Bring Your Own Agent

(Was this newsletter forwarded to you? Sign up here.)

In 1987, if you hired a great salesperson, you weren't paying for their charm. You were paying for their Rolodex. That leather-bound book of contacts was the asset. The person was the delivery mechanism. A salesperson without a Rolodex was just someone who was good at talking.

That model held for decades. The network was the moat. You hired rainmakers because they brought relationships your company couldn't build on its own. The best ones knew it, too. They guarded those contacts like trade secrets, because they were.

Then the internet unbundled the Rolodex. LinkedIn made everyone's network visible. CRMs made relationships a company asset instead of an individual one. Sales intelligence tools made it possible to find, qualify, and reach any prospect without knowing them personally. The Rolodex stopped being the differentiator.

So hiring shifted. You started evaluating people on skills and potential. Can this person run a demand gen engine? Can she build a measurement framework from scratch? Does he understand how to structure a media mix model? The asset moved from what you knew (people) to what you could do (capabilities). That's where most companies still are.

It's about to shift again.

The inference budget

Jensen Huang stood on stage at GTC last week and said something that should reframe how every executive thinks about talent. Nvidia is giving every engineer an annual "inference budget," an AI token allocation worth roughly half their base salary. For a $300K engineer, that's $150K in AI compute on top of salary. He said he'd "go ape" if a $500K engineer wasn't consuming $250K in tokens annually.

Read that again. Nvidia is spending $2 billion a year on tokens for the engineering team. The company's vision is 75,000 human employees working alongside 7.5 million AI agents. 100 agents per person.

That's a compensation philosophy, not a perk. The token budget says: we expect you to show up with an army.

The third shift

Amjad Masad is seeing this from the platform side. The Replit CEO, whose company just raised $400 million at a $9 billion valuation, says the roles are already collapsing at his own company. "We have designers shipping code, engineers shipping design, and salespeople shipping code." His conclusion: "The particular skill is not the bottleneck anymore. It is how ambitious you are, how generative you are, how creative you are, and how good you are at utilizing these tools."

That's the shift in a sentence. The tool fluency is the new Rolodex.

Think about what's actually happening. The best marketers I work with aren't just skilled. They've built systems around themselves. One has a Claude Code setup that pulls live campaign data across six ad platforms, formats weekly reports, and flags anomalies before the Monday standup. Another built custom GPTs for competitive analysis that know her industry's jargon, pricing models, and buying cycles. A third has an agent pipeline that monitors search term reports across 40 accounts and surfaces wasted spend automatically.

These aren't people using AI. These are people who've built their own infrastructure. Their own agents. Their own toolkits. The value isn't just the person anymore. It's the person plus the system they've assembled around their domain expertise. Last week I wrote about how the knowledge barrier has collapsed and "I can't" has become "I won't." The willing ones didn't stop at learning the tools. They built armies out of them.

Dennis Mortensen, who founded the AI scheduling company x.ai, wrote about this concept back in 2016. He called it BYOA: Bring Your Own Agent. Mortensen asked whether you were hiring Rebecca, or Rebecca augmented by nine agents. And whether it even mattered.

Ten years later, that's no longer hypothetical.

What this means for hiring

Tobi Lütke sent a memo to Shopify requiring every employee to reflexively reach for AI as a first pass on everything. His argument: people who master these tools will "sequester all the best careers to themselves," and it's unfair not to tell everyone that's what the game looks like now. On the Acquired podcast, he described the future as "super highly enabled individuals" who operate with founder-like leverage because of what they've built around themselves.

Bret Taylor, who chairs the OpenAI board and co-founded Sierra, put it more directly on Acquired. Are you generating demand for people, or for their agents? His co-founder Clay Bavor revealed they have a bet on the year when more than 50% of conversations with Sierra's agents will be with people's personal agents. Agents talking to agents.

The hiring implications are uncomfortable. When you interview a senior marketer in 2027, you won't just ask what they've done. You'll ask what they've built. What agents do they run? What workflows have they automated? What does their personal stack look like? The resume becomes a portfolio of systems, not a list of accomplishments.

And the candidates who haven't built anything will look like the salespeople who showed up in 2015 without a LinkedIn profile. Technically qualified. Practically behind.

The Rolodex parallel

The Rolodex-to-skills shift took about 15 years. The skills-to-agents shift will take about three.

The pattern is the same. An individual advantage becomes visible, then transferable, then table stakes. The Rolodex was an individual's network made portable. Skills were an individual's capabilities made legible through credentials and track records. Agent stacks are an individual's operational capacity made deployable.

Each shift increases the leverage of a single person. A great salesperson with a Rolodex could sell more. A skilled marketer could execute better campaigns. A marketer with a personal agent stack can do the work of a team.

That last part is why Jensen Huang is budgeting 100 agents per person. It's why Tobi Lütke is rewriting what Shopify expects from every employee. And the people who figured it out first are already pulling away.

The question for CMOs

If you're running a marketing org, using AI isn't the bar. That bar is too low. The real question is whether your next hire shows up with their own agent stack, their own custom tools, their own automated workflows. And whether you even know how to evaluate that.

Because the Rolodex carriers who didn't adapt to the skills era got left behind. The skills-only professionals who don't build agent systems will face the same reckoning, faster.

The asset has moved again. It's no longer just who you know or what you can do. It's what you've built to multiply what you can do.

The best hires in 2027 won't just bring skills. They'll bring an army.