Why 'AI-powered' is doing less work in vendor pitches than it used to

"AI-powered" as a phrase in a vendor pitch used to signal something relatively specific: a real machine learning component doing meaningful work in the product. It's become close to a mandatory checkbox on any software vendor's homepage, applied to everything from a genuinely sophisticated model to a single API call wrapping a basic rules engine, and the phrase alone no longer tells a buyer much.

The evaluation question that actually matters now isn't "is this AI-powered," it's "what specifically does the AI component do, and what happens to the product's core value if you removed it." A tool where the AI is doing real, differentiated work fails this test differently than a tool where the AI is a thin layer over what was already a perfectly good piece of software.

This isn't a claim that AI-powered features are usually fake, most are doing something real. It's a claim that the label stopped being informative once every vendor adopted it regardless of how central the AI actually is, which means the label can no longer substitute for actually asking what the system does and verifying it against real use.

The buyers getting this right aren't skeptical of AI claims generally, they're specific about which claims they're evaluating: not whether a tool is AI-powered, but whether the specific capability it's claiming actually works on their specific data, which is a question no amount of marketing language on a homepage can answer for them.