The Question Before Hiring: Can AI Do This Instead
In April, Duolingo CEO Luis von Ahn sent an email to his 1,500 employees that captured a shift happening across business: the company would go "AI-first." Before any team could justify hiring, they'd need to prove AI couldn't handle the work.
Around the same time, Shopify's Tobias Lütke made a similar commitment. Using AI effectively "is now a fundamental expectation of everyone at Shopify," he wrote. Teams would need to demonstrate why AI couldn't solve their problem before asking for headcount.
These weren't isolated moments. They signal a philosophy taking root: the time to rebuild around AI isn't in five years. It's now.
The Inversion of Hiring Logic
For decades, business growth followed a simple script: identify work, hire people, scale output. AI is inverting this logic.
Kevin Terrell, whose AI-native startup BirchAI was acquired by Sagility, describes the shift: "Even with Fortune 500 healthcare clients, the workload per engineer is minimal." His company achieved what would have required dozens of employees with a fraction of the traditional headcount.
The new AI companies are reaching roughly $3M in annual recurring revenue within their first year while quadrupling year-over-year growth with about $164K in revenue per employee—metrics that would have seemed fantastical in the SaaS era.
Starting From Scratch
What distinguishes this moment is the willingness to rebuild rather than retrofit. Von Ahn was explicit: "Making minor tweaks to systems designed for humans won't get us there. We need to start from scratch."
But here's the crucial part: "We can't wait until the technology is 100% perfect. We'd rather move with urgency and take occasional small hits on quality than move slowly and miss the moment."
This acceptance of imperfection marks a departure from previous enterprise technology adoptions. The AI-first cohort trades polish for speed, betting that competitive advantage accrues to those who learn fastest.
Ron Gabrisko, CRO at Databricks, embodies this: "I can predict my revenue within 1 or 2%. I can see which customers might churn. We have the next best action on every single customer." This isn't aspirational. It's an operational reality.
What Gets Delegated, What Stays Human
The practical question: which parts of the business can genuinely be run by AI today?
The answer is expanding faster than most anticipated. AI now handles recruitment screening, legal document review, financial analysis, code generation, and content creation at scale. In healthcare, logistics, and financial services, AI is automating insurance claims, legal briefings, and revenue cycle management.
But the boundary between AI capability and human necessity remains contested. Professional services firms built on the apprenticeship model face particular upheaval. With entry-level roles shrinking, firms must shift from hiring for grunt work to hiring for leadership potential.
Why the Urgency?
Perhaps the most striking element is the rejection of gradualism. Traditional enterprise wisdom counsels patience: pilot projects, phased rollouts, test over quarters.
But when Microsoft, Meta, Amazon, and Alphabet are committing $360 billion in AI capital expenditure this year alone, waiting becomes a risk in itself. The five-year planning horizon now looks dangerously complacent.
The Uncomfortable Questions
For leaders contemplating this shift, the questions are less about technology than identity:
What is the business when a dozen people can do what once required a hundred?
How do you build culture when your "team" increasingly consists of AI agents?
What happens to the apprenticeship model that builds institutional knowledge?
These are real costs, even if they don't appear on financial statements. The efficiency gains are measurable. The losses are diffuse and difficult to quantify until they're missed.
The Moment
The AI-first moment feels different because of its compression. Cloud computing took a decade to move from curiosity to standard practice. AI is demanding decisions in quarters, not years.
The companies making bold commitments now—Shopify, Duolingo, Box, Zoom—aren't gambling recklessly. They're reading market signals and concluding that the risk of moving too slowly exceeds the risk of moving too fast.
For the rest of us, their example poses an uncomfortable question: if AI capabilities are doubling in months, not years, what does a five-year wait actually buy beyond certainty that you've waited too long?
The time to rebuild around AI isn't safe in the future. It's uncomfortable in the present. The only question is whether you will rebuild your business model or have someone else do it.
What's your perspective on the AI-first approach? Where do you see the limits of automation in your industry?
