1. The High-Voltage Gap Between Aspiration and Readiness
There’s a line that Genpact CEO Balkrishan “BK” Kalra shared recently that has been stuck in my mind ever since: “Aspirations are really high. Readiness is low.” It’s one of those deceptively simple statements that captures the exact moment so many companies are living through right now. Everywhere you look, leaders are excited about what AI could do. They read the headlines, they see the demos, and they imagine a world where intelligent software is quietly making their business faster, smarter and more efficient. But then they look inside their own organizations and realize that the ground beneath them isn’t quite ready for all that ambition. Kalra made these remarks during a Newsweek “AI Impact Forum” webinar, and he wasn’t there to celebrate progress. He was there to tell a hard truth: the era of pilots and proof-of-concept experiments is over. You can’t keep running small tests forever and call that a strategy. The hard part, Kalra argued, is no longer whether the models are capable enough — it’s whether the enterprise is ready to actually absorb them. The technology may be brilliant, but data is still scattered across disconnected systems, business processes still vary from team to team and market to market, and employees still don’t always understand the tools well enough to help redesign how the work should get done. A more powerful model doesn’t fix any of that on its own. Kalra put a name to these frictions that may sound familiar to anyone who has spent time in the messy middle of a digital transformation: data debt, process debt and talent debt. These are the quiet forces that keep companies from turning a great demo into a real business outcome. It’s not that the AI is broken. It’s that the organization around it isn’t ready. And until that changes, even the most capable technology will struggle to make it out of the pilot lane.
2. The Consent Problem No One Saw Coming
If there’s one place where that readiness gap becomes painfully real, it’s in the quiet, unglamorous world of customer consent. Adam Binks, the CEO of Syrenis, a company that builds enterprise consent and preference management software, has seen a very specific version of this problem in recent months. The issue isn’t that companies lack customer data. Most large organizations have years’ worth of contact histories, purchase records and behavioral signals sitting in their systems. The problem is that the permission attached to that data may not actually cover the new ways they want to use it. Binks described sitting in meetings where an AI project stalled for months, not because anyone found a specific problem, but because nobody could say with certainty that a problem didn’t exist. He told the story of a retailer that had built a sophisticated AI-driven outreach model using years of customer contact data. The model was ready. The audience was defined. And then, right before launch, someone asked the question that stopped everything: did the permissions collected for earlier campaigns actually allow this? The answer, it turned out, was no. Nobody had acted in bad faith. The permission simply did not stretch that far. The right lesson here isn’t that companies should stop using customer data. It’s that consent is not a one-time event. It’s not a checkbox that gets clicked once and then forgotten. It’s tied to a purpose, and purposes change. When a customer gives you information for one thing and you want to use it for something new, you have to go back and have an honest conversation. Binks noted that the retailer in his story eventually did exactly that. They went back to customers with a plain-language explanation of the new purpose, relaunched with a smaller audience, and the smaller audience actually performed better than the original list would have. That’s the deeper insight here: when you respect the boundaries of permission, you often end up with better, more engaged customers. The same principle applies to how enterprises are structured. Binks observed that most large organizations designed their technology stacks around departments, products and individual business problems rather than around the customer. Marketing has one system. Customer service has another. Every acquisition adds another platform, and every platform thinks it holds the definitive record of what the customer chose. Even the meaning of something as simple as “do not contact” can vary across contexts.
3. The Ordinary Work of Making AI Ethical and Governable
What makes this so difficult is that consent and preference management is not just a technology problem. It’s a governance problem, a privacy problem and a business problem all at once. Binks pointed out that often what fails is the space between functions. Privacy assumes technology has implemented the rules. Technology assumes marketing understands the permission. Marketing assumes the data it receives is safe to use. And somewhere in that chain of assumptions, things fall through the cracks. The answer, as he framed it, is to make consent a living part of the workflow itself. AI systems should check the current consent state before using data or taking an action, rather than relying on a historic copy of the record. They should be able to trace where data traveled, which systems used it, and whether later customer choices were applied across all of them. And when an AI agent takes an action — say, predicting that a customer is likely to churn and initiating a retention offer — there need to be clear boundaries about what it’s allowed to do, what information it can use and when a human has to step in. This is all deeply unglamorous work, but it’s the work that separates companies that are genuinely ready for AI from companies that are just talking about it. And as Binks pointed out, the organizations that get this right end up using more of their data, not less, because they can prove what they’re allowed to do and stop arguing about it internally. It’s a useful reframe. Too often, consent is seen as a constraint, a way of limiting what you can do with your biggest asset. But the real competitive advantage comes from being able to move quickly because you know exactly where you stand. That same tension between aspiration and readiness will be on display in the coming weeks in two upcoming Newsweek webinars. The first, presented by Cognizant, focuses on overcoming barriers to AI transformation in legacy industries, with leaders from Yahoo, Qualcomm and Microsoft discussing how established organizations are connecting AI to existing systems and clarifying governance. The second features Firdaus Bhathena, the chief technology and transformation officer at S&P Global, in conversation with Newsweek’s AI Impact Forum host, Dr. Ranjit Tinaikar. They’ll explore how large companies are turning advances in AI into measurable business results, and why trusted, differentiated data matters more than ever.
4. The Medical Device Revolution Nobody Predicted
In one of the more thought-provoking contributions in this week’s newsletter, a reader shared an insight about medical devices that felt almost like a revelation. Medical devices have always presented a paradox. They are foundational to modern medicine, playing a central role in how clinicians diagnose, monitor and treat disease. And yet they have never defined the venture category in quite the same way biotech or software has. The challenge, the reader argued, has never been clinical impact. It’s how the category fits into the standard venture model. Device companies have historically faced what he called an “unholy trinity”: long development timelines, regulatory risk and slow clinical and commercial adoption. Even when a device was truly life-changing, the business opportunity often remained bounded by its initial market. But AI may be changing that risk-reward equation. Instead of being a discrete, purpose-built clinical instrument tied to a single intervention, an AI-enabled device can become a clinical intelligence platform. It can support earlier detection, fuel novel biomarker discovery and help create a more anticipatory model of medicine. The reader suggested thinking about devices across three dimensions. First, capability: once a device is deployed, each additional validated algorithm can expand its clinical utility and addressable market. Second, setting: when guidance and interpretation are built into the tool, technologies that once required a specialist can move into primary care and other points of care. Third, compounding data value: devices used in real-world settings generate datasets that support further validation, new algorithms and additional clinical use cases. The example he shared was Optain, a company that makes a low-cost retinal camera. The camera can bring diabetic retinopathy screening directly into primary care, and additional validated algorithms could eventually help assess broader indicators of vascular, metabolic and neurological health. In a resource-constrained healthcare system, the scarce resource isn’t just capital. It’s clinical attention, workflow capacity and organizational bandwidth for change. Once a device is integrated into clinical workflow, AI can make that deployment more valuable over time — supporting more algorithms, enabling broader insights and delivering greater value to both patients and clinicians. It’s a vision of healthcare innovation that isn’t just about better gadgets, but about building intelligent systems that meet patients where they already are.
5. When Broken Equipment Meets a Helpful Colleague
Not all AI impact lives in the C-suite. Sometimes it shows up on the factory floor, in the middle of a breakdown, with a technician desperately trying to figure out how to fix a piece of specialized equipment that came with almost no documentation. That was the reality for Proterra, the company that makes battery systems for commercial vehicles and equipment, at its manufacturing plant in Greer, South Carolina. When the original installers left, they took their knowledge with them. The documentation was minimal, and the automated machinery had no history file to consult. So Proterra did something smart: it started using MaintainX, a maintenance and operations platform, to manage equipment maintenance and build its own record of previous repairs and troubleshooting steps. Nick Haase, co-founder and senior principal at MaintainX, explained how it works. Technicians started centralizing work orders, procedures and screenshots. They built a record of how problems were solved, what worked and what didn’t. Then, when a time-sensitive breakdown occurred, they could ask the AI to search years of work orders for similar failures, summarize how they were handled and recommend next steps. The same model, trained on years of relevant work orders and manufacturing expertise, starts to behave less like a search bar and more like an experienced colleague — one who has seen the failure before and always remembers what happened. Haase said the manufacturer reduced downtime by making previous fixes easier to find, and the AI could also identify recurring failure patterns that helped technicians refine preventive maintenance schedules. Proterra reported $250,000 in cost savings and more than 192 hours of avoided unplanned downtime from its overall use of MaintainX. Those figures cover the broader maintenance operation, not the AI feature alone, but the point stands. Over four years, those records grew into a living reference library for equipment that arrived with little documentation. And each repair adds another example technicians can retrieve the next time a machine develops a similar problem. That’s a story about AI in its most practical, no-nonsense form: not replacing people, but making their expertise more useful and more durable. That same spirit is showing up across the broader economy. Travel companies are using AI to move trip discovery toward conversational recommendations while also applying it to booking and airport operations. HeadFirst Global is consolidating 11 businesses under a new brand called Vertage, building a platform to help enterprises define work by outcomes and decide how to blend people with AI agents. A Cisco and Omdia survey of 1,000 IT leaders found that 51 percent already run agentic AI in production, and 24 percent are comfortable with AI operating without human oversight. McDonald’s is scaling its generative AI-enabled ArchIQ ordering system as part of an $8.5 billion, 10-year investment in technology. SAP, meanwhile, has updated its global AI ethics policy, adding mandatory ethics assessments and human oversight to its governance process.
6. The People, Promotions and Unexpected Magic of AI
Beneath all of this strategizing sits a simple truth: in the AI economy, the competitive advantage is quickly moving to the people and companies that can actually turn ideas into working tools. The executive moves in this week’s newsletter make that clear. Jared Coyle, after serving as chief AI officer for the Americas at SAP, joined Anaplan as executive vice president of AI, where he’ll lead its AI Center of Excellence from ideation through enterprise delivery. Chris Hart left Eli Lilly to join ProQR as chief data and AI officer, helping shape AI strategy across drug discovery and development. Florian Quarré joined Premier as chief AI and analytics officer from HealthScape Advisors. Todd Lohr, who spent years as KPMG’s national managing principal, is now vice chair and head of Client Technology & Innovation, where he’ll focus on building AI-native businesses. And Wei Zheng, who led development of Conductor AI, was promoted to CEO as the company pivots toward answer engine optimization. The pattern is unmistakable: the people who were once in charge of “exploring” AI are now in charge of actually running the business with it. But perhaps the most human note in the entire newsletter came at the very end, in a reflection on how AI has become personal rather than professional. One leader described how his “CEO stack” used to be a collection of off-the-shelf tools like Claude, ChatGPT and NotebookLM. Now, he’s built his own custom apps in Lovable and Claude Code. There’s “Later Gator,” named after his daughter’s favorite book, which manages his to-do list and forces him to make a decision on anything untouched after 48 hours. There’s “Seldon,” his virtual second brain, which briefs him on his day, drafts his emails and monitors his most important documents. And there’s “Murrow,” which pulls news from sources he’d never find on his own. None of these required an engineering background. He built all of them in just a couple of hours. That’s the real magic of AI right now — not some far-off future where intelligent machines take over everything, but the present moment where an ordinary person can teach a machine to help them think, remember and decide. It’s the same instinct that makes a factory technician trust an AI that remembers every previous repair. It’s the same spirit that makes a retailer brave enough to ask customers for fresh permission rather than quietly bending the rules. The aspiration isn’t high because we’re all ready yet. It’s high because, for the first time, the distance between the idea and the proof has never been shorter. The readiness may still be low, but it’s no longer a question of whether the technology can do something. It’s whether the rest of us can learn to build with it.
