Ex-Tesla President: AI Agents Replace 12-Month ERP Rollouts

How a Tesla Veteran Is Deploying AI Agents in Enterprise

Jon McNeill served as President of Tesla during its hypergrowth from 4 to 40,000 employees, working directly alongside Elon Musk on everything from hiring philosophy to sales operations. He’s since built and sold five startups and now leads a company applying agentic AI to supply chain optimization — a space where traditional ERP implementations take 9-12 months.

On AI agents replacing consultants: McNeill’s latest venture pulled the team that built Tesla’s ML-powered supply chain platform (circa 2017) and rebuilt it with agentic AI. The result is striking: “Their agents could go in and understand the work rules of that ginormous platform within hours and then design a system and a workflow within hours.” They recently onboarded one of the largest grocery delivery platforms in the country — a project that would normally take nearly a year with standard ERP systems.

On the AI-as-labor question: The hosts pressed McNeill on whether AI labor replacing human labor means even new businesses won’t create human jobs. His honest answer: “I haven’t heard the case for why we do that is compelling enough.” He points to the spreadsheet analogy — 1950s skyscrapers full of human calculators were replaced by digital spreadsheets, but those buildings aren’t empty. Spreadsheet software enabled derivatives pricing, options exchanges, loan syndication — trillions in market cap that nobody predicted.

On the 1-800 number parallel: McNeill’s most vivid metaphor for AI disruption comes from telecom history. When electronic switches replaced 800,000 human telephone operators in the 1970s, the hand-wringing was real. But free long-distance calls enabled toll-free numbers, which created the entire call center industry — “A few years into the 1980s, there were now millions of people employed in call centers. Hundreds of software firms that were created.” His first startup was one of them.

On AI models as a tooling layer: Rather than betting on which AI company wins the model race, McNeill sees the hyperscalers (OpenAI, Google, Anthropic, xAI) as building infrastructure — comparable to web browsers in the internet era. “We’re getting really excited like we did about Netscape and Internet Explorer… but what you really want to be looking at is what businesses are going to get built on top of this tooling.”

On talent selection at Tesla: While not AI-specific, McNeill’s hiring philosophy directly informs how to evaluate AI-era talent. Elon’s method: go super deep on a real problem, determine if the candidate actually did the work or claimed their team’s output. McNeill has since adopted this — present a current problem of yours and evaluate curiosity, analytical depth, and the ability to simplify complexity within minutes.

5 Takeaways from Tesla’s Former President on AI and Enterprise

• AI agents compress implementation timelines by orders of magnitude - What took 9-12 months with traditional ERP now takes days with agentic systems that learn client workflows autonomously
• Every tech revolution creates more GDP and more jobs - McNeill can’t name a single technological breakthrough in history that resulted in fewer total jobs
• Second-order effects are where the value lives - The first-order effect (job destruction) is visible; the entrepreneurial opportunities on the other side are not, but they’re always bigger
• AI models are the tooling layer, not the business - Like browsers enabled Facebook and Airbnb, AI models will enable the real businesses that haven’t been imagined yet
• Humans-plus-AI is the current winning formula - Supply chain experts using AI as an “exoskeleton” — domain expertise to judge what agents bring back, then redirect when needed

What This Means for AI-Powered Organizations

McNeill’s supply chain story is the clearest enterprise proof point yet for agentic AI: domain experts wielding AI agents to deliver results in days that legacy systems take a year to achieve. The pattern — expert + agent > either alone — maps directly to how organizations should think about AI adoption. Don’t replace your experts; give them agent infrastructure that multiplies their output by 10x.