September 8, 2025
Benas Bitvinskas, Co-Founder at Supercamp
How to Become an AI-First Company
April 2025. Shopify CEO Tobi Lütke sends an email to the entire company. AI usage is now mandatory. Teams have to prove why AI can't do something before they can ask for more people or resources.
"Reflexive AI usage is now a baseline expectation at Shopify."
Message received. This is how you do it.
What Shopify got right
Shopify didn't just mandate AI usage and hope for the best. They built their own AI workspace first. Custom assistants with tools connected via MCP running on their own infrastructure. When Tobi mandates AI usage, he's talking about controlled, purpose-built strategy that understands Shopify's business.
This is the blueprint every company should follow.
But here's what most companies do instead: CEO sends the mandate, IT points everyone to ChatGPT, and suddenly your entire company is using external AI tools with zero training or guardrails.
This is why most AI transformations fail.
Step 1: Build infrastructure before mandates
Samsung learned this lesson the expensive way. In 2023, their employees started using ChatGPT for code reviews and debugging. Within weeks they'd accidentally shared confidential source code, internal meeting notes, and customer data with OpenAI's servers.
The lesson? Infrastructure first, mandates second.
Smart companies flip this order. They deploy AI workspace platforms like Supercamp on their own infrastructure before anyone touches external tools. Your data stays internal. Your compliance stays intact. Your costs stay predictable.
Step 2: Create unified AI experiences
Here's what happens when you skip this step: Your marketing team uses ChatGPT, sales uses Claude, finance uses Gemini, and engineering builds custom tools.
Result? Fragmented workflows, inconsistent results, and nobody learning from each other.
AI-first companies build one unified platform instead. Same interface for everyone. Consistent capabilities across teams. Shared learning and best practices.
Your developers get advanced prompt engineering features with custom models. Your HR team gets simple English instructions. Both work in the same system with the same security policies.
Step 3: Design for learning, not just usage
Most companies mandate AI usage and wonder why results are inconsistent. They skip the most critical piece: building systems for getting better at AI.
Smart companies saw this coming and built learning into their AI rollouts:
AI usage questions in performance reviews
Peer feedback on prompt engineering
Dedicated channels for sharing what works (#revenue-ai-use-cases, #ai-centaurs)
Regular reviews that include AI integration discussions
A culture of sharing both wins and failures as teams experiment
They didn't just mandate AI usage. They created systems for mastering it.
Step 4: Solve the skill gap problem
Your developers understand prompt engineering and model limitations. Your HR team just knows "AI makes work faster." This skill gap kills AI transformations.
AI-first companies bridge this gap with progressive disclosure. Complex AI capabilities hidden behind simple interfaces. Advanced users get power features. Beginners get guided experiences.
The result? Everyone can be productive from day one, but experts aren't limited by simplified tools.
Step 5: Build governance that scales
AI-first companies don't just mandate usage - they build systems that scale with adoption. Platforms like Supercamp include usage monitoring and cost controls from day one. Automatic compliance with whatever regulations apply to your industry - GDPR, HIPAA, SOC 2.
The AI literally can't violate regulations or expose confidential data because the guardrails are built into the system, not bolted on afterward.
One unified workspace with consistent security policies instead of dozens of scattered AI subscriptions. When your team grows from 50 to 500 people, your AI governance scales automatically.
Step 6: Measure what matters
Most companies track the wrong metrics. They count AI tool usage and wonder why productivity isn't improving.
AI-first companies measure outcomes instead:
Time saved on specific workflows
Quality improvements in deliverables
Cost per unit of work completed
Employee satisfaction with AI tools
Business impact from AI-enhanced processes
They track leading indicators (adoption, training completion) and lagging indicators (productivity gains, cost savings). This creates a feedback loop that improves AI implementation over time.
Start small, think big
The companies that nail AI transformation don't try to boil the ocean. They pick one team, one workflow, one problem to solve really well.
GitHub started with AI code completion for their own developers. Proved it worked internally before rolling out GitHub Copilot to millions of users.
Shopify built AI tools for their own merchant support team first. Learned what worked, what didn't. Then scaled those insights across the entire platform.
Your marketing team struggling with content creation? Build them the perfect AI writing assistant. Your sales team drowning in prospect research? Give them AI that actually understands your ideal customer profile.
Perfect the experience for 10 people before you roll it out to 1,000. The insights you gain from that small group will save you months of company-wide confusion later.
Once you nail it for one team, the others will start asking when they get access. That's when you know you're building something that actually works.
Why most companies get this backwards
The pressure to "become AI-first" makes executives want immediate results. They skip the foundation work and jump straight to mandates.
But here's what's frustrating: the blueprint already exists. Companies like Shopify, GitHub, and others have shown exactly how to do this right.
Launch your own AI workspace first, then mandate its use. Custom tools, controlled environment, data stays internal. Learning systems built in from day one.
Yet most companies are doing the opposite. Mandate first, figure out the infrastructure later. Point everyone to public AI tools and hope nothing breaks.
The window is closing
Your employees are already using AI tools. They're copying sensitive data into ChatGPT, asking Claude to write emails, and using Gemini for research. It's happening whether you like it or not.
You have two choices: let this happen in the shadows with zero control, or build something better.
The companies winning at AI aren't the ones with the biggest AI budgets. They're the ones who solved the right problems first. They built tools so good that using anything else feels like going backward.
Your smartest competitors are already building this infrastructure. They're either implementing platforms like Supercamp or building their own AI workspaces from scratch.
Start tomorrow. Pick one team. Find their biggest frustration. Build them something that doesn't just use AI - something that makes AI feel like magic.
Because here's what most executives miss: the best AI transformations don't feel like transformations at all. They just feel like better work.
What's the one problem in your company that AI could solve completely?