The golden rule of AI analytics: the human stays in charge

Francois Ajenstat's new tool: a slider that keeps the AI in its place

My old friend, former Tableau chief product officer Francois Ajenstat was telling me all about the new tool he’s launching, Golden Analytics. Then I let slip that the name sounded to me like a Chinese restaurant.

He smiled and built on the joke: “I’ll go one step further,” he said, “I’ll invite you into the kitchen to cook for yourself” -- with a professional cook at your side.

Golden Analytics arrives on the market at a time ripe for successors to legacy Tableau, Power BI, Looker, MicroStrategy, and others -- each of which requires deep expertise.

“Our goal with Golden is to let a competent business user connect a Snowflake table and has a working dashboard in under 2 minutes,” he said. “No data team in the room.” And no black box either.

“I hate the AI chatbots for data,” he said. “They try to be convincing in giving you the answer, but the reality is that you still need experience and you need to understand what it’s doing. I’m trying to bring the best of both worlds.” Like a standby cook.

It’s inspired by Cursor, the code editor with native AI, unlike the bolted-on AI of Tableau or Power BI. You can write code manually, you can ask it to write some code and you can modify what it writes, or it can do everything for you. “Cursor didn’t replace the engineer,” he said. “It made the engineer 10x faster while keeping them completely in control of the code. That’s exactly what we’re building for the analyst.”

Users control the degree of autonomy with a slider called, of course, the Slider of Autonomy. Set it at zero and the AI merely stands by. But set it all the way up and the AI steps in, letting the user stand by to watch.

As appealing as this may sound, it still raises a question. How can something as complex as AI be reduced to a a single axis, of zero AI to 100 percent AI? Data analysis is a collection of distinct judgments, each with its own complexity and its own cost if it goes wrong. If you get problem-framing or causal-inference wrong, for example, your analysis answers the wrong question.

The explanation for the “single-axis” question is subtle, Ajenstat says. “I don’t think there’s a black and white answer.” AI helps by removing some of the drudgery in reaching insight or getting the work done. “It’s not about removing the human,” he said. The slider’s work remains visible. “There are no black boxes.”

It sounds like the only surprise is inside your fortune cookie.

Golden Analytics will launch when it’s ready, no doubt soon. Until then, put your name on the wait list at goldenanalytics.com.