Axion Founder Daniel First on Solving the $4 Trillion Manufacturing Problem with AI

Axion bets manufacturing's hidden quality failures are worth trillions

American manufacturing runs on $30tn of revenue, and $2-4tn of that goes to quality failures companies cannot see until customers complain. Axion's bet is that the field data already exists across sensors, support tickets and technician visits, and that linking it to root causes turns quality from a cost center into a product roadmap. First argues enterprise AI stalls because pilots stop at insight, so Axion automates the path from data to root cause to action.

Takeaways

• Quality failures cost manufacturers trillions. Quality eats $2-4tn of manufacturing's $30tn in annual revenue, and much of it stays invisible until recalls or complaints arrive. Manufacturers market themselves as high quality but rarely have field data on what is actually breaking for customers.
• Axion links scattered field data to root cause. The platform is a command center used across aerospace, medical devices and consumer electronics. It combines IoT and telematics data, customer feedback and technician investigations so manufacturers see the top issues hitting customers and why they happen.
• Enterprise AI pilots mostly fail to reach production. Daniel cites a study this year finding 95% of AI pilots fail and watched C3 and BCG Gamma engagements stall. He says Axion was built to close the gaps that kill pilots, since flashy demos rarely survive production.
• Manufacturers will compete on learning speed. Rather than cutting costs, Daniel expects American manufacturers to compete on how fast they learn from customers. AI agents will propose product iterations, and today's practice of solving problems only on the top 10% of SKUs will give way to constant iteration.
• The data center bet rests on AI delivering. The build-out is spiking HVAC, energy and electrical demand on the expectation it shows up in GDP. Daniel warns that if enterprise pilots keep failing, leaders who already fear losing trust in innovation may never see the payoff.
• First runs six-month experiments, keeps fast people. He sets experiments on a six-month horizon, kills what does not work and measures how quickly customer needs reach the product. He keeps people who move at the company's pace and says founders must reason from first principles because 2010s advice no longer applies.

What customers found in the data

• At Harley-Davidson, 12 departments were on the platform within a couple of months, using quality data to shape both the current product's messaging and the next generation.
• SharkNinja's ice cream machine was used most for protein ice cream it was not designed for, and the company later built a creamier, protein-focused machine.