⏰ Today in 5 Minutes or Less

Discover how to stop your team's AI-generated prototypes from becoming engineering's worst technical debt nightmare. Drawing on candid advice from AI pioneer Ajay Thakur, you’ll get engineering rules to keep your team's code scalable and questions you can ask to align with your engineering counterpart before shipping.

✦ WHAT'S HAPPENING ✦

What We Can Learn from Engineering to Become Great AI Code Designers

We are living through the biggest shift in the design-to-engineering workflow in twenty years.

For decades, the division of labor was clear: designers crafted simulations of user interactions using tools like Figma, and engineers translated them into stateful, functioning software.

Today, AI coding assistants and generative UI tools have collapsed that boundary. Increasingly, designers are expected to generate interactive prototypes, functional frontend components, and even pull requests.

It is an extraordinary superpower. Yet, if we wield it without understanding how software is architected, we won't accelerate our teams, we will drown them in AI-accelerated technical debt and problems the like of which we saw last week on a global scale including global Microsoft 365 and Cloud Outage, U.S. Nationwide Aviation Ground Stop, U.S. Banking & Payment Delays. To become great AI code designers, we have to also be learning the foundational disciplines of engineering.

From Translation to Co-Authorship

While researching my book, Leading by Design: The Insider’s Playbook for Tech Leadership, I spoke at length with Ajay Thakur, one of the sharpest engineering leaders I have ever worked with. Today, Ajay is CIO at Movius; previously, he served as Head of Cloud Computing and led the Center of Excellence for Generative AI at Uber, spearheading their first production AI deployments.

Ajay captured the sheer leverage AI brings to building software:

❝

"Imagine leveraging AI to scale our engineering capabilities, turning a single engineer into a team of hundreds. This would allow us to tackle complex problems faster and create more sophisticated and innovative products."

Ajay Thakur

When designers use that same leverage to generate code, our relationship with engineering shifts from translation to co-authorship. Engineers no longer have to guess how a micro-interaction should feel when you build a working version yourself.

However, Ajay issued a warning about generative AI that every designer writing code must heed:

❝

"The advent of generative AI has introduced a new layer of uncertainty. While this technology offers immense potential, its outputs can be unpredictable, making establishing robust quality control measures essential."

Ajay Thakur

AI makes generating code effortless; maintaining it in production is still expensive and unforgiving. If a designer prompts an AI to build a UI that ignores component architecture or API limits, an engineer has to throw it away and rewrite it. That destroys velocity and trust.

✦ THE LEARNING ✦

How to Get AI-Generated Code Accepted Into Real Products

Ajay summed up what separates average design partners from extraordinary ones in one concept, technical literacy:

❝

"I've found that the most successful partnerships occur when designers have a solid understanding of technical constraints. A good designer knows what is achievable within our engineering limitations and can push back constructively to ensure we strive for maximum utility in our products."

Ajay Thakur

When AI builds a feature, it usually assumes everything goes smoothly. To make sure what you generate actually survives real-world use and gets approved by the team, use these habits:

1. Plan for Messy Real Life, Not Just the Ideal Case

AI tends to design for a perfect world where everything loads fast, text fits neatly, and users already have plenty of history. Real life is messier: connections slow down, accounts start empty, and long names spill onto two lines.

Instead of asking AI to build a basic piece of screen, like a user activity card, ask it to handle the bumps along the way too: a loading placeholder, a friendly message when there is nothing to show yet, neat cut-offs for long text, and a retry option if loading fails.

2. Check Data Limits Before You Build

Every live part of a screen asks the system for information. Asking too often or pulling too much data at once can slow down the app, drain a user's mobile data, or rack up usage costs.

Before asking AI to build a data-heavy view, take 15 minutes with someone on the technical team to check three basics:

  • Availability: Is the data already ready to use, or does it have to be fetched fresh each time?

  • Cost and limits: What happens if lots of people trigger this at the same moment?

  • Weight: How can you load information in smaller chunks or pause brief typing/scrolling bursts so phones stay snappy?

3. Use Existing Building Blocks Instead of Inventing New Ones

Left to its own devices, AI loves inventing custom styles and extra tools from scratch. That creates extra work for everyone maintaining the product later.

Point AI toward what your team already uses, your existing layout rules, visual styles, accessibility basics, and naming habits. Sharing an example of a real, working component alongside your team's style and AI guidelines helps AI build with familiar parts rather than inventing new ones.

✦ THE PLAY BOOK ✦

4 AI Code Readiness Questions for Engineering.

Before your team pushes AI-generated prototypes or pull requests, align with your engineering counterpart using these four questions:

Focus Area

Question to Ask Your Engineering Lead

Why It Matters

1. Guardrails

"Which parts of our frontend codebase are safe for designer-generated code, and where is the architecture too fragile?"

Prevents you from introducing bugs into critical or legacy systems.

2. System Sync

"How can we feed our production component library and coding standards directly into our AI tools?"

Ensures AI generates modular, reusable components instead of throwaway code.

3. Hard Limits

"What API rate limits, data latency, or bandwidth constraints do our AI-generated interfaces need to respect?"

Keeps locally brilliant prototypes from collapsing under real-world traffic.

4. Quality Control

"What is our definition of 'done' for a designer's pull request. What testing or accessibility checks do you need?"

Respects engineering’s quality KPIs and turns code reviews into coaching.

✦ THE TAKEAWAY ✦

🧠 The Short of It Is

Learning from engineering how to push production-ready code feels like a superpower. Until the handoff wall vanishes completely, and suddenly, we are co-authoring the product directly in live code. That breakthrough leaves our entire creative community staring down a set of uncomfortable, uncharted questions: What do we call ourselves now? What happens to our compensation when our work looks indistinguishable from frontend engineering? And when production inevitably breaks, whose neck is on the line?

Ready to Stop Going in Circles?

Book a Discovery call with me today. Let’s map out a bulletproof strategy to reclaim your executive presence, lock down your professional narrative, and fast-track your path back into a leadership role that matches your true strength, passion and expertise.

That’s all for this week!

PS: Would you like to receive more leadership strategies? Here are a few ways I can help:

With ❤ from Sally

Welcome to Leading by Design!

Hello and welcome to Leading by Design! If you’re new here, add your email to ensure that you receive my next piece in your inbox.

REPLY

Avatar

or to participate