What AI Native means?
Let's start from the beginning
When we talk about a company needing to become AI Native, and after speaking with people from several companies about this, one of the most common responses I heard was:
How much of our code is written with AI?
Maybe even your CEO has asked you that question. In my opinion, that’s not the right question to solve—at least not at the beginning.
While that question focuses on usage metrics, our real concern should be redefining internal processes and company culture.
So I ask:
What exactly do we mean by AI Native?
This was the first question that came to mind when I first heard that a tech company should aim to achieve this goal in the coming years.
☝🏼 And that’s exactly what I want to define in this first edition of the newsletter.
The first response was:
I think we should define a strategy to become an AI Native company, just like we did a few years ago to become Cloud Native.
Do you like today’s edition? Share it with your friends 👇🏻
☝🏼 I believe this kind of transformation is not just about using OpenAI APIs or tools like Cursor/Windsurf to write code.
👉🏼 It’s about changing the way we work at the organizational level and putting humans at the center of the process.
Moreover, I believe this change doesn't only apply to engineering teams, but to every department in a company or startup.
☁️ → 🤖 From Cloud Native to AI Native
A few years ago, we learned how to become Cloud Native by adopting specific architectures and practices. Many of those principles can now be adapted to the AI world:
Infrastructure as Code → Models as Code
Version your models using tools like MLflow or DVC.
Microservices and API-first → Inference services that are API-first
Design REST/GRPC endpoints so any team can call your models.
Auto-scaling and resilience → Auto-scaling AI endpoints
Set up inference clusters that scale on demand.
I know, these are three big steps, and we still don’t know how to fully get there. Step by step. For now, a practical analogy:
Before, you used to save your Excel file on a disk or USB drive.
Then you migrated it to the cloud and your Excel lived in a shared folder.
Now imagine that same file “living” inside an AI service.
🏩 An AI platform for the entire company
This isn’t just about exposing AI APIs to Engineering.
It’s about turning every department into a consumer and prosumer (yes, that’s a word!) of AI. Some examples I foresee:
Sales: shared prompts to write emails, qualify leads, and deliver more effective demos.
Product: using Figma AI + Jira AI + Confluence AI to craft clear requirements and export them directly to tickets.
Legal / Finance / HR: AI assistants that review contracts, analyze budgets, and automate responses to internal FAQs.
☝🏼 We’re just getting started.
We’re still far from achieving this, and we’ll need to make some changes even before starting the transition to AI Native (just like we did before becoming Cloud Native)—like being API-first across the board.
Thinking this way, I believe your company can achieve some important goals on the path to becoming AI Native. Here are a few, with evidence to show it’s not just hype:
🚀 Productivity improvements by avoiding prompt and pipeline reinvention.
Zhang et al. (2023) analyzed data from 1,200+ manufacturing firms in China and showed that every 1% increase in AI adoption led to a 14.2% increase in Total Factor Productivity (TFP). The effect remains after controlling for endogeneity and robustness checks, driven by added value, skilled labor, and tech modernization MDPI.
Li & Wang (2023), using data from listed Chinese firms, confirm that AI-driven innovation boosts TFP mainly through cost reduction, increased use of skilled labor, digitization, and more efficient R&D ScienceDirect.
China’s all-in, nothing new under the sun.
🧪 Boosting experimentation with fast feedback loops.
Koch et al. (2022), using data from Germany’s Community Innovation Survey (CIS 2018), found that companies using AI achieved €16B in “world-first” innovation sales and 6% cost savings in process innovation. Applying multiple AI methods across departments further boosted innovation results ScienceDirect.
Müller & Bley (2022) analyzed 956 press articles and found that 64% of AI implementations in R&D were exploration-oriented (new markets), compared to just 5% for exploitation. Also, AI tends to augment human tasks rather than automate them, favoring rapid test-learn cycles ScienceDirect.
💸 Attracting investors by showcasing a mature internal AI platform.
This is real: the earlier you get into AI, the more capital you can raise.
Goldfarb et al. (2022) analyzed how AI adoption intensity impacts revenue growth. Only high levels of adoption lead to significant revenue increases—especially when paired with complementary investments (cloud, databases, internal R&D strategy). This growth makes companies more attractive to institutional investors ScienceDirect.
Maarouf et al. (2024) developed a large language model (LLM) to predict startup success based on Crunchbase descriptions. It accurately identified which startups would receive funding, showing how AI not only drives business growth but also integrates into the investment decision-making process arXiv.
⚙️ Let’s make it tangible
And because I don’t want to stop at fancy talk, here are some actionable steps you could consider taking this year. Small (?) things that could lay the foundation for something much bigger.
Create a prompt repository
Choose a versioning tool (GitHub/GitLab).
Define a prompt template (context, input/output examples, tags).
Agree on a review workflow: author → technical reviewer → publish.
Protect sensitive data (PII)
Identify critical fields in your database(s).
Implement a wrapper/proxy that services calls instead of directly exposing AWS Bedrock (for example) to interact with LLMs.
Automate security tests to ensure no data is ever leaked.
Adopt Model Context Protocol (MCP)
Start with a low-risk use case (e.g., support ticket analysis).
Measure latency and costs, fine-tune, and gradually scale toward more strategic cases.
AWS is doing it. Might be for a reason?
Have you implemented any of this already?
If you feel like sharing how it went, email me at hello@ainativenewsletter.com or click this link 👇🏻