As we close out 2025, I keep having the same conversation with CEOs, CTOs, and operations leaders: “We invested heavily in AI this year. We’re not seeing the returns we expected. They’re not alone. And the answer is clearer than most realize.
The Gap That Defined 2025
McKinsey’s latest survey shows 88% of enterprises now use AI regularly. Only 39% see measurable bottom-line impact. That 49-point gap is the story of 2025. I watched it play out at AWS re:Invent last month. Werner Vogels devoted his entire keynote to “Simplexity,” arguing we’ve overcomplicated AI when we should focus on fundamentals. The message landed because it matched what everyone in that room was experiencing: the barrier to AI impact isn’t the models. It’s the data feeding them.
What We Learned Serving Defense, Aerospace, and Manufacturing
We spent 2025 working with operations leaders who kept describing the same problem: “Our AI looks great in the lab. It fails when we scale it across actual operations.” The issue was never their AI strategy. It was that their AI made decisions based on yesterday’s spreadsheet, last week’s inventory count, or assumptions about what was happening on factory floors. Two examples crystallized this for us.
When Deployment Speed Actually Matters
A Tier 1 aerospace manufacturer needed operational visibility across 150,000 square feet of production space. Traditional approaches, running cable, integrating with legacy systems, and configuring software, would have taken weeks. We deployed Thinaer in 1.5 days with two people. The speed wasn’t just convenient. It revealed something important: the barrier to digital transformation often isn’t technology. It’s deployment complexity. When connecting physical operations becomes fast and simple, transformation actually happens. When it’s a six-month IT project, it dies in committee.
When Machine Learning “Fails,” Data Usually Failed First
Another aerospace customer had invested heavily in a machine learning system for predictive maintenance. The results were inaccurate enough that they were ready to write off the entire AI investment. The problem wasn’t the model. It was the data feeding it: batch uploads, manual entries, readings that were hours old by the time they reached the algorithm. Once they had real-time operational data from Thinaer, the same ML models that were “failing” started delivering. Within 90 days, they’d moved from writing off the AI investment to expanding it. They didn’t need better AI. They needed accurate data to give their AI something reliable to work with. This pattern repeated across customers all year: AI initiatives don’t fail because the models are bad. They fail because the operational data foundation isn’t there.
Two Realities That Will Shape 2026
Agentic AI will reward infrastructure, not ambition. The agentic AI market will grow roughly 6x by 2030. These aren’t chatbots, they’re autonomous systems that coordinate workflows, make procurement decisions, and optimize operations without constant human oversight. Here’s the problem: 62% of organizations are experimenting with agents, but most can’t deploy them beyond a single function. Why? Agents make decisions based on accessible data. If that data is incomplete, delayed, or disconnected from physical operations, your agents make expensive mistakes, faster than any human could. The 2026 question isn’t whether to invest in agentic AI. It’s whether your data infrastructure can keep up with autonomous decision-making. Operational visibility isn’t preparation for transformation. It is the transformation. When boards hear “asset tracking” or “RTLS,” they hear logistics optimization. They’re missing it. The companies capturing real value from AI-powered operations, 50% fewer unplanned breakdowns, 20-40% longer asset life, aren’t winning because they have better algorithms. They’re winning because they finally know where things are, how they’re performing, and what’s happening right now. The AI is almost beside the point. It’s table stakes. The differentiator is the operational data that makes AI actually work.
What I’m Telling Our Board
PwC’s latest AI analysis confirms what we saw all year: companies that crowdsourced AI initiatives without an enterprise strategy keep posting impressive adoption metrics while capturing minimal business value. Here’s the framing I’m using with our board, and what I’d encourage you to consider for yours: The 2026 success metric won’t be AI tools deployed. It will be whether your operational data enables those tools to act intelligently. The 39% seeing real AI impact aren’t the ones with bigger budgets or flashier models. They’re the ones who understood what AI actually needs.
A Different Conversation for 2026
We’ve spent three years watching enterprises chase AI capabilities while ignoring the foundation those capabilities require. 2025 exposed the disconnect. 2026 will belong to leaders who act on it, not with more pilots, not with more models, but with the operational visibility that makes AI actually deliver. The future isn’t about choosing between operational excellence and AI transformation. They’re the same thing. Bryan Merckling CEO, Thinaer If you’re working through how to move AI from demos to impact, I’m happy to compare notes. We’ve spent years learning what works—and what doesn’t—in Defense, Aerospace, and Manufacturing operations. Learn more below.