Bringing predictive analytics to the agentic AI era

This one comes labeled clearly: it was produced by Insights, MIT Technology Review’s custom content arm, not its editorial staff. We’re covering it because the thesis is one a lot of enterprise buyers are hearing right now — that predictive models have already won the argument over traditional statistical forecasting, and the new frontier is letting those systems make decisions on their own.

The piece frames the central question as how to let predictive systems act on their own conclusions without drifting away from what the business actually intends. That’s the right question. It’s also the one with the most money and risk riding on it.

We’re broadly for this direction. Real-time training that updates models continuously instead of on a quarterly cycle is a genuine improvement for inventory, demand and risk work. Pulling in messy, unstructured data — the conversations and interactions that never made it into a tidy spreadsheet — can surface signals older forecasting simply missed. Everest Group partner Vishal Gupta, quoted in the piece, says enterprises are done with a backward-looking view. Fair enough.

But I’d push on two things. First, the claim that this argument is already settled is doing a lot of work. Whether a deep-learning forecast beats a well-tuned statistical baseline depends heavily on the data, the domain and how honestly it was tested. Sponsored content doesn’t have to show its benchmarks, and this excerpt doesn’t.

Second, moving from prediction to autonomous decision-making changes who is accountable. A forecast that’s wrong gets argued about in a meeting. An agent that reorders stock, reprices a product or flags a customer on its own acts before anyone argues. Continuous retraining also means the model you approved last month may not be the model running today. If this becomes the default, the harm shows up as quiet drift — decisions that slide away from policy, not one dramatic failure.

Our read: the useful version keeps humans setting the guardrails and auditing outcomes, with clear limits on what the system can do without sign-off. Gupta’s argument that analytics is simply becoming AI may be true as marketing. The enterprises that win will be the ones that still know which decisions they actually delegated.

Predictive analytics is a broad discipline — modeling, data prep, analysis workflows, interpreting results and decision-making applications. The piece argues deep learning and generative AI are folding all of that into one category the industry now just calls AI.