Industrial AI: Five Conditions to Meet Before You See a Return
Velotic's CEO outlines five conditions plants must meet for industrial AI to deliver value. The weakest, data, is also the most common bottleneck. One reported case cites a 6% OEE gain.
On 17 November 2026, IIoT World published an article based on a video interview with Brian Shepherd, CEO of Velotic, and his white paper "The Journey to Value in Industrial AI". The message is clear: AI on the shop floor only pays off if certain foundations are already in place, and the first of them is data.
Plenty of data, little structure
Plants collect huge volumes of machine data, but most of it lacks the structure AI needs. According to Shepherd, building a solid data foundation is the most frequent bottleneck he encounters in conversations with customers. For anyone working on OEE, this is familiar territory: a downtime record without context (shift, order, product, cause) is of little use, even for the most traditional analysis.
The five conditions
The white paper identifies five conditions; the weakest one sets the pace for the entire organization:
- choosing the right tool for the specific operational problem;
- building a solid data foundation, with plant data that is structured and contextualized;
- deploying AI at the right point and the right cost, targeting bottlenecks with a clear financial return;
- building in reliability from the start, so engineers can trace and verify every AI decision;
- choosing between autonomous and augmented operations.
The white paper also proposes the Value Cube, a three-dimensional framework that plant managers can use to self-assess their maturity against the five conditions. The article also frames factory AI as "physical AI" in three stages: sense, think, act. A failure in any one of these stages blocks the entire implementation.
Augmented operations first
According to Shepherd, in the short to medium term the fastest value comes from augmented operations: copilots and real-time recommendations for line personnel, with experienced operators always kept in the decision loop. In the coming years, he argues, most operators will work in this mode.
The cases cited
Two cases are reported. Church & Dwight, a consumer goods manufacturer, achieved a 6% OEE gain. Papa John's cut quality audits across 11 centers from a full week to about three clicks. Both connected machines, sensors and control systems, combining the data with input from shop-floor staff, and followed the same path: start small, learn fast, then scale.
"Most of the time, this is not about a technology challenge… It's about an organizational change management problem." — Brian Shepherd
A note of caution on the source
The source is commercial in nature. Velotic was formed in March 2026 by combining GE Vernova's Proficy business with PTC's ThingWorx and Kepware businesses, and the results cited, including the 6% OEE gain, are those claimed by the company. The article provides no baseline, timeframe or details on the plants involved, so the figure should be read as an indication, not a benchmark.
What a manufacturer can take away
Vendor aside, the checklist is useful, especially for SMEs that want to improve OEE through data. The MES and data contextualization layer comes before AI, and the conditions are largely organizational, not technological.
A practical approach:
- check whether downtime, scrap and production data are collected in a structured way and linked to order, line and shift;
- pick a single bottleneck with a measurable economic impact;
- start with tools that support operators, not with automated decision-making;
- make sure every recommendation is explainable and verifiable by technicians;
- involve shop-floor staff from the outset, because organizational change is the real challenge.