Engineering Excellence In The Age Of AI

Forbes

5 days ago

Abhi Shimpi is the Vice President of Software Engineering in a Financial Services organization.
As engineering leaders, many of us are racing to integrate GenAI into our development life cycles. The tools are powerful, the potential is massive, but amid all the buzz about velocity and automation, I believe we're overlooking a critical element: Engineering Excellence.
If we don't start reshaping our engineering culture for AI, AI will reshape it for us, and it might not be in our favor.
I don't mean just a technical shift, but also a cultural one. If we lose sight of the foundational practices that make engineering sustainable, secure and scalable, then we're moving forward recklessly.
What Engineering Excellence Used To Mean
Before GenAI entered the scene, engineering excellence had a clear definition. We talked about code quality, test automation, secure development practices, peer reviews, resiliency, architecture rigor and continuous delivery. We had internal maturity models to measure and reinforce those principles. Those models gave teams an understanding of what 'good' looked like and how to build clean, maintainable and trustworthy software at scale.
It was about process and discipline. We created feedback loops, fostered coaching and mentorship and we made space for design thinking and technical judgment.
Now, GenAI is rewriting the rules, and we need to make sure we don't allow it to erase those fundamentals along the way.
Speed Without Discipline
AI has transformed the developer experience. Tools like GitHub Copilot, Google Gemini and Microsoft Copilot can generate code for entire functions or workflows in seconds. Non-technical users can build apps using natural language prompts. In theory, this is empowerment. In practice, it's often chaos.
I've seen firsthand how easy it is to bypass core engineering principles in the rush to adopt GenAI and ship faster. A developer asks Copilot for a script, drops it into a PowerApps and deploys. No design review, no security scan and no consideration given to how security is handled or data is managed. It works, but it doesn't scale. It creates anti-patterns that violate the architectural standards we've spent years putting in place.
And it's not just developers; citizen developers (those with minimal technical training) are building and deploying internal applications without understanding the implications. What kind of data are they handling? What access are they exposing? What guardrails are missing?
And it's happening across industries. The real risk isn't that GenAI makes mistakes, it's that we stop asking questions.
FOMO Is Not A Strategy
Let's be honest: A lot of organizations are embracing GenAI out of fear of missing out. Once the floodgates opened, everyone rushed in. The intent was good, but the pace? Unsustainable.
There's nothing wrong with moving fast if you're moving with intention, but if you don't know what you're measuring, you're just reacting. And when you prioritize output over outcome, you miss the real opportunity.
This is why I keep emphasizing outcome over output. GenAI can help you generate more code. That doesn't mean it's better code. We need to slow down just enough to ask: Does this solution create long-term value? Is it secure? Is it explainable? Is it maintainable?
Rebuilding Development Culture For AI
Embedding AI into our workflows is not enough. We have to embed engineering judgment alongside it.
That means reinvesting in the things that made us strong in the first place: coaching, mentorship, engineering excellence and craftsmanship. Peer reviews still matter, clean architecture still matters, release/maintenance still matters and code design is not optional.
In one example from my experience, developers unfamiliar with a programming language were able to deliver time-sensitive solutions using GenAI tools faster. We layered in strong governance: design reviews, peer oversight, security assessment and architectural alignment. Without those guardrails, the same project could have introduced serious risks.
Hence, AI doesn't eliminate the need for engineering culture. It amplifies the consequences of not having one.
Redefining Maturity For An AI-First World
We used to measure engineering maturity using KPIs like velocity, defect rates, time to market and code coverage. Those still matter, but they're no longer enough.
Now we need to measure how efficiently and responsibly we're using AI. That includes measuring aspects, such as:
• How much human oversight is required?
• Are AI outputs explainable?
• Are they aligned with our architectural patterns?
• Do we trust the AI engine's recommendations? And if not, why?
If we allow AI to review our code, we must also define a trust framework. What is the trust score? What patterns is the AI referencing? Do those patterns match what we've codified as best practice? Which LLM should be used?
The maturity model must evolve and be assessed continuously. Otherwise, we're shooting in the dark.
Psychological Safety And Performance In A Machine-Driven World
There's another piece to this puzzle—psychological safety. When we're using AI, safety is about trust in systems.
We need to build environments where developers feel safe questioning AI outputs, rejecting them when necessary and adding human judgment. Blind faith in GenAI is just as dangerous as blind rejection.
At the same time, we need to hold teams accountable for performance and outcomes. The tools may change, but excellence still requires clarity, consistency and commitment.
What Good Looks Like
So, what does success look like?
From our experience, it includes:
• Less rework
• Fewer defects
• Lower tech debt
• Faster and more efficient onboarding, even for junior engineers
• Enhanced developer productivity and satisfaction
In the example I shared earlier, we saw measurable gains using GenAI. Faster delivery, broader developer capacity and successful outcomes even when teams were new to the tech stack. But those benefits only came after we added extra oversight to ensure architectural compliance and secure development practices. Over time, that governance load decreased because the cultural foundation was strong.
That's the path forward. Short-term governance for long-term gain.
Shape Or Be Shaped
The real test of GenAI is cultural. Tools will continue to evolve. But if we fail to adapt our engineering practices and mindsets, those tools will define our future for us.
The future is about moving with purpose.
If we can redefine our maturity models, enforce meaningful guardrails and keep engineering excellence at the center, AI will be a powerful ally. If we don't, it will become a force we no longer control.
And by then, it might be too late.
Forbes Business Council is the foremost growth and networking organization for business owners and leaders. Do I qualify?