No-nonsense lessons from 200+ investments and 2 exits.

I Spent 3 Years Perfecting AI Onboarding. Here's the Hard Truth.

Doing less is the only way to get more out of AI onboarding.

That sounds wrong, doesn't it? But after advising over 200 companies, from giants like Scale AI and OpenAI to tiny startups, I’ve seen millions of dollars lit on fire chasing the opposite idea. I’ve spent three years figuring out what actually works. The answer is simple, but it’s not easy.

At my last company, RemoteTeam, we were acquired by Gusto. Before that, MovieLaLa was acquired by Gfycat. Both times, we were building for a new world of work. And both times, I saw how critical it was to get new hires up to speed. Now, as an investor in companies like Anthropic and Hugging Face, I see the same challenge, but amplified by a factor of a thousand: AI.

Companies are burning cash on AI. They’re buying every tool with a “.ai” domain, running marathon training sessions, and creating beautiful, useless internal wikis full of “AI best practices.” The result? Nothing. Worse than nothing. The tools become shelfware, adoption is a rounding error, and new hires are more confused than when they started.

They’re all making the same mistake. They think the goal is to use AI.

The real goal is to solve a business problem. For a new hire, the biggest problem is feeling useless. They want to feel smart and capable, fast. The secret isn’t a bigger toolbox; it’s a single, well-sharpened knife.

The “More is More” Myth is Killing Your AI ROI

I remember sitting down with the founders of a B2B SaaS startup in my portfolio. Super sharp guys. They had just raised a solid seed round and went all-in on AI. They bought 15 different tools. A writer, a coder, a meeting bot, a design thingy… the works. They were sure this was how they’d scale their remote team and build the next unicorn.

Six months later, I’m looking at their metrics. The adoption rate across all 15 of those expensive tools was under 5%. I talked to a few of their recent hires. They told me about their “AI training” on day three. It was a firehose of new platforms and logins. They smiled, nodded, and as soon as the Zoom call ended, they went right back to Google Docs. It was just too much noise. The tools weren’t connected to the actual job they were hired to do.

This isn’t a one-off story. It’s the norm. Everyone is so scared of being left behind that they’re just throwing money at the problem. That’s not a strategy. It’s a panic attack. It creates tool fatigue and a culture where AI is just another corporate chore.

I saw another company, a fantastic e-commerce brand, spend over $250,000 on a suite of AI tools for their marketing team. They had AI for copywriting, AI for image generation, AI for social media scheduling. The whole nine yards. The idea was to turn every marketer into a one-person content machine. A year later, their actual output hadn't changed. Why? Because nobody knew how to use the tools in a way that fit their existing workflow. The AI was a detour, not a shortcut.

My “Aha!” Moment Across 200+ Investments

The pattern became impossible to ignore. The companies in my portfolio that were actually getting value from AI weren’t the ones with the biggest AI budgets. It was the ones who were focused. I’m talking ruthlessly, obsessively focused.

They understood the first week for a new hire is sacred. You’re a sponge, soaking up culture, product knowledge, and trying to figure out where the bathrooms are. The last thing you need is a dozen new passwords for tools you don’t understand.

The breakthrough was this: The only goal of AI onboarding should be to integrate ONE core, AI-powered workflow that solves a real, painful, recurring problem for that specific role. And it has to happen in the first five days.

That’s it. Not ten workflows. Not five. One.

The “Do Less” Framework: One Workflow to Rule Them All

This isn’t about being lazy; it’s about being brutally effective. Here’s how I tell my founders to do it.

First, Find the Core Pain. (And Don’t You Dare Guess.)

Stop brainstorming in a conference room. Go talk to the people who just went through the pain. Ask your newest hires, “What was the most frustrating, time-sucking, or confusing part of your first month?”

For a new sales rep, it’s probably getting up to speed on customer stories. For a junior engineer at a company like Anthropic, it might be trying to understand a massive, complex codebase. For a marketer, it’s figuring out the brand’s actual voice, not the one in the style guide.

You need to get specific. Don't accept "learning the product" as an answer. Dig deeper. Is it understanding the technical architecture? Is it knowing which customers to reference for which use cases? Is it finding the right assets to send to a prospect? The more granular the pain, the easier it is to solve with a targeted AI workflow.

Second, Build a Single, Integrated AI Habit

Once you know the pain, you build a system to kill it. I don’t mean you give them a ChatGPT Plus license and say “good luck.” You build a machine.

Let’s use the sales rep example. The pain is not knowing the customer history. Instead of telling them to “use AI,” you build this for them:

• The Tool: You use a single platform that has already ingested every sales call, support ticket, and CRM note your company has ever created. Tools like Gong or Chorus, when paired with a large language model, can be incredible for this.
• The Prompt: You give them a dead-simple, pre-made prompt. “I’m a new AE. I have a call with [Company Name] in an hour. Give me a one-page briefing. Include a summary of their business, our full history with them, the top 3 objections they will likely have, and three talking points that connect our product to their specific needs.” This prompt engineering is key. Don't leave it up to the new hire to figure out.
• The Process: On day two, the sales manager sits down with the new rep. They don’t “train” them. They run this exact workflow together for a real call. The new rep watches a powerful, useful brief appear in 30 seconds—a brief that would have taken them a full day to piece together manually.

Suddenly, AI isn’t an abstract chore. It’s the tool that helps them hit their number and make money.

Let's take another example: a junior engineer. Their biggest pain is often just navigating the codebase. It's a tangled web of undocumented features and legacy code. Instead of weeks of painful exploration, you can use an AI code intelligence tool like Sourcegraph. You build a workflow where on day three, their task is to fix a tiny, low-risk bug. Their only tool? The AI assistant. They learn to ask it questions like, "Where is the code for the user authentication flow?" or "What are the dependencies of this function?" They fix a real bug, push real code, and feel like a contributing member of the team on day three, not day thirty.

Third, Make it a Ritual, Not a Reference

This is where most companies fail. Don’t just show them once. Make it a mandatory part of their first week. The sales rep must generate and share one of these AI briefs for every single call they shadow. The engineer must use the AI codebase tool to document one small feature.

By making it a required output, you force the habit. But because it’s so damn useful, it doesn’t feel like force. It feels like a cheat code. They aren’t “learning AI.” They’re learning how to be great at their job, and AI is the vehicle.

How to Measure Success

So how do you know if it's working? Forget about "AI adoption" as a metric. It's a vanity metric. Who cares if 90% of your team logged into the tool once? The only metrics that matter are the ones tied to the original business problem.

• For the sales rep: Is their time-to-first-deal shrinking? Are they ramping to full quota faster than your historical average? Are they more confident on calls?
• For the engineer: Is their first pull request submitted sooner? Are they taking on more complex tasks earlier in their tenure? Are they asking fewer basic questions of senior engineers?

Measure the outcome, not the input. If the business metric is improving, the AI onboarding is working. If it's not, go back to step one and make sure you identified the right pain point.

This is How You Actually Build an AI-Native Culture

When you do this, something amazing happens.

That new hire becomes a true believer. They start getting curious. “Wow, if it can do this, what else can it do?” They start playing. They start showing other people. They become your evangelists, organically.

This is how you build an AI-native company. Not from the top down with mandates and a stack of invoices. You build it from the bottom up, one success story at a time. You create pull, not push.

Stop overwhelming your new people. Stop bragging about your AI stack. Find the single biggest problem you can solve for them in week one and build a dead-simple, integrated AI habit to fix it. Do less, but do it a thousand times better.

It took me three years and watching hundreds of companies to figure this out. You can start on Monday.

Frequently Asked Questions

What experience informs this perspective?

This perspective comes from over a decade of building companies in Silicon Valley, two successful exits (RemoteTeam to Gusto, MovieLaLa to Gfycat), and investing in 200+ startups including Anthropic, OpenAI, and Scale AI. I write about what I've lived.

How can I apply this thinking to my own situation?

Start by identifying the core principle behind the opinion, not the specific example. Then ask yourself: does this principle apply to my context? If yes, test it in a small, low-risk way before going all in.

How has this view evolved over time?

My thinking on most topics has changed significantly over the years. Early in my career, I held many conventional views that experience proved wrong. I try to update my beliefs when the evidence changes.

What's the most common pushback you get on this?

People often push back by citing exceptions or edge cases. And they're usually right that exceptions exist. But building a strategy around exceptions rather than patterns is a losing game for most founders.