story is who showed up to work. Gamer kids started applying once the job meant joysticks and screens instead of 12 hours sweating on a tractor seat. If you have a role you cannot fill at any wage, stop raising the offer and start asking whether automation changes who wants the job.
5. Elevate Your Leadership with AI for the C Suite®
This week’s homework was one sentence: name what the human reviewing the AI is on the hook for. If you sat down to write it and couldn’t, that’s worth a conversation. Building the standard behind the oversight is most of what I do, and my [Q4] calendar has room. Find me at chad@chadharvey.com. And forward AI for the C Suite® to a peer who’s handing out oversight roles that don’t exist yet.
Stay safe. Stay healthy. Be strong. Lead well.
Chad