$1.66. That's what it cost to run an AI agent through 500GB of retail data, across 100 rounds of analysis, in under three minutes.
That number shifts the conversation about AI and data work more than any benchmark ever could.
The experiment involved feeding half a terabyte of retail data to an AI agent with zero retail domain knowledge and asking it to develop a business strategy. The agent completed 100 analysis passes in 134.9 seconds and produced strategic recommendations the author described as unexpected - directions that wouldn't have surfaced through conventional analysis.
What the Numbers Actually Mean
A mid-level data analyst in the US costs roughly $40-80 per hour. A single day of analysis runs $320-640. A full strategy engagement involving multiple analysts and domain experts can easily reach $5,000-50,000. The agent did it for less than a bus fare.
The 500GB scale matters just as much as the price. Most small and mid-sized businesses have retail data scattered across spreadsheets, databases, and point-of-sale systems that nobody ever fully analyzes - because the cost of hiring someone to dig through it rarely makes sense for the expected return. At $1.66 per deep analysis run, that math changes entirely. You can run it every week. You can run ten different prompts on the same dataset and see what falls out.
The 134.9 seconds is almost secondary, but worth noting: a human analyst running 100 separate hypothesis tests across 500GB of data would take weeks, if they'd even attempt it.
The Part That Deserves Skepticism
Here's where I push back on the "end of data analysts" framing: the hard part of data analysis was never the computation. It was knowing which questions to ask, and knowing when an answer is wrong.
The author had zero domain knowledge and produced an unexpected strategy. Unexpected in a useful way, or unexpected in a "the model generated a confident-sounding retail recommendation from misaligned data" way? Without domain expertise, those two outcomes look identical at first glance. AI agents produce fluent, structured outputs whether the underlying patterns are real or noise.
Data analysts earn their value not just from running queries, but from knowing when the data is dirty, when a statistically significant result is operationally irrelevant (meaning it shows up in the numbers but wouldn't change any real business decision), and when to distrust a clean-looking conclusion. A $1.66 run doesn't fix those problems - it just surfaces more hypotheses faster.
The Role Shifts, It Doesn't Disappear
For small business owners who've never had the budget for dedicated data analysis, this is a genuine opening. Retail data that was previously too expensive to interrogate systematically is now accessible to anyone willing to set up a basic agent workflow.
For working analysts, the job moves upstream. Less time spent querying, cleaning, and formatting outputs. More time deciding which AI-generated hypotheses are worth testing and explaining to stakeholders why a $1.66 recommendation should or shouldn't be trusted.
The analysts who treat this as a threat are going to be right. The ones who treat it as a faster shovel are going to be busy.