The Ready

Roughly 5% of integrated AI pilots generate measurable ROI. I see that number cited everywhere, usually to support some version of "AI is overhyped." But that's the wrong read. The question isn't whether AI generates value. It's whether your organization knows how to act on what it learns. Most don't. And they didn't before AI showed up either. Johnson & Johnson ran nearly 900 generative AI pilots. Their CIO looked at the data: 10-15% of use cases were driving about 80% of the value. So they killed most of the pilots, dissolved the central governance board, and concentrated resources on a handful of bets that actually worked. Moderna built their own internal AI tool, ran a direct comparison against ChatGPT Enterprise, found they'd been outperformed, and shut their own product down. PepsiCo ran broad pilots, named four or five big bets, and stopped resourcing everything else. These aren't stories about lucky technology choices. They're stories about organizations willing to use what they learned. Each one required someone to look at evidence and then kill something that had real organizational weight behind it. Most organizations struggled with this long before generative AI arrived. Sunk-cost thinking, risk-averse leadership, no real mechanism for cutting what isn't working. The pilot runs for six months and the data comes back thin. Instead of stopping, someone extends the timeline, adjusts the success criteria or quietly moves the goalpost. The evidence is clear but killing it means admitting the bet didn’t pay off, and most organizations don’t have the muscle for that. AI didn't create that pattern. It inherited it and makes it more expensive. Stanford's Digital Economy Lab published a study last month looking at 51 successful AI deployments across 41 organizations. 61% of successful projects were preceded by at least one failed attempt. They named "proof-of-concept factories" as the dominant failure mode: organizations where experimentation never converts to value because no one does anything with what they learned. Experimentation is how you find out where to focus and invest. The companies getting traction aren't better at running pilots. They're better at using what the pilots reveal, including when that means stopping.

Our impression of how most change rollouts sound: "Shipped it! I'm available for questions. Thanks!" Effective rollouts look different than the moves we've inherited. The deck goes out. The Q&A gets scheduled. A calendar invite appears with three !!!s in the subject line. Announcements, no matter how enthusiastic, don't move people. Rodney Evans has a different approach. It's 5 moves that each serve a clear purpose. 1️⃣ Early Signal Telegraph the tension months before a proposal surfaces, because people need time to feel the weight of a problem before receiving a solution. If the first time someone hears about a structural change is the day you roll it out, you've handed them an answer to a question they haven't asked yet. Think about how long it takes you to make your own decisions. The weeks spent sitting with something and turning it over. You can't compress that timeline by announcing change harder. 2️⃣ Origin Story When the new thing finally arrives, most people lead with the asset. Instead, lead with the reason. Share a video, a conversation, or a piece of writing that explains why this work matters and what it means to those doing it. People resist change when they don't know why it's happening, or they sense the why is being kept from them. An origin story does two things at once: it hands people the meaning behind the change, and it signals that the people leading it believe in it. 3️⃣ Specific Ask The launch itself should include a request that's purposefully narrow. Not "Feedback welcome." More like: "Here's a proposal. We're governing it on Friday. I need your eyes on slides 4-6 with this client persona in mind." Vague asks create anxiety; specific asks create movement. Most rollouts skip this step entirely, or water it down into an open-door policy nobody walks through. 4️⃣ Interactive Forum Give people a place to show up and be heard. This can be generative ("What should this become?") or governance-oriented ("Are we aligned on this?"). The format matters less than the fact that it exists. You're creating a moment where people's reactions have somewhere to go besides a Slack DM. The trap most organizations fall into is treating this as a checkbox. A forum only works if people believe their input will affect something. 5️⃣ Ongoing Reinforcement Launch isn't your finish line. It's your halfway point. After a change lands, narrate the adherence back. Make the change visible by pointing to the people who are living it: "Here's a team that used the new approach on a real client problem. Here's what happened. Here's what it looked like." This is what keeps a change from quietly dying post-launch, when attention drifts and old habits reassert themselves. People need proof that change is real. That others are doing it and it's not a blip. Change that lands the first time is better than an announcement you have to repeat 7 times. The arc laid out above isn't extra work. Skipping it is.

We Oprah-ed AI at work. You get AI. You get AI. You get AI. Broad deployment, everyone in, build capability, do more with less. Oprah could afford that. Most organizations are about to find out their pockets aren’t endless. There's the financial cost, which is getting very real very fast. But there's also the cost that comes with change, with the distraction AI has created, and the cost you impose on people when you ask your workforce to continuously adapt to tools that keep shifting underneath them. None of that got factored in when the urgency narrative made everything feel inevitable. I've been in a lot of client conversations lately about AI. Almost none of them start with "We have a problem and we think AI can help us solve it." They start with "We know we need to be doing something with AI." I often ask, “To what end?” and am met with “That’s a good question.” Uber is becoming the cautionary tale I see referenced everywhere. They burned through their entire budget by April and now the CTO is back at the drawing board. By most accounts, they were deploying broadly, incentivizing engineers to use AI, and rewarding adoption as the metric without a clear answer to what all of that was supposed to produce. That's not a story about AI failing. It's what happens when speed becomes the strategy and cost assumptions turn out to be wrong. We told ourselves AI would make us better, faster, stronger. In pockets, it has. I've felt it in my own work. I see it in my colleagues. Individual AI that makes people sharper is real, cheap, and genuinely hard to meter. That might be the most defensible investment organizations have made. The financial pressure could force a question most organizations never actually answered: What does AI need to do for our specific business, in a way that's worth the full cost of getting there? The organizations that rushed hardest to get ahead may have the most catching up to do. The ones that stayed deliberate have something most don't right now: room to choose. The ROI question looks different now that the investment is real. What actually justifies the cost? 

Communicating change and exercising authority over change are different things. Organizations do a lot of the first and very little of the second. They treat clarity about consequences as unkind or heavy-handed, when people often experience it as a relief. Here's what usually gets said: "We're migrating to a new sales tool. Here's why it's better and here are resources to help you get started." Here's what rarely gets said: "Starting Monday, your metrics come out of the new tool. The old one goes dark on Friday at 5 pm. That's the timeline." The second version doesn't require anyone to love the change. It just requires them to make it. One reason why the first version keeps failing: Organizations aren't good at protecting people's time to think about work and structure their days, so even employees with the best intentions (who *want* to watch the tutorial) will deprioritize a vague ask indefinitely. Clear deadlines remove the cognitive weight of wondering whether you've done enough. When you're honest about what kind of change you're making—needs hearts and minds vs. needs execution—the communication becomes more straightforward. Clarity isn't a demand. It's a gift. People would rather know the hard deadline than live with a nagging sense that they're supposed to be moving in some direction but aren't sure if they're doing it right. If you're leading a change right now, ask yourself: "Have I communicated it or have I actually led it?" They're different acts, and only one helps things move forward.

Overheard in a meeting: “I’ve never met a Chief of Staff who didn’t do the bidding of an individual leader.” It got a reaction. Many people have noticed or experienced this dynamic. If “every system is perfectly designed to get the results it gets” (Demming), that got me curious: How are Chief of Staff roles designed? I jumped into LinkedIn and went down a bit of a rabbit hole. These accountabilities stood out to me: “Work directly with the CEO” “Increase the CEO’s effectiveness” “Be the trusted extension of the CEO” “Serve as the President’s strategic advisor” “Manage and protect the CEO’s calendar” The role is often framed around proximity to a leader. But what do you actually want your Chief of Staff to do? Do you want them to be an extension of one leader, or help the organization move amidst complexity? If it’s the latter, the role has to be designed differently. Starting with the job description. When the role is too anchored to one leader, a few patterns can emerge: - The loudest need gets prioritized over the most strategic one - The organization gets interpreted through a single perspective - The Chief of Staff becomes highly reactive instead of system-aware When I stepped into this role, I had my own conditioning about what a Chief of Staff does, shaped by the job descriptions I’ve considered and examples I’ve seen. But in this role, I’ve noticed a shift in my own work when I stopped thinking about myself as a Chief of Staff to a team and started thinking about myself as a steward of the operating system. I lift my gaze. I spend less time responding to what just surfaced and more time sensing what the organization needs in order to move toward its goals. I spend less time on “fire drills,” and I spend more time shaping the operating conditions: designing clear ways of working, governance, and operating rhythms so we all can work better. And I can do that because my role is designed to: - Protect important work, even when urgent requests appeared - Look across the business instead of within one domain - Pay attention to patterns, friction, and feedback loops (not just immediate problems) - Design structures and rhythms that help the organization move more effectively (rather than solving one problem at a time) A durable organization designs roles that can evolve as it does; it has roles built not for an individual, but instead for a system. That means the job descriptions have to look different. What accountabilities would you prefer to see in a Chief of Staff job description? ⬇️

I’ve done this. You’ve done this. The smartest boss you’ve ever worked with has done this: You announce a change, do that clappy gesture with your hands to signal "job done," and wait for everyone to cheer and shift what they’re doing in under 24 hours. That's never how it goes. That’s not how change works. (But to quote Matthew McConaughey, it'd be a whole lot cooler if it did.) Sam Spurlin recently broke down this breakdown on The Ready’s podcast in a helpful way. It looks like this: Sharing information doesn’t imply people will get the info you’re sharing. And people getting the info you’re sharing doesn’t imply they’ll change their behavior. But we believe pulling off step 1 means we’re also pulling off step 3. If you’ve ever spent 13 hours preparing a perfect announcement (of a new tool, a new strategic direction, or a new program) only to be met with crickets and side-eye from both your team and your customers, check out the episode: https://lnkd.in/esebfDJB Skepticism of change probably isn't an indictment of the idea itself. Your update could be wanted! needed! appreciated! and people will still struggle to get down with it. But there’s a way through the crevasse of resistance. It just has more to do with how you change and less to do with what you’re changing.

The bill has arrived. AI labs have been selling compute at a fraction of its actual cost by using a pricing model that favored adoption. That era is ending now. Anthropic is splitting subscription usage into metered pools starting June 15. Legacy enterprise seats are being retired and new contracts bill token consumption at the real cost. GitHub's flat-fee Copilot model dies June 1. This was always going to happen. Hook organizations on cheap access, make AI indispensable, get it embedded in how work actually gets done, and then start charging what it actually costs. It worked. Uber burned through its entire 2026 AI budget by April. Their CTO is back to the drawing board. I'm willing to bet they're not an outlier. This changes your strategic picture. The economic model now has to be a key part of the design. You can't build workflows around AI, hire around AI, or commit to a vendor without knowing what it costs at scale and what it's actually worth. Which means the CFO question is coming: What are we getting for this? Seat utilization isn't an answer. Usage rate isn't an answer. "Our people are more productive" isn't an answer when the bill triples. If you've been running deliberate experiments (think: scoped bets, clear hypotheses, actual measurement on impact), you're probably in a decent position. You know where the value is and where it isn’t, and are making decisions accordingly. If you haven't, now is the time to start to figure out where AI actually moves the needle for your organization. That's the only answer that holds up when the economics get real. And if you've been moving slowly with light enablement, with a wait-and-see posture, you might have been more right than you knew. The organizations that didn't build load-bearing workflows on subsidized pricing have more room to be deliberate now.

One pattern plagues every organization: Getting anyone to do anything new. You’ve got high conviction but low uptake, so something’s getting stuck. You’ve seen this movie before: You choose a new sales tool. There’s an announcement, there are training videos, and there are office hours. The swap should be simple. Stop using tool A and start using tool B. But some people waver. Some read the announcement email at 1:15 am and forget about it. Besides, log-ins to tool A still work. It’s what people know, so they don’t move right away. They’ll get to it, eventually. Maybe. Swap "tool" for "strategy," "structure," or "hiring platform" and the pattern holds. But resistance to a new comms stack is one thing. Resistance to your organization’s pivot away from its core business is higher stakes. The way we believe change happens is flawed. It’s why organizations still try comms-ing, stakeholder-ing, project managing, training, and socializing things to death. But if those moves reliably worked, this pattern would be broken by now. What does work? In this week’s podcast episode, Rodney Evans and Sam Spurlin pull apart what we assume about change and explore the moves that actually get people moving. Hit play now: https://lnkd.in/edR5RY8s

In case you missed the million articles on this: Anthropic and OpenAI just announced they're standing up entirely new companies to embed engineers directly inside enterprise business units. The concept: use what they’re calling Forward Deployed Engineers to identify high-value AI opportunities, redesign workflows, and capture measurable ROI. Honestly, I think the model is compelling, because technical expertise has been one of the real gaps in enterprise AI adoption. It’s one we wrestle with at The Ready as a small company. The "everyone will be a builder" narrative has always felt like a bit of a dodge. I’ve spent days vibe coding my way into something I couldn't figure out how to replicate or maintain when a connection broke. It’s not a skill most orgs are going to develop at scale, and I’m not sure we’ve been honest enough about that. (Cue the comments about how that’ll be the differentiating skill for who gets to keep their job in the future.) Bringing actual technical horsepower into the BU makes sense. What I am stuck on is that both companies are explicitly naming adoption as the primary challenge, and their solution is to increase deployment speed.   That’s not an adoption strategy. Confusing deployment and adoption is going to cause a lot of harm. Getting a workflow redesigned faster doesn't mean the organization will actually change how it works. That’s a different problem and a harder one to solve. The FDE can build the thing. Whether the thing sticks is still up to the humans around it and whether the org makes it possible for those humans to work differently. An FDE isn’t going to change that. Neither can McKinsey, whose involvement with OpenAI, I’d flag as a reason for skepticism rather than confidence. Most people I know who've engaged with the big consulting firms walk away with a nice plan on paper and weak implementation. I’m not sure having them at the table makes the adoption problem any easier. Curious what people here think: Is this the future of enterprise AI adoption? Or just another way for the companies who stand to benefit most from adoption to make us reliant on them?

Put your organization under a microscope for a moment: Does work get done through strategy, operations, or relationships? Here’s a visual 👀 explaining this breakdown. (H/T to Andrea Robb for sharing this idea with Rodney Evans a while back.) When strategy is your clearest organizing principle, work moves forward because everyone understands the direction, trade-offs, priorities, and what to optimize for. But if your strategic direction is muddled, work moves forward through operational muscle. People double down on processes, they rigorously plan, and they run hard at execution. If operational norms are clear and healthy enough, progress is still possible. But if your operations aren’t codified, work moves forward through relationships. Think back-channeling. Favor-trading. Influence campaigns. Reading the room. Knowing someone who knows someone. Relationships at work matter. But they shouldn’t be the only lever you can pull to get things done. When your strategy is humming, you stop hearing "Let me just check with so-and-so first" and you start seeing people move. That’s the difference between an organization that runs on clarity and one that runs on access.