Author: Keith Baldwin

Many organizations want AI. Few are willing to do the foundational work that makes it successful. Many organizations feel pressure to “add AI.” Sometimes that pressure comes from leadership.Sometimes from competitors.Sometimes from board decks, annual reports, or vendor presentations. The problem is not interest in AI.The problem is jumping straight to tools and models before […]

“Prompt engineer” is one of the fastest-spreading titles in AI. It is also one of the most misleading. Prompts matter.Good prompts help. But treating prompt engineering as a standalone job role is how organizations confuse tooling with engineering—and eventually ship fragile systems into production. This article explains why prompt engineering is a skill, not a […]

AI workloads break systems in ways traditional software rarely does. Not because the code is bad.Not because the models are wrong. But because AI introduces latency, unpredictability, and cost spikes that synchronous systems were never designed to handle. Async processing and queues aren’t performance optimizations for AI.They’re survival mechanisms. AI Workloads Behave Differently Than Traditional […]

In enterprise environments, AI rarely lives alone. It lives inside: The AI model is often the least fragile part of the system. What fails are the things surrounding it. Enterprise-grade AI engineering means treating AI as one component in a larger operational system — not the system itself. 1. Architecture That Separates Intelligence From Responsibility […]

AI-powered “vibe coding” has become popular because it feels fast, creative, and liberating. I use it myself — for small, single-user systems, internal tools, and prototypes. It’s a powerful way to explore ideas quickly. But in medium and large organizations, production software lives under a very different standard. When something goes wrong, there are meetings.When […]

“Can we just add AI to this?” It sounds harmless.Optimistic, even. In practice, it’s one of the fastest ways to destabilize a production system. Most AI failures in enterprise environments don’t happen because the models are bad.They happen because AI is treated like a feature instead of what it actually is: A cross-cutting system that […]

For many executives, human-in-the-loop sounds like a concession. A sign that the AI “isn’t ready yet.”A temporary crutch until models improve.A tax on speed and automation. For experienced engineers, it signals something very different: Maturity. In production AI systems, human-in-the-loop is not a workaround for weak technology.It is a deliberate safety mechanism — one that […]

In traditional software, errors are usually obvious. A service throws an exception.A request fails.A user sees a broken screen. In AI systems, the most dangerous errors don’t crash anything. They look like success. That’s why error handling matters more in AI than it ever did in traditional software—and why teams that reuse old assumptions quietly […]

If AI were actually failing at the rate people claim, production systems across finance, healthcare, logistics, and government would already be collapsing. They aren’t. What is failing—quietly, repeatedly, and expensively—is engineering discipline applied to AI systems. This distinction matters, because blaming “AI” is comfortable.Blaming engineering discipline is uncomfortable.And uncomfortable truths are exactly what production systems […]

When AI systems fail, the first question is always the same: What happened? Without logging, that question has no answer. AI systems operating without proper logging aren’t just harder to debug — they are business liabilities. They expose organizations to legal risk, operational blind spots, runaway costs, and irrecoverable trust loss. This isn’t an engineering […]

And How Engineering Prevents It AI prototypes almost always work. That’s the problem. Demos succeed in controlled environments, with curated data, friendly prompts, and no real operational pressure. Production systems, on the other hand, are messy, adversarial, cost-constrained, audited, and unforgiving. When AI prototypes collapse in production, it’s rarely because the model “wasn’t smart enough.”It’s […]

What the 2025 McKinsey AI Report Confirms — and What Enterprises Are Still Missing Disclaimer: This article is an independent analysis and commentary on the publicly available 2025 McKinsey AI Report. McKinsey & Company does not endorse, sponsor, or have any affiliation with AInDotNet or the viewpoints expressed here. The Hard Truth McKinsey Confirmed According […]

AI Isn’t Special. It Isn’t a Strategy. And It Isn’t a Shortcut. Disclaimer: This article is an independent analysis and commentary on enterprise AI adoption trends and publicly available research, including the 2025 McKinsey AI Report. It is not affiliated with or endorsed by McKinsey & Company. All opinions are my own. AI Is Not […]

Most organizations struggle with AI not because the technology is weak, but because expectations are wrong. A common assumption is that AI should do the work—design the system, automate the process, optimize operations, and deliver results. When that doesn’t happen, leaders conclude that AI is unreliable or immature. It’s a lot of work to use […]