The foundational elements of AI architecture that IT leaders need to scale
MIT Technology Review MIT Technology Review Insights
MIT Tech Review lays out four AI architecture basics IT leaders can bank on: data quality, context engineering, governance, and people.
Based on reporting by MIT Technology Review, MIT Technology Review Insights — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
The pitch here isn't another breathless take on agentic AI. It's a reminder that the boring stuff underneath still decides whether any of it works. MIT Technology Review's Insights team, drawing on comments from Elastic CIO Adnan Adil, argues that four foundational pieces of AI architecture matter more than whatever model or feature just shipped this week.
Start with data. Adil is blunt about it: without clean, connected, real-time-accessible data, models simply don't run well, or they run with the wrong context. Most companies are still dragging around legacy systems, fragmented ownership, and incomplete datasets, and no amount of AI horsepower fixes that on its own. The stakes aren't abstract, either. Gartner's number, cited in the piece, is stark: 60% of AI projects could get abandoned through 2026 if the data underneath isn't AI-ready.
The second pillar, context engineering, is subtler than prompt tweaking. It's about deciding what information a model actually sees for a given query, and just as importantly, what it doesn't. Adil's line is that minimum context, current data, and machine-readable formatting beat dumping everything into the model and hoping for the best. Too much context, the piece notes, dilutes answers, drives up costs, and slows things down. That's a very different problem than
My take — AI-written commentary, not fact-checked reporting
The people point is the one that should get more attention than it does. Nearly 70% of Deloitte's surveyed tech execs say they're growing teams because of generative AI, not shrinking them, which cuts against the layoffs narrative that dominates headlines. If governance and observability have to be built in from day one rather than bolted on later, that's an admission that most companies are already behind, not ahead, on the boring infrastructure work that actually determines whether AI pays off.
Read more about this at: MIT Technology Review