What to Look For in People Now

Adapted from two keynotes I gave in the past week — the 2nd Bay Area Young Entrepreneurs Day, and the Lingnan College (SYSU) Hong Kong Alumni Annual Gala.

Another question I hear all the time in founder groups: in the age of AI, do we still need people?

Yes. But not the same kind of people.

The mechanical, rule-following work — filling out forms, transcribing data, sending template replies, doing first-pass document review — AI does this faster and cheaper than humans now. This layer genuinely doesn’t need people anymore. But the layer above AI is more in demand than ever: defining the goal, dividing the work, managing quality, taking responsibility.

I call this “upgrading to manager.” We used to manage and deploy people; now we have to manage and deploy AI. Two faces of the same job.

1. AI Makes the Organization Transparent

The biggest side-effect of AI landing in a company isn’t replacement — it’s making the gap between people visible to the naked eye.

Start at the micro level. Same role, same hours, look at the KPIs. Worker A uses AI for email triage, customer research, and follow-up scripting — and lands 200 new prospects this week, 30 of them in real follow-up. Worker B is still on handwritten notes and Excel — same hours, 20 prospects, 5 follow-ups. A 10× gap.

We used to look at effort — and lose our nerve. Now we look at output — and it just sits there.

Now the meso level. Same company, different teams. The old investment-research squad: 3 analysts plus 2 associates — 50 deals reviewed per quarter. The new squad: 2 AI-native twentysomethings plus 1 open-minded MD, running AI for sector maps, financial extraction, founder-interview synthesis, and first-draft IC memos — 200 deals per quarter.

Precision, in the end, is about moving HR decisions from “gut feel” to “measurable performance gap.”

Here’s one more leverage stat. Revenue per employee at the top US tech companies: NVIDIA $5.2M, AppLovin $4.2M, Netflix $3.1M, Apple $2.9M — and even an “old guard” name like Microsoft sits at $1.3M per head (trueup.io, 2025). Traditional manufacturing, by comparison, runs ¥300k–¥800k per person — roughly $4k–$11k.

Tens of times apart.

In Silicon Valley, big headcount is the new embarrassment; revenue-per-head is the new pride.

2. Hiring New Blood: Look for Mindset, Not Hands

At Causally, when we hire, we don’t look at how polished the resume is or how many projects are listed. We look at three things: a mindset that starts from the value of the outcome, not the task list; the collaboration skill to treat AI as a teammate; and the manager’s instinct to break problems down, divide work, and round up resources.

None of these three are easy to interview for directly. So we designed a 48-hour take-home test that drops the candidate into a real, cross-disciplinary, single-person project — the workload and knowledge breadth far exceed what any one person could plausibly cover the old-fashioned way. Anyone who can’t use AI simply won’t finish.

Older industry hands hear about the task and the first reaction is always “Are you crazy? How could anyone do that in 48 hours?” But over the past six months, we’ve received plenty of submissions from candidates who did the test extremely well.

What was impossible before is doable now. The people who can’t do it aren’t the people we want.

3. Deploying People: The Leverage Squad

Hiring new blood is necessary but not sufficient. The most effective combination I’ve seen in the AI era is “new blood × open-minded old hand”: a few AI-native twentysomethings — young, no baggage, ready to learn — paired with one open-minded veteran with ten-plus years in the trade, who knows the pain points, knows the customers, and is already using AI. Two of them working the front lines together beats ten new hires every time.

A few weeks back I came across a packaging factory in Guangdong, mostly serving US consumer-goods brands in food and beauty. The old sales process went like this: a client sends a brief, sales and engineering jointly cook up a first quote, and then the revisions begin — switch the material (PCR-PET vs. virgin PET), change the finishing (hot stamping, embossing, UV printing), restructure, adjust MOQ — and with a 12-hour time-zone gap, each round of back-and-forth eats one to two days. A typical deal runs a dozen rounds over two or three months, and getting cut by a competitor mid-cycle is routine.

The boss himself is that “open-minded old hand” — a decade-plus in the business, knows every pain point cold. He fed the past 5 years of quotes, emails, spec sheets, and sample feedback into AI, and had a couple of twentysomething hires build a sales Agent. When a client brief comes in now, the Agent parses the reference images with vision, pulls the three closest past projects from history, and produces a first-draft quote with 3D renderings within half an hour. Every time the client revises, the Agent runs overnight in China — which is the US daytime — preparing the English reply, the updated quote, and the revised structural drawings. The salesperson comes in the next morning and just reviews, signs, and sends.

The result: average brief-to-order cycle dropped from 9 weeks to 2. The same three salespeople now close more than twice as many deals per quarter.

The time-zone gap — which used to be the single biggest drag on Chinese exporters — has flipped into a 24-hour-coverage advantage.

This isn’t hiring. This is leverage.

4. Refreshing the Roster: Upgrade, Re-deploy, Release

For existing employees, I use a three-step ladder.

Step one — Upgrade. Anyone who can grow into an AI manager is gold for the company. Twenty years of experience plus managing AI is the rarest combination on the market right now — worth more than any twentysomething.

Step two — Re-deploy. There are still positions AI can’t replace: management, training, client relations, mentor-apprentice work. People who aren’t ready to upgrade but are still worth keeping go here.

Step three — Release. People who won’t upgrade, sit in a core role, and actively block the transformation — there’s only one option left for them.

Keeping people who refuse to change is the single biggest injustice you can do to the colleagues who are breaking their backs to embrace it.

Being ruthless isn’t being cold-hearted. It’s responsibility. Re-deploy before you release — always.

Closing

A thought experiment I keep coming back to when I’m trying to explain to founders what AI-era management actually looks like:

You’re a hunter-era tribal chief, leading a 10-person crew, running on shared instinct. Everyone eats. One day you’re transported to a feudal court and find yourself in charge of 1,000 retainers — and your first reaction isn’t “what should we hunt today?” but utter paralysis.

That’s the situation you’re about to walk into. Overnight, you don’t have 10 employees — you have 10 employees plus 1,000 Agents.

The instinctive coordination that worked for 10 people will not work for 1,000 Agents.

Upgrade your new hires. And upgrade yourself.