Many people have tried AI tools and come away unimpressed. I get it – many demos promise magic, but in practice the results can be difficult.
Therefore, I want to write this not as a futurist prediction, but from lived experience. Over the past six months, I’ve made my engineering organization AI-first. I’ve shared before about the system behind this transformation—how we built workflows, metrics, and safeguards. Today I want to get away from the mechanics and talk about what I have learned from this experience – about where our profession is headed when software development itself is turned upside down.
Before I do, a few numbers to show the magnitude of the change. Subjectively, it feels like we are moving at double speed. Objectively, this is how bandwidth is developed. Our total engineering team has grown from 36 to 30 at the beginning of the year. So you get ~170% productivity with ~80% manpower, which corresponds to a subjective ~2x.
Zooming in, I picked out a few of our senior engineers who started the year in a more traditional software engineering process and finished it first in AI. (Discounts apply to holidays and off-site):
Note that our PRs are tied to JIRA tickets, and the average volume of those tickets hasn’t changed much over the year, so it’s as good a proxy as the data can give us.
In terms of quality, when I look at the value of the business, I actually see an even higher upside. One reason is that when we started last year, our quality assurance (QA) team couldn’t keep up with the pace of our engineers. As a company leader, I was not happy with the quality of some of our first releases. As we progress through the year and build our tool AI workflows to include writing unit and end-to-end tests, our coverage improved, the number of bugs decreased, users became fans, and the business value of the engineering work increased.
From grand design to rapid testing
Before AI, we spent weeks perfecting user flows before writing code. Change made sense when it was expensive. Agile helped, but even then it was too expensive to test multiple product ideas.
This trade-off was lost when we first went to artificial intelligence. value experience collapsed. An idea can go from whiteboard to working prototype in a day: From idea to AI-generated product requirements document (PRD), to AI-generated technology specification, to AI-powered application.
It has shown itself in some amazing transformations. Our website, the hub of our acquired and inbound demand, is now a production-scale system of hundreds of individual components, all designed, developed and maintained by us directly in code. creative director.
Now, instead of validating with slides or static prototypes, we validate with working products. We’re testing ideas live, learning faster and releasing major updates every two months, a pace I couldn’t have imagined three years ago.
For example, Zen CLI was originally written in Kotlin, but then we changed our mind and moved it to TypeScript without losing the speed of the release.
IOur UX designers and project managers code features instead of mocking them up. When the release time crunch hit everyone, they stepped in and ironed out dozens of small details with production-ready PR to help us ship a great product. This includes changing the UI layout overnight.
From coding to validation
The next turn came where I least expected it: Confirmation.
In a traditional organization, most people write code and a smaller group tests it. But when artificial intelligence makes up most of the application, the leverage point moves. The real value is in defining what “good” looks like – in being clear about what’s right.
We It supports more than 70 programming languages and countless integrations. Our QA engineers have become system architects. They build AI agents that create and maintain acceptance tests directly from requirements. And these agents are embedded in coded AI workflows that allow using the system to achieve predictable engineering results.
This is the true meaning of “swipe left”. Verification is not an independent function, but an integral part of the production process. If an agent cannot validate its work, it cannot be trusted to generate production code. For QA professionals, this is a moment of reinvention, where with the right upskilling, their work becomes a critical enabler and accelerator. Adoption of AI.
Product managers, CTOs, and data engineers now share this responsibility as correctness determination has become a cross-functional skill rather than a role limited to QA.
From diamond to double funnel
For decades, software development followed a “diamond” shape: A small product team was handed off to a large engineering team, then narrowed back down through QA.
Today, this geometry is reversed. People engage more deeply in the beginning—defining the intent, exploring the options—and still end up confirming the results. The average performed by AI is faster and narrower.
It’s not just a new workflow; this is a structural inversion.
The model looks less like an assembly line and more like a control tower. Humans set direction and constraints, AI drives execution at speed, and humans step back to validate results before decisions go into production.
Engineering at a higher level of abstraction
Every major leap in software has raised our level of abstraction—from punch cards to high-level programming languages, from hardware to the cloud. AI is the next step. Our engineers now work at a meta-layer: AI organizes workflows, adjusts agent instructions and skills, and identifies safeguards. He builds cars; man decides what and why.
Teams now routinely make decisions about when it’s safe to integrate an AI product without review, how tightly to lock in agent autonomy in production systems, and what signals actually show accuracy at scale—decisions that didn’t exist before.
And that’s the paradox of AI-first engineering – it’s less like coding and more like thinking. Welcome to a new era of human intelligence powered by artificial intelligence.
Andrew Filev is the founder and CEO of Zencoder
