Why AI agents might not be right for you
SiliconANGLE Jason Bloomberg
Opinion — commentary, not a factual news event.
AI agents are everywhere, but this piece says they’re often the wrong tool. Their unpredictability can be the feature and the flaw.
Based on reporting by SiliconANGLE, Jason Bloomberg — 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
AI agents are getting treated like the answer to everything: coding, customer support, IT incident handling, even messy enterprise tasks. Jason Bloomberg’s argument is simpler and sharper than the current hype cycle. A lot of the time, agents are a shiny way to solve a problem that deterministic software already handles just fine.
His case starts with how these systems work. Most agents sit on top of large language models, which do not give the same answer every time for the same input. That nondeterminism is what makes them feel capable of judgment. It also makes them less reliable than the software businesses have relied on for decades, where the same input is supposed to produce the same result.
That matters most when the task has one correct answer. Payroll is the obvious example. So are cases where a bad outcome would be catastrophic, or where the cost of retries, extra model calls and tool use outweighs the benefit. Bloomberg points to incident mitigation in complex IT environments, weather-dependent planning and business forecasting as situations where agents can help, but only when a perfect answer is not the point.
His practical test is blunt: calculate an error budget. If the acceptable failure rate is zero, don’t deploy the agent. And even when some misfires are tolerable, the expected benefit has to beat the cost of failure by a wide margin, because those estimates can be wrong. Bloomberg also warns that the effort to govern agents may end up shrinking them into something far less useful than the hype promised. In his view, the future probably looks less like agents everywhere and more like agents used sparingly.
My take — AI-written commentary, not fact-checked reporting
This is the rare sane take in a market that keeps confusing cleverness with usefulness. Agents are fine when the job is messy and the downside is limited; they’re absurd when people start asking them to impersonate judgment, especially in places where failure hurts. The industry’s favorite habit is still to build a rocket, then bolt on brakes, seatbelts and a warning label and call it governance.
Read more about this at: SiliconANGLE
