Levels to the game

Hello again my friend,

Geoff Charles is the CPO of Ramp, a $32B+ finance/tech company that released 500+ features last year and crossed $1 billion in revenue. An extremely successful company that’s nailed modern tech products.

Geoff matters because he said something that we’re all kind of aware of, but in a scarier, more serious way. I’ve thought about it a lot.

"If you're not using Claude Code this year, no matter what your role is, you're probably underperforming."

Not engineers. Not product managers. Every role.

He went further. Ramp has a framework, Level 0 through Level 3, for measuring AI skills across every single employee. Here's what each level actually means:

"Our job is to get everyone in the organization up the ladder,"

Now, Ramp is wildly ahead of the game, but remember, the game’s fast.

A year ago, if you used ChatGPT at all you were ahead. Now he’s saying that it’s not enough. Time’s changed quickly. Soon, everyone will think like this across all kinds of companies, not just industry leaders.

And the problem is that people like me who pay attention to this stuff have to write out messages that boil down to “use AI to be more productive, or suffer,” and it kinda sucks. So at the end of this, I have a section dedicated to what we can all do to stay ahead of the game individually (regardless of how the companies views all this).

Let's lock in.

There is a quiet arms race happening inside the most successful companies right now.

It has nothing to do with which AI model got the best benchmark. It has nothing to do with fundraising rounds or flashy product launches. It's about something more fundamental.

How fast can your team actually move?

It’s been a thing for a while. I’ve written about it often describing my day job at beehiiv. But before, we moved faster based on priorities and efforts. It’s different now, because it’s not just about working nights and weekends anymore.

Yes, people are waking up to the fact that internal productivity is a serious competitive advantage because of how much they can unlock with AI. Before this, you needed to hire people who were 1 in 100,000. Now, you give Anthropic $100/month.

The current META (most effective tactic available), another video game term, is not marketing, or advertising, or distribution at all, in the traditional sense.

Here's what that looks like:

Ramp built Glass, a fully pre-configured AI workspace that every employee gets on day one. All their tools connected via single sign-on. A marketplace of 350+ reusable workflows built by colleagues for colleagues. Persistent memory. Scheduled automations. The key insight is what happens at the marketplace level: when one person reaches L3 and figures out a great workflow, it immediately becomes everyone else's L1 starting point. One breakthrough compounds across the whole company instantly.

Stripe built Minions, homegrown coding agents responsible for more than a thousand pull requests merged every week. Humans review the code. Minions write it, start to finish.

Linear built AI agents that are full members of your workspace, assignable to issues, added to projects, mentioned in comment threads. Not a tool you open in a new tab. A teammate you delegate to.

This is how the best-run companies in the world are operating right now.

Here's where it gets uncomfortable, especially if you're career-focused.

If AI is this good, what does a team of two that uses it, look like compared to a team of twelve?

That math has a name: headcount reduction. I’ve talked about it before, and it’s dark, but important.

The flip side, if you're the person who figured it out, is that the same math creates room for fewer people to do more meaningful work at a level they couldn't before (and probably earn way more money).

The companies that haven't figured this out are paying for twelve-person teams to produce what two people could. That gap is going to close. One way or another.

I’ve said it before, and I’m saying it again, please consider using this tech if you’re not already. But, watch for the traps.

When you start using AI properly, you can suddenly manage 50 things at once instead of 10. That sounds like a win. But what actually happens is you end up with 50 half-finished threads instead of 10.

The anxious, overwhelmed feeling doesn't disappear just because AI is doing more of the work. It comes back in a different form, a pile of open loops where the AI pushed paper forward, but nothing actually got done.

I've felt this building Lumiere and Stacko alongside a full-time job. AI makes starting so frictionless that you can spin up five new directions in an afternoon. Before long you have a dozen half-baked drafts, three research docs nobody acted on, and a to-do list that got longer, not shorter. You're busy. Nothing shipped.

The fix isn't using AI less. It's being ruthless about priorities and actually closing the loop. Before you start any task with AI, define what "done" looks like. A sent email. A published post. A decision made and communicated. If the task comes back to you in a slightly different shape with more options to consider, that's just a fancier version of procrastination.

It hit me so hard over the last few weeks.

It’s a cultural thing around being disciplined around the right priorities. Do the work you have to do, don’t start another new app just because Claude can build it for you.

That same discipline is what separates winners and losers. Which brings us to what this actually means going forward.

You don't need to work at a $32B fintech to think this way. But it would help to treat your productivity the way these companies treat theirs, as a competitive advantage worth investing in deliberately.

And I don’t want anyone thinking it’s about cold plunges and infrared therapies. I don’t think that. We’re not trying max personal productivity with a five hour wellness routine here.

Here's the practical version.

Figure out your current level. Be honest. Are you occasionally opening ChatGPT to clean up an email? That's L0. If you review which level you are on now, this can help you reach the next:

Audit your week for repeated patterns. Do you find yourself always doing the same few things every morning? Answering emails? Responding to Slack msgs? Start with these things. Maybe you can get a summary of emails. Maybe you can say “read all of my emails from yesterday, write drafts for me to approve or edit, don’t send them.” Again, there’s an incredible amount of work there done for you within minutes instead of an hour.

Take other people’s ideas. See something on IG or Twitter (X) you like? Screenshot it, send it to the AI, and ask “can you do this?” It’s unreal how often it can. If it can’t, ask “can you tell me exactly how to do it step by step?” and again, you’ll be surprised.

Stop waiting for your company to give you a system (spend a little money). Most companies are still figuring this out. The people who win in the next two years won't be the ones who waited for IT to approve a tool. If your work is against it, don’t tell them (and don’t get caught if there are punishments).

Always close the loop. Every time you use AI on something, ask: is this done, or did I just move it? Push the AI to finish with you, not just think with you.

Back to Geoff Charles (the guy from before), who said it well with one line: "Your job is to automate your job."

Just don’t tell your company that you did.

And I hope this does something to push you to try it, and to push further if you’re already on it.

Thanks as always for reading.