Episode 63: Harmar's Stephen Hightower on the data problem hiding behind AI in manufacturing

Download, Listen, and Subscribe

Apple  |  Spotify  | YouTube

Or search "Manufacturing the Future" wherever you listen to podcasts!

"They have a data problem that's wearing an AI costume. Because if you don't have accurate data, it doesn't matter what you do, you just do it faster." – Stephen Hightower

A sharp wake-up call from Stephen Hightower, Chief Technology Officer at Harmar, who put a moratorium on development until his team fixed the data underneath. The problem he's solving isn't just data quality, it's the operational waste that builds up when three integrated enterprise systems don't agree with each other, manual workarounds become tribal knowledge, and spreadsheets quietly replace the systems people no longer trust. When your data is broken, AI doesn't fix it. It just does the wrong thing faster.

In This Episode:

Stephen walks through how he's rebuilding data confidence at Harmar, a manufacturer of mobility and accessibility solutions. He describes launching a Master Data Initiative using CRUD and RACI matrices to assign clear ownership for every critical data element across engineering, ERP, and customer systems. He explains why cycle time, quality, and unit cost are the only three metrics that matter in manufacturing, and how tracking integration error rates across systems exposes where data breaks down. Stephen also shares how his Lean Six Sigma background, starting in the late 90s at Lockheed Martin, shaped his approach to treating broken tech stacks as waste to be eliminated through root cause analysis and corrective action. He describes putting monitoring tools in place to gain insight the ERP system wasn't providing, deploying a digital worker to handle customer care calls so his team can move up the value chain, and using the financial close cycle time as a diagnostic for how well a business is really running. He's also candid about AI's limitations: he uses Claude for data analysis and engineering cycle time problems, but stresses that every AI output requires human verification because it will, in his words, "lie to you all day long."

Topics

• Why most AI failures start with a data problem wearing an AI costume
• Running a master data initiative using CRUD and RACI matrices for clear ownership
• Tracking integration error rates across enterprise systems to expose waste
• Applying Lean Six Sigma to broken technology stacks
• Finding the hidden spreadsheets that signal system waste and broken trust
• Using monitoring tools to gain real operational insight the ERP wasn't providing
• Deploying a digital worker for customer care operations
• Why financial close cycle time reveals operational health
• The "go find the spreadsheets" test for any manufacturing business