Friday, May 15, 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. This week I’m highlighting the intersection of AI and core values. Here’s what you need to know:
1. Sound Waves: Podcast Highlights
This past Monday on AI for the C Suite®, I dropped a solo jargon watch on the term most executives haven’t yet heard but will define the next two years of competitive positioning: harness. New analysis pins 65% of enterprise AI failures on harness defects, not the model itself, which means two companies running the same frontier LLM in 2026 will have completely different reliability, P&Ls, and moats. It’s a technical distinction that matters, so tune in for that plus 14 other AI terms to know in ’26.
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: Cut, Coast, or Redeploy
AI isn’t a full-blown replacement for people… yet. But it can do an awful lot of what people used to do.
That sentence is the honest starting point for any C-suite conversation about workforce strategy in 2026. The capability question is settled enough. Your customer service reps, your analysts, your associates, your operations team, your finance staff. Each of those roles has measurable pieces that AI now does competently. Not all of the role. Just pieces. The pieces vary by function, by industry, by company. The pattern doesn’t.
What’s not settled is what you do about it.
The current consultancy debate fixates on framing. Treat AI as a tool. No, treat it as an employee. No, treat it as a colleague. A recent HBR piece authored by BCG researchers ran a 1,200-manager study finding that anthropomorphizing AI as an “employee” reduced error identification by 18% and shifted accountability away from humans by nine percentage points. Useful research. Genuinely useful. It answers an organizational design question. The workforce decision is something else entirely.
The workforce decision is sitting on your desk. AI just absorbed some percentage of what one of your people used to do. Now what?
There are three responses. Each is sometimes right.
Cut. Eliminate the role, book the savings, exit the relationship. This is the path of least resistance for publicly traded companies whose quarterly reports demand visible operating leverage. The math is uncomplicated and the optics are clean. Wall Street typically rewards it. So it’s what happens.
Coast. Leave the role in place, let the human keep doing the work they’ve always done, layer AI on top where it helps and ignore the inefficiency where it doesn’t. Business press dismisses this option as defeatist. I disagree. For workers within a few years of retirement, who built their careers on craft, whose institutional knowledge is the moat your organization has, coast can be the responsible answer. I’ve watched advisors and CTOs in my groups quietly make this call for specific senior staff. Those calls reflect a judgment that some careers should finish on their own terms. It also unlocks opportunities to map the value and experience that human brings to the organization. And there’s real long-term value in that exercise.
Redeploy. Take the time recovered from the AI-absorbed task and rebuild the role around what the human is now uniquely positioned to do: judgment work, relationship work, exception handling, customer ambiguity, cross-functional translation. This is the option the consultancy decks advocate for in theory and that almost nobody operationalizes well in practice. It’s expensive in management attention. It requires you to genuinely know what your people are capable of beyond their current job description. Which most organizations do not. The job descriptions in your HR system describe roles, not capacity. The roles your people currently occupy describe what they were hired to do years ago, not what they’re now capable of becoming. Redeployment starts with the uncomfortable admission that your HR data is mostly historical fiction.
What redeployment actually looks like, when it works, is uncomfortable in ways most operating playbooks aren’t built for. Audit capacity rather than role: figure out what each person can do that you’ve never asked them to do. Move the human onto exception handling, judgment calls, and the customer ambiguity AI can’t navigate. Measure them on outcomes you couldn’t have measured a year ago, not on the throughput metrics from the job they used to do. That’s three sentences of guidance. The execution is several quarters of management attention you won’t get back.
Publicly traded companies will mostly default to Cut. The structural incentives produce the result. Boards don’t have to be cruel to get there. Their incentive architecture rewards near-term cost compression and punishes the multi-quarter ambiguity that thoughtful redeployment requires. Their boards are not built to absorb the explanation. So they don’t try.
Mid-market firms have something their public peers structurally lack: a window of genuine choice. You can run a more patient calculation. You can absorb a couple of soft quarters while you rebuild a finance team around AI-augmented decision support. You can hold onto a senior account manager whose AI-resistant relationships represent more enterprise value than any restructuring spreadsheet would suggest.
The window only stays open if you walk through it.
Which brings me to The Hudsucker Proxy, the 1994 Coen Brothers film worth revisiting as an AI workforce metaphor.
Sidney Mussburger, played by Paul Newman, runs Hudsucker Industries’ board. After the chairman’s untimely defenestration, Mussburger’s plan is to crash the stock so the board can buy it back cheap. He needs a CEO so visibly unqualified that the market will panic. Enter Norville Barnes, a mailroom clerk played by Tim Robbins, whom Mussburger installs precisely because he’s expendable. Norville carries a piece of paper with a circle drawn on it. Mussburger doesn’t ask what the circle is.
The circle is the hula hoop.
The thing the board treated as disposable headcount turns out to be carrying the next decade of the company’s value creation. Mussburger’s contempt for Norville reads as strategic insight. Look closer and it’s accountancy thinking dressed up as strategy. The man optimizing for short-term board outcomes literally cannot see the asset standing in front of him.
You have Norvilles in your mailroom. So do I. So does every mid-market organization I’ve worked with. Not every employee is a Norville. That’s the honest part. But some are, and your current operating model probably can’t distinguish. The defining question of this AI moment is whether you have the management bandwidth and values clarity to find them before you reflexively cut them.
A while back I wrote about the Core Values Pyramid, a five-level model I developed that ranges from None through Wallpaper up to Embodied. Most organizations sit in the middle, with values that are written, posted, and quietly ignored when commercial pressure arrives. AI is now that commercial pressure, and it has arrived.
How you respond to the capability gain is the most honest values test your organization will face this decade. Every employee will watch. They will know which people got cut, which got coasted, which got redeployed, and what the pattern says about who actually matters here. You will not be able to talk your way out of the pattern with town halls or values posters. The decisions will speak.
If your values turn out to have been Wallpaper, your best people will figure it out within a quarter. They will start looking. And the workforce capability you needed for the redeployment option, the institutional knowledge and judgment work and relationship capital, will walk out the door on its own schedule.
My takeaway? Mussburger’s mistake at Hudsucker Industries was a vision failure. He optimized for the metric in front of him and missed the asset already in the building. The board never had a values problem to solve. They had a values problem they couldn’t see.
The mid-market AI decision is the same shape. Cut where it’s honest. Coast where it serves the human and the business. Redeploy where the latent capability is real and you have the management discipline to develop it. Before any of that, ask yourself the harder question.
Who’s the Norville in your mailroom, and would your organization recognize them if they walked in carrying a circle on a piece of paper?
3. Research Roundup: What the Data Tells Us
Whose Values Is Your AI Actually Running On?
You’ve spent years getting clear on what your company stands for. So here’s a question worth your attention: when you deploy AI, whose values is it actually following? Most AI tools ship with values baked in by the vendor. New research from Peking University shows you can keep your values in a separate, swappable layer instead.
The numbers that matter: Across four different AI models, this approach left only about one in five bad outputs slipping through, roughly a 70% reduction versus no intervention. The next-best method let more than one in four through. Output quality held up, with manageable added cost.
What this means for your Monday morning: Think of it like the difference between hiring people already trained somewhere else versus running your own people through your own playbook. Most AI today hands you somebody else’s playbook. Modular value alignment lets your company’s values, your industry’s standards, and your jurisdiction’s rules live in a layer you control, without rebuilding the system every time something shifts.
The catch: Even the best result here still means roughly one in five bad outputs gets through. That’s real progress, not a finish line. Values guide the system. They don’t replace human judgment on anything that touches your customers, your patients, or your money.
Action item: Ask your AI vendor this week: “Are the safety rules locked into the model, or can we plug in our own values and update them?” The answer tells you whose company is really running.
Read our full analysis of this research at AI for the C Suite®.
4. Radar Hits: What’s Worth Your Attention
Who Benefits From AI? New Studies Offer an Answer. Census, Anthropic, and a new Federal Reserve analysis converge on the same finding: workers in the most AI-exposed jobs earn 47% more on average, and 53% of high earners say AI made them more productive versus 30% of low earners. The pattern for executives: the companies widening their lead aren’t buying more licenses. They’ve broken work into tasks and matched the right tasks to the right tools. If your AI rollout is mostly procurement, you’re losing ground to firms doing workflow redesign first.
Behind the Scenes Hardening Firefox with Claude Mythos Preview. Mozilla shipped 423 security fixes in April, up from a baseline of 20-30 a month, after pointing AI agents at their own codebase. One run alone caught 271 bugs in a single Firefox release, including flaws that sat undetected for 15 and 20 years despite extensive fuzzing. Defenders just got a step change in capability. So did attackers using the same tools. This is exactly what I wrote about on April 3, 2026. Worth asking your software vendors and security team this week: are you running AI-driven code audits, and if not, what’s the timeline?
5. Elevate Your Leadership with AI for the C Suite®
If you’re staring at the Cut/Coast/Redeploy decision on a specific team and want a peer-level read before you commit, that’s the conversation I make time for. My Q2 calendar is committed, but I’m scoping a small number of Q3 engagements now. Start a thread at chad@chadharvey.com and we’ll see if there’s a fit.
If this issue gave you something to take to your leadership team, forward AI for the C Suite® to one peer wrestling with the same workforce calls. That’s how this list grows.
Stay safe. Stay healthy. Be strong. Lead well.
Chad