AI-First: Aspiration vs. Implementation

Box is the latest company to announce going AI-first.

But they're not.

Aaron Levie is the third CEO to echo similar memos we have covered from Shopify & Duolingo.

The memos aren't about revolution but evolution.

All three CEOs use remarkably similar language:

→ Box: "We need to ask what work would look like if we started in 2025"

→ Shopify: "Reflexive AI usage is now a baseline expectation"

→ Duolingo: "Making minor tweaks to systems designed for humans won't get us there"

We need to make a clear difference between what is meant by AI-First and AI-Enabled.

AI-First companies:

→ Start with AI as the core engine

→ Design products entirely around AI

→ Organize teams explicitly around AI capabilities

→ Could not exist without AI

Perplexity rebuilt search around AI. Google simply added AI to search.

AI-Enabled companies:

→ Add AI capabilities to existing products and processes

→ Enhance their current business model with AI

→ Use AI to improve efficiency and extend capabilities

→ Could survive without AI, but choose evolution

This is the path Box, Shopify, and Duolingo are following and it's not hard to see why.

For established companies with working business models, this approach offers a safer evolution:

→ They have proven products and loyal users

→ Some have only recently become profitable or stable

→ The innovator's dilemma is real, but so is business continuity

But 'safer' doesn't guarantee safety.

True AI transformation would require radical steps:

→ Rebuild products around AI

→ Restructure teams for AI workflows

→ Develop proprietary, defensible AI

→ Cannibalize existing revenue streams if needed

Most established companies can't become truly AI-first without starting over.

Their architecture, business models, and customer expectations create constraints that startups don't face.

The question isn't whether they become AI-first.

It's if they'll transform deeply enough to avoid being displaced by those who truly are.