United States – October 28, 2026 -- Path2Response bans generative AI from production modeling and decisioning, and its co-founder and CTO Michael Pelikan says competitors moving to a single generative AI model for data science put their clients at risk.
Pelikan separates statistical modeling from generative AI
Pelikan, who built neural networks at NASA's neurocomputing institute in the 1990s, distinguishes two categories: statistical models, machine learning and expert systems on one side, and large language models that synthesize content from training-data patterns on the other. He argues the second category does not meet the needs of statistical modeling.
His example is a response model for a specific offer. Such a model must find real signals in real data, he says. A generative system produces output that looks right but has no knowledge that the offer is a seasonal promotion, that a 20% coupon is on the cover, or of the macroeconomic conditions of the current mail season. Pelikan says data scientists with domain expertise supply that context, and he predicts companies going all-in on generative AI for modeling will see results fall, with some already doing so.
Path2Response runs on elastic cloud with same-day and next-day turnaround
Pelikan says the platform was built without the data-cooperative playbook, drawing on his background processing large, complex data sets for airlines. It ingests data in its natural, unstructured form, with no fixed-width fields or translation loss, and processes it at high recency.
Infrastructure scales from hundreds of servers to thousands of nodes on client demand, then back down. Same-day and next-day model turnaround is standard, which he says enables a loop of running, reviewing, adjusting and rerunning a model.
Overspecified pre-selects degrade model performance
When clients and brokers load up pre-selects and push a model toward a predetermined answer, performance drops, Pelikan says. Removing those constraints and applying the right variables lets the model outperform human pre-selection, in his account. He says this requires human expertise at the front end to provide context, followed by the discipline to step back.
Path2Response's formal AI usage policy prohibits generative AI in production modeling or decisioning. Client data is never used to train models and is never shared with external AI systems.