Be AI First launched this week. But the decision to build it started years ago—back when saying AI will become the operating system of business got you polite nods and immediate topic changes. Here's the real story.
The Bet I Made in 2018
In 2018 I bought beaifirst.com with a conviction that felt obvious to me and dramatic to everyone else: Companies wouldn't just use AI. They'd eventually need to reorganize around it. Not as a side tool. Not as a pilot. Not as a "Center of Excellence." As an operating model.
Back then, it sounded premature. Today, it's just risk management. Learn more about Be AI First →
The Long Runway (2014 → 2023)
I started building machine learning models around 2014, during a quantitative PhD in psychology. Data science felt like discovering leverage: you could turn messy reality into signal, and signal into action.
My career path surprised people (and honestly, it surprised me too): I transitioned from being a rabbi into data science, and in 2016 took my first data scientist role at Travelport. Over the next several years, I worked in AI and analytics at large companies—including BP and Priceline—and kept repeating the same message internally: AI isn't a feature. It's infrastructure.
The models were narrower then. The tooling was clunkier. But the trajectory was clear.
2018: Agents Talking to Agents (and the Blank Stares)
That same year, I wrote an article predicting something that sounded borderline insane at the time:
- People wouldn't spend their lives "browsing the internet."
- They'd have agents that did the legwork.
- And business would increasingly become agents talking to agents.
Most people didn't buy it. I'm very familiar with the blank stare—the one that says: "Nothing looks different around me… what are you talking about?"
I wrote about this prediction at the time — see How AI Will Drive the Future of Travel (2019).
So I kept building. I started Invown (a regulated crowdfunding platform), continued investing in real estate, and kept using AI in the background—helpful, but not yet decisive. Until late 2025.
The Threshold Crossing (Late 2025 → February 2026)
Something fundamental shifted. Not in theory—in capability. Since late 2025, models crossed from assistants to operators: shipping code, operating inside spreadsheets, running multi-step workflows end-to-end. One person with a well-designed agent stack can now operate like a much larger team.
This is why "AI-first" matters. The advantage isn't perfection. It's compounding iteration speed with guardrails.
I turned down real vacations around this time—not because I'm allergic to fun, but because it felt like watching a new operating system get installed on the economy. I didn't want to look away mid-install.
The Weekend That Broke the Old Timelines
Here's where it stopped being philosophical. I asked my team to build a new module for Invown. We're fintech. Security and compliance aren't "nice-to-haves." They're table stakes.
They told me: 8 weeks. I pushed back. They revised: 5 weeks. Still no.
So I took the requirements, pushed them through modern AI tooling, and built a working solution in a weekend—including security credentials, a UI harness to test it, and a production deployment.
And here's the point: I'm not a software developer. I'm a data scientist. I can reason about systems. But I'm not a "ship production modules solo" person. Except, suddenly, I was.
That was the moment I fully accepted: the timeline assumptions most teams are using are now obsolete.
PurposeLife: My PhD Research Finally Scales
Around the same time, I rebuilt something even closer to home: PurposeLife—a purpose-development program based on my PhD research, validated across multiple samples.
The reason it never scaled wasn't because it didn't work. It didn't scale because it required coaching bandwidth: schools don't have enough staff for high-touch coaching, and adults don't have time for heavy 1-on-1 intervention.
But now, AI can coach the process—consistently, patiently, and at scale. That's not a "feature." That's a bottleneck removal. See our proof →
This is why I don't lead with "chatbots." I lead with operating leverage.
The Demand Signal I Couldn't Ignore
As I started running my businesses this way, my output stopped being "better." It became… weird. Not 2×. Not 10×. More like 50–60×—getting done in a week what used to take a team months.
Clients noticed. They saw the speed, the quality, the follow-through. And they kept asking the same question: "How are you doing this—and how do we do it?"
That's when Be AI First stopped being a domain I owned and became a company I had to build.
Why Be AI First Exists
I built Be AI First because I got tired of "AI transformation" advice from people who have never shipped an AI workflow into production. This isn't theory for me. I've led AI teams inside large enterprises and I run multiple real businesses AI-first today. Learn about our team →
The playbook I teach is the one I use every day.
Even the website is a micro-proof: I described what I wanted to an AI agent, went to sleep, and woke up with a working build I could test. That's the world now.
What "AI-First" Means (and What It Doesn't)
AI-first does not mean:
- Buying a chatbot,
- Writing a strategy deck,
- Running a pilot that never ships.
AI-first means:
- Identifying the highest-ROI workflows,
- Designing guardrails and evaluation up front,
- Shipping pilots with real users and real KPIs,
- Then productionizing with governance, monitoring, and training.
In other words: an operating model. See our playbook →
Ready to Become AI-First?
If you want to see what this looks like inside your business, start here:
- Book an AI-First Fit Call — we'll map the highest-ROI workflow and what it would take to ship a real pilot fast.
- Or start with the AI-First Audit — ranked strike list, ROI model, and a 90-day execution plan.
Because the blunt truth is this: If you're still "software-first," you'll be out-iterated by companies that are AI-first. This isn't coming. It's here.
Related: How AI Will Drive the Future of Travel (2019 article predicting agents in travel)