Many people think of AI as the train. In reality, AI is the railway. The real shift is the infrastructure being built today: compute, energy, talent, and digital stack. Countries that invest in this infrastructure will shape the future of AI. My conversation with @BCG at MWC.

AI can warn a maintenance team that a machine may fail, but McKinsey argues that the opportunity lies in what the organization does next. A prediction is useful when it changes how maintenance is planned and carried out. That requires more than a model. Microblog by @antgrasso Suppose a warning arrives days before a possible breakdown. The repair still waits because the schedule has not changed or the required part is unavailable. The model has supplied information that the process cannot yet use. Who can act on that warning while there is still time? McKinsey estimates potential cost reductions and additional uptime across asset-heavy industries. Its analysis emphasizes redesigning the maintenance process so that insights can influence decisions and lead to timely interventions. I would put maintenance teams at the center of that redesign, giving them access to information while they are planning work and enough authority to adjust priorities. #IndustrialAI #Maintenance

Finance puts AI agents to a serious test: accuracy, controls and clear audit trails. Hyphenate (@HyphenateAI) raised $45M in Series A to expand its autonomous finance platform for accounting, revenue and treasury workflows. More: bit.ly/4BqqII9 #hyphenatepartner

New in CodeRabbit Secure: Deep Scan now follows a vulnerability across your linked repos, then replays the exploit in a throwaway sandbox to prove it's real before it flags anything. Hendrik walks through this and everything else in the new update:

A familiar name in your inbox can feel reassuring. But is it really who you think it is? Attackers exploit trusted identities and everyday workflows. Making verification a habit is an important way to protect ourselves. @MSFTIssues link bit.ly/4i7ihdb v @antgrasso

AI creates business value when strategy, workflows, and execution work together. Leaders need to connect these layers, redesign how work gets done, and build the skills that turn AI capabilities into measurable outcomes. Source @McKinsey Link mck.co/4hVixKN v @antgrasso

Keeping up with the market means knowing which AI platforms work best for different tasks and training our teams to use them effectively. Source @VisualCap via @antgrasso

If you use AI, do not stop using your brain. An AI suggestion is one possible option, not a decision already made for you An answer may sound convincing and still leave something out. Give it the same attention you would give any advice affecting your work Microblog @antgrasso Before accepting a recommendation, check what it is based on. Does the system have enough information about your situation? Would you accept the same advice from a colleague without asking a question? Explore another option, even when the first suggestion seems reasonable. You can ask AI to propose alternatives, then compare them with what you know. If the original recommendation still makes sense, follow it. There is no need to disagree with AI just to prove you are in control. But keep the possibility of changing your mind. When new information arrives, reconsider the decision. What would make you choose differently today? AI can suggest a direction. You should remain free to take another one. #HumanCenteredAI #AIDecisionMaking

Two patients have the same diagnosis. Would the same treatment be right for both? Clinical decision support helps clinicians examine the differences. Patient information supplies the context. Medical knowledge helps interpret it before a decision is made. Microblog @antgrasso Consider a clinician preparing to prescribe a medication. The patient record contains kidney function results that may affect the appropriate dose. A clinical decision support system can combine that information with medical references and flag a possible dosing concern for review. Now the clinician has something to examine. Are the patient data current? Does the recommendation make sense in this particular care situation? The software can bring relevant evidence to the attention of the clinician while the decision is being considered. The clinician interprets that evidence and decides how to proceed. The purpose is to make medical knowledge available in the context of the patient, without transferring clinical judgment to the software. #ClinicalDecisionSupport #DigitalHealth

Who decides whether a Bitcoin payment is valid? Not a bank. Independent nodes check it against shared rules, while miners add blocks. That allows BTC to move without a central operator, but protecting your private keys remains your responsibility. Microblog by @antgrasso

Cybersecurity is a business decision. Connect AI governance, identity protection, and cryptographic readiness to clear ownership and measurable risk reduction. Make resilience a leadership priority. Source @Gartner_inc link gtnr.it/4jjawSk via @antgrasso

AI consumes energy in 2 phases: training and inference. The first has a finite cost per training run; the second keeps accumulating with use. One training run has a defined end. Inference follows demand, whether the model handles a few requests or millions. Microblog @antgrasso Training requires computation to develop the model. Once a training run is complete, the energy consumed during that run is a fixed total. The trained model can then be used repeatedly without being trained again for every request. But every request still requires computation. The more frequently the model is used, the more inference energy accumulates. The amount also depends on how much processing each request requires. A simple question and a complex analysis do not necessarily consume the same amount of energy. For organizations deploying AI, estimating energy consumption means considering both the training process and the workload generated throughout the life of the service. #AIEnergy #AIInfrastructure

AI adoption does not automatically build trust. With 30% of Americans worried about its pace of development, leaders need to make AI's value tangible: show measurable results, explain its limits, and make human accountability clear. Source @StatistaCharts via @antgrasso

A KPMG survey finds that 89% report measurable value from AI, but only 12% consistently assess that value against its cost. The figures raise a question: what can organizations conclude about AI economics if costs and benefits are not compared consistently? Microblog @antgrasso In its Global AI Pulse Q3 2026, KPMG reports that 61% review AI costs during approval and 59% monitor them once systems are running. Consider an AI application that processes supplier invoices. The finance team may know how much it costs each month. But how much time does it save once the manual corrections are taken into account? Without that comparison, deciding whether to expand the application is harder. Among organizations reporting established ROI, KPMG finds that 48% consistently compare AI value with cost. Financial control tells an organization what it spends. Connecting cost to results helps it decide whether further investment makes sense. #AIEconomics #AIManagement

AI-powered wildlife alerts on rural roads are a useful application of connected technology, especially with measurable speed reductions during testing. @TMobileBusiness Partner