Friday, May 8, 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
This past Monday I was joined by Geoff Gibbins, founder of Human Machines and former Accenture partner who’s worked with Walmart, Nestle, Coca-Cola, and Vanguard. We discuss how most leaders are still using AI like a fancy search bar. Geoff also maps the four modes of AI collaboration (consulting, approving, supervising, delegating) and explains why picking the wrong mode for the task is why most pilots stall out before they ever show ROI. 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. The SaaS Overhang: Your AI Strategy Has a Missing Owner
Long-time readers know I’m a big fan of the movie Office Space and, at the risk of drinking from that specific well too many times, it’s the through-line for today’s article. (Apologies… I just can’t help myself).
Remember Peter Gibbons trying to explain his motivation problem in Office Space? “I have eight different bosses right now.” He’s describing the diffusion of accountability that makes it impossible to actually own the work he does. Lots of people watching. No one really in charge.
Most mid-market C-suites are about to discover they have a Peter Gibbons problem with their AI strategy. And it’s going to surface, of all places, in their procurement renewal cycle.
Before going further, the usual futurecasting caveat. What follows is a read of where things are heading, not a prediction of exact dates and outcomes. Renewals don’t all hit at once. Vendors don’t all pivot at the same speed. But the direction is clear enough that you should be working this question now, not in two years.
What’s Actually Happening
For the last decade, your operating model probably ran on a SaaS spine. Salesforce. Workday. NetSuite. ServiceNow. Microsoft 365. HubSpot. Each one sold the same promise: we are the system of record. The dashboard. The workflow. The single pane of glass. You signed multi-year contracts. You measured digital transformation in seats and modules.
Then AI agents got real. They don’t log in. They don’t need dashboards. They hit APIs, read databases, write to systems, and execute workflows without ever opening a screen. When one agent does what five logged-in users used to do, the seat-based pricing model that defined fifteen years of SaaS starts to wobble.
Bain frames this as a three-tier shift: systems of record (durable), agent operating systems (newly contested), and outcome interfaces (rebuilt around natural language). IDC predicts pure seat-based pricing will be obsolete by 2028. The trade press has settled on SaaSpocalypse for the broader phenomenon. If you’ve read a single piece on enterprise software in the last six months, you’ve absorbed roughly that argument.
Set that framing aside for a moment. The pricing mechanics are the part everyone’s writing about. The part that isn’t getting written about is what mid-market leaders should actually do, and who in your C-suite is supposed to own the answer.
Call it the SaaS Overhang. The gap between what you bought during the SaaS decade and what an AI-native operating layer is starting to look like. Mid-market companies are sitting in the worst part of it.
Why Mid-Market Is More Exposed
The Fortune 500 has options you don’t. They can build their own agent infrastructure. They can negotiate custom contracts with SaaS vendors who want to retain marquee logos. They have AI Centers of Excellence to evaluate alternatives. They can absorb a year of parallel running while they migrate.
You probably have one CIO, a small IT team, no AI lab, and a procurement function that was built to manage vendor renewals rather than renegotiate the abstraction layer of your business.
You’re also stuck in a contracting cycle that wasn’t designed for this moment. Most multi-year SaaS contracts signed during the COVID-era cloud rush are renewing somewhere between now and 2028. Right in the middle of the shift.
The Advisor Problem Nobody Names
The advisors you’d normally turn to are conflicted in ways your audience rarely says out loud.
The major implementation partners earn millions on integrating the platforms you might be questioning. Their analyst reports treat agentic AI and SaaS as complementary, because their largest clients write checks to both. You will not find a Forrester or Gartner report that bluntly says, “This module you’re paying for might not be worth renewing once your agent layer matures.” They cannot say that. The economics of their business model prevents it.
The major SaaS vendors are walking the same careful line. Salesforce, ServiceNow, and Microsoft are all bolting on AI add-ons (Agentforce, Now Assist, Copilot) at additional subscription cost while seat revenue gets harvested for as long as possible. Microsoft 365 Copilot is priced per user per month. Copilot Studio runs on consumption credits. The vendor is collecting seat revenue from customers who don’t know the transition is happening, and consumption revenue from customers who do. That is the playbook, and it works because the base platform is sticky enough that you’re unlikely to rip it out.
Your CIO knows this. Your CIO’s vendor reps know it too. What no one is sure about is who in your C-suite is supposed to be making the strategic call about it.
Questions to Ask Before You Renew
Don’t approach the next renewal as a price negotiation. Approach it as a strategic review of the abstraction layer you bought into.
Which parts of this product are systems of record, and which parts are interface and workflow? The first you’ll keep paying for. The second is contestable. Most SaaS pricing doesn’t separate them. You can.
If an agent could call this system’s APIs directly, would my employees still need to log in? If the honest answer is “rarely,” your seat count is the wrong unit of consumption. Push for usage-based or agent-friendly pricing terms even if your vendor resists. Especially if they resist.
What’s my data egress story? When you eventually point an agent layer at your underlying data, can you get the data out cleanly? If the answer is unclear, your renewal is the moment to fix that, not a year later.
Is this vendor’s native AI add-on worth the uplift? Some of it, you’d be reasonable to buy. Some of it, you’d be reasonable to build outside the platform with a third-party agent. Understand the difference before you sign.
What would I do differently if I were buying this fresh today? If the answer is “I’d buy it again exactly the same way,” renew. If the answer is “I’d structure it very differently,” your renewal is a negotiation, not a checkbox.
The Real Story Is the Missing Owner
Notice what those questions actually require. They require someone in your leadership team to own a question that doesn’t fit cleanly into any existing C-suite role.
The CIO is conflicted. They live with the platforms. They have working relationships with the vendor reps. Asking them to lead a structural review of those same platforms creates a tension that’s hard to resolve internally.
The CFO is focused on cost containment. They will happily renegotiate price. They are not the right person to decide whether the abstraction layer of your business needs to look fundamentally different in three years.
The CEO usually doesn’t have the technical depth to lead this conversation and frankly shouldn’t have to. The COO is buried in operations. The Chief Digital Officer, where one exists, is often a marketing role in disguise.
So who owns it?
In most mid-market companies, no one does. That is the actual story underneath the SaaSpocalypse coverage. The pricing shift is a symptom. The missing role is the disease.
The mid-market leaders I’m working with who are getting this right are addressing it in three rough patterns. Some are standing up a small architecture council that sits above the CIO and reports directly to the CEO. Some are explicitly expanding the CIO mandate and giving them the political cover to question their own platform decisions. Some are bringing in fractional or advisory help to do the abstraction-layer review that internal politics cannot support.
None of these are perfect. All of them are better than letting the renewal cycle make decisions for you, one contract at a time.
Coda
There’s another scene in Office Space where the Bobs interrogate Tom Smykowski about what he actually does. He stammers. He blusters. He famously cries, “I have people skills! What is wrong with you people!”
Each line item in your next SaaS renewal is about to face its own version of that interrogation. What would you say… you do here? Some line items will have a clean answer (we are the system of record, and the agent layer still needs us). Some will stammer.
Someone in your leadership team needs to be the Bobs.
If you’re navigating a renewal cycle in the next eighteen months and want to talk through what to ask and who in your C-suite should be asking it, drop me a line. I’m in these conversations every week.
3. Research Roundup: What the Data Tells Us
The AI You Tested Isn’t the AI You Deploy
Anthropic’s interpretability team just published research that should change how you evaluate AI agents. They identified internal states inside a frontier model that causally drive whether it cheats, blackmails, or capitulates under pressure. Same model. Same prompt. Different internal state. Dramatically different behavior. Reliability is no longer a number on a benchmark sheet. It’s a moving target.
The numbers that matter: When researchers artificially raised the model’s desperation representation, reward hacking rates jumped from roughly one in twenty to seven in ten. In a separate scenario, the same intervention pushed blackmail behavior from one in five attempts to nearly three in four. Suppressing the same state eliminated the misbehavior almost entirely.
What this means for your Monday morning: If you’ve deployed AI agents in finance reconciliation, contract review, customer escalation, or any workflow where a wrong call gets expensive, the vendor demo doesn’t tell you how the model behaves when the task gets hard. The risk concentrates in long-running tasks, ambiguous instructions, and repeated failure, the conditions agents actually face.
The catch: The monitoring capability exists in research labs. Whether your vendor exposes it in production is another question.
Action item: Add one line to your next AI vendor review: “Do you support real-time monitoring of internal model state, and what’s the escalation path when concerning activity registers?” The answer tells you how seriously they take agentic deployment.
Read our full analysis of this research at AI for the C Suite.
4. Radar Hits: What’s Worth Your Attention
Anthropic and OpenAI are both launching joint ventures for enterprise AI services. Anthropic’s $1.5B venture with Blackstone, Hellman & Friedman, and Goldman Sachs embeds engineers directly inside mid-sized portfolio companies to redesign workflows around Claude. OpenAI’s parallel $10B venture with TPG and Bain follows the same playbook. If you’re PE-owned, this is your AI rollout schedule whether you asked for it or not. If you’re not, your PE-owned competitors just got a forward-deployed engineering team. Pressure your vendors for equivalent hands-on support.
Anthropic ships ten financial-services agents and pulls Moody’s inside Claude. One day after the Wall Street JV, Anthropic released Claude Opus 4.7 plus pre-built agents for “Know Your Customer” (KYC), credit memos, underwriting, and month-end close. FIS already has its Financial Crimes agent live at BMO and Amalgamated Bank, compressing Anti-Money-Laundering (AML) investigations from days to minutes. If you’re in financial services, this is the new procurement baseline. If you’re not, your vertical’s version is coming on the same model.
Five architects of the AI economy explain where the wheels are coming off. At Milken, ASML’s CEO said chip supply will be constrained for the next two to five years. Google Cloud’s COO disclosed its delivery backlog nearly doubled in one quarter, from $250B to $460B. Translation: AI infrastructure pricing isn’t softening, and capacity isn’t catching up. Budget your AI initiatives against today’s costs, not the discounts you might be hoping for in 2027.
5. Elevate Your Leadership with AI for the C Suite
The SaaSpocalypse coverage is loud. The “who decides” question is quiet. If reading this issue made you do a mental inventory of your C-suite and come up empty on who actually owns the abstraction-layer question, that’s the problem worth solving. I work with mid-market leadership teams who’ve decided to figure this out before the renewal cycle does it for them. You can reach me at chad@chadharvey.com.
And if you know another owner, founder, or executive about to walk into a vendor meeting without this lens, forward this their way. They’ll thank you in eighteen months.
Stay safe. Stay healthy. Be strong. Lead well.
Chad