The downstream petrochemical market is literally being flooded by data. The unfortunate result is that much of that information is being ignored, according to panelists at Reuter’s 2026 Downstream USA conference this week in Houston.
The abundance of data, much of it now being propagated by AI, has only served to overwhelm users much of the time. Panelists were speaking on the first day of the conference, held July 15-16 at the George R. Brown Convention Center in Houston.
AI has been a pervasive topic during the conference, with few panels not touching on the subject. “Leading with the business problem, avoiding the random acts of digital or AI, and really focusing on those problems that will benefit are key,” says Natalie Thompson, vice president of reliability at BP in Baytown, Texas.
Otherwise, shareholders won’t continue to support AI due to minimal ROI. “At Marathon, we have hundreds and hundreds of dashboards,” says Jacques McGregor, vice president of digital transformation for Marathon in San Antonio, during a panel discussion. “How are our engineers supposed to solve a problem when they don’t know which dashboard to look at? If your data isn’t right, not only can you over maintain, you can also miss failures.
“Using the right AI tools can pull all that data, simplify it and help your engineers make the right decisions, saving all sorts of time and not missing failures.”
At present, AI is currently only exacerbating the problem in many cases. “We build a separate dashboard here, a separate dashboard there, a separate AI agent here and so on,” he adds. “It’s not integrated with the workflow. If you want adoption, you need to understand the workflow and build the solutions in the workflow so people aren’t pivoting from one tool to another to another to get their job done.
“It’s got to be embedded in the workflow.”
The key, panelists say, is in the value of the data. “Poor master data, poor engineering data, poor procurement data etc. is what we’re seeing in many cases,” says Walter Pesenti, vice president of manufacturing excellence at Borouge International in Los Angeles. “That creates inconsistent coding, disconnected storing systems and inconsistent inspection data. The consequences to all that is unnecessary maintenance, unexpected failures, inefficient capital spending, lost production and reduced profitability.”
Improving the data quality is key. “A lot of CEOs come back from a meeting with the board or conferences and the initial reaction is we need to put AI in everywhere,” BP’s McGregor says. “That’s totally the wrong approach. You need to figure out what are the most important processes that are going to move the needle for the company. Then figure out where AI is best suited.”
