Geoff Charles, CPO of Ramp, told me something that stuck: "My job is to automate my job. And all our jobs is to automate our jobs."
Not "use AI to be more productive." Not "adopt new tools." Automate your job. Make yourself obsolete at what you're doing right now — so you can move to the thing that matters next.
This is the operating philosophy behind one of the fastest-growing companies on the planet. Ramp hit $1 billion in annual revenue with roughly 25 product managers shipping over 500 features in a single year. That's 20 features per PM. Most companies struggle to ship 20 features per team.
And the uncomfortable part? If you're not thinking this way in 2026, you're already behind.
The Consensus Is Wrong
The prevailing narrative right now goes something like this: AI is replacing jobs. PMs are dead. Engineers are dead. Everyone's dead.
The counter-narrative is equally useless: "Don't worry, AI is just a tool, your job is safe."
Both miss what's happening on the ground. The people winning right now aren't protecting their roles. They're racing to automate themselves out of their current tasks — because the prize for doing it first is enormous.
At Ramp, 50% of all production code is AI-generated. That's up from 30% in December. Geoff expects it to hit 80% within months, with 90-100% as a realistic end state. But here's what most people miss: this isn't just an engineering story. PMs are submitting pull requests. Designers are shipping code. Account managers are building internal tools. Salespeople are automating their own workflows.
The entire company, every function, is expected to build.
What Self-Automation Looks Like at Scale
Strip away the buzzwords. Here's what a company running this philosophy looks like from the inside.
Ramp built a Voice of the Customer agent that sits in Slack. A PM asks a question — "What feedback are we getting on our procurement product?" — and the agent crawls 90 days of support tickets, chat logs, Gong recordings, Salesforce notes, and in-app surveys. Eight minutes later, it delivers categorized pain points with links to the underlying source material. The same analysis would take a human eight days to complete manually.
They built a data analysis tool where anyone — not just analysts — can ask natural language questions against their Snowflake database. Sales reps find customers in specific regions. Marketers check campaign performance. Support teams identify common issues. By making it easy to ask questions about data, Ramp didn't just speed up analysis. They increased the number of people who actually ask questions. The company became more data-driven as a byproduct.
And the punchline: they've already moved past that tool. Now they use Claude Code with custom skills connected to their database. Instead of asking a question and getting an answer, they tell AI their goal — "build me a full report on procurement performance with the top 10 blockers in our funnel and draft 10 growth ideas" — and get a complete HTML report backed by live data.
The shift from Q&A to goal-oriented work is the real unlock.
The PM Role Isn't Dying — It's Splitting
This is where the career implications get sharp. Geoff put it bluntly: the way we've trained PMs has been terrible. Stakeholder management. Prioritization frameworks. Communication. All of those are commodity skills now that code is free.
What matters in 2026 is two things. First, can you build? Not write specs — build. Geoff demo'd a feature he built in five minutes using Ramp's internal coding tool: a full accounts payable dashboard with overdue bills, upcoming payments, and total outstanding amounts. Front end and back end. Production-ready code. Pull request submitted to the engineering team.
According to GitHub's Octoverse data, AI now writes 46% of code on the platform. Microsoft's CEO says 30% of their code is AI-generated, projecting 95% by 2030. The cost of producing code is approaching zero. If you're a PM who can build, you're suddenly 10x more valuable than one who writes a perfect spec.
Second, can you think like a GM? With infinite builders available, the bottleneck shifts to strategy: which problems to solve, how to position against competitors, how to monetize, how to win. The craft and the business thinking — those are the two surviving paths.
The System Behind the Output
The deeper lesson from Ramp isn't about individual productivity. It's about systems.
When Geoff reviews a product and spots a bad user experience, he doesn't give feedback. He asks: what broke in the process? What prompt failed? What skill failed? What design system gap caused this? Fixing the person is a band-aid. Fixing the system means the error never happens again.
He told his team ten times that the call-to-action needs to be above the fold. Classic conversion optimization, backed by six years of A/B testing. Instead of saying it an eleventh time, they encoded it into their automated design review process. Now it's caught before anyone needs to review it.
This is the mindset shift: your job isn't to do the work. Your job is to build the system that does the work. And then improve that system. And then automate the improvement of that system.
Ramp tracks AI token usage per employee across every tool — Notion AI, ChatGPT, Claude Code, their internal apps. Not to micromanage. To find who's pushing the frontier and who needs help getting there.
The Uncomfortable Timeline
Ramp has an AI proficiency framework with four levels. L0 is someone who occasionally uses ChatGPT. L3 is a systems builder — someone who's automated core parts of their job and is now building infrastructure for others.
The expectation is simple: L0s won't last at the company. L1s need to become L2s. L2s need to become L3s. There are designated experts whose entire job is evangelizing, onboarding, and unblocking people. New hires must demonstrate AI proficiency in their interviews. For PMs, that means showing a product you've actually built — not a spec, not a prototype, a working product.
Meanwhile, only 6% of companies globally achieve measurable business impact from AI, according to McKinsey. Two-thirds are still experimenting. The gap between Ramp's operating model and the average company is widening by the month.
Geoff plans 90 days out — because in 90 days, you can now accomplish what used to take three years. Three-year roadmaps are a relic.
The Career Advice Nobody Wants to Hear
Geoff said management is "probably dead" as a career optimization. Not because managers aren't needed — but because right now is not the time to build management skills. Right now is the time to be the best builder in the world.
That means going back to IC mode. Fewer meetings. Less alignment theater. More hours in the tools, building, shipping, learning what works. Geoff himself cut back on meetings and one-on-ones to spend time building with AI tools — and he's the CPO of a $32 billion company.
The people who automate their own jobs fastest get to decide what comes next. Everyone else gets to watch.