Stan Smith

Founder & CEO Gradient AI

Last Reviewed: September 19, 2026

Executive Summary

Stan Smith is the founder and chief executive of Gradient AI, a company that applies machine learning to the two decisions insurers make most often and most expensively: whether to write a risk and at what price, and what a claim is going to cost. Its platform spans property and casualty, workers’ compensation and group health, each with its own general manager, and its central asset is data — models trained across a pooled body of industry claims and underwriting history rather than on any single carrier’s book. The company has an unusual origin. Smith built it as a practice inside Milliman, the actuarial and insurance consultancy, and then bought the business unit from Milliman outright in August 2018, since when it has run as an independent software company. He came to insurance from nearly thirty years in enterprise technology, with founding or executive roles at MatrixOne, Agile Software and OpenRatings.

Career Highlights

• Founded gradient A.I. as a practice within Milliman focused on the hardest analytic problems in risk management and insurance, then acquired the business unit from Milliman on 20 August 2018.
• Has led the independent company since, building it into a multi-line insurance analytics platform with dedicated general managers for the property and casualty and health segments and a data science organisation under a PhD-level VP.
• Led development of several patented technologies, among them systems that predict corporate bankruptcies, a global database for improving supplier performance, and technology supporting performance management in lean initiatives.
• Founded and served as CEO of OpenRatings, a supplier performance ratings business.
• Served as Executive Vice President and General Manager at Agile Software and as a Vice President at MatrixOne, both enterprise product-lifecycle and supply-chain software companies.

Professional Journey

Smith’s career before insurance was spent in enterprise software concerned with supply chains and supplier risk. At MatrixOne and then at Agile Software — two of the significant product lifecycle management vendors of the period — he held vice-presidential and general management roles. He then founded OpenRatings, which built ratings and predictive models for supplier performance, including the bankruptcy-prediction technology among the patents he is credited with developing. The common thread is predicting how a counterparty will behave using data drawn from many organisations rather than from one, which is precisely the problem structure he would later take into insurance.

The move into risk came through Milliman, one of the largest independent actuarial consultancies in the world. Smith established gradient A.I. there as a distinct practice, aimed at applying machine learning to insurance problems that traditional actuarial methods handled slowly or not at all. Milliman was a natural host: it had the domain credibility and the client relationships, and its clients had the data. But a consulting practice and a software company have incompatible economics — one sells expert time per engagement, the other sells a product that improves as more customers contribute data to it — and the practice had outgrown the model.

On 20 August 2018 Smith acquired the gradient A.I. business unit from Milliman. It was a management buyout rather than a venture-funded spin-out, with Smith taking the business personally; the announcement emphasised client support for the transition and a continuing willingness to work with Milliman where opportunities arose. Since then Gradient AI has grown as an independent SaaS company, extending from its original ground into group health and property and casualty alongside workers’ compensation, and building out a full executive team including a chief technology officer, a chief product officer, and segment general managers.

Smith’s public argument for the business is captured in the title of a podcast the company promotes on its own leadership page: data first, AI second. In insurance that ordering is not a platitude. A model trained on one carrier’s claims reproduces that carrier’s historic selection biases; the value of a pooled, cross-carrier dataset is that it sees risks the individual insurer’s own book never presented. That is the asset the Milliman years accumulated and the one the buyout secured.

Education

Dartmouth College

Bachelor’s degree

The field of study and graduation year are not stated in any public source.