Hema Ramaswamy is a software engineering leader with 30+ years of industry experience. She has launched mission-critical enterprise products for Oracle and Hewlett Packard, as well as multiple startups, and transformed 8 product engineering organizations to scale to their hyper-growth needs. Most recently, she served as a Senior Vice President of Engineering at Tracer Labs.
As a Senior Vice President of Engineering, I've seen firsthand how AI is transforming software development. It seems like every business wants to push out AI-powered products. However, I don't find this explosion to be all that different from the rise of the internet or the mobile phone.
My 30-year career in tech has taught me that there will always be a shiny new object. Whether that object is AI or cloud computing, the impulse tends to be the same: hop on the bandwagon as quickly as you can. Unfortunately, this can be at the expense of strategy.
The rush to adopt new technology can have significant pitfalls. When you don't step back and apply engineering principles, your products are more likely to be misaligned with existing systems, user needs, and business goals. You may pour resources into features that fail to solve meaningful problems — all due to anxiety about falling behind.
To prepare my team for the AI era, I've approached it like any other technological advancement:
Pay attention to trends, but don't let them dictate your strategy. Instead, embrace change with a measured approach that prioritizes business impact.
Here is an overview of how I've helped my developers become "AI ready," both to optimize workflows and enhance our products with AI.
You may notice that the steps below sound familiar. That's because they are rooted in general best practices for software development — the foundation for all great products!
This article is based on a recent conversation with the DevPath team. You can watch the full recording here:
1. Clarify your "why"
Before diving into an AI project, consider the business reasoning behind it. Why do you need an AI solution?
Every business wants to market themselves as "AI-powered," so I understand the rush to add AI wherever possible. However, this approach prioritizes activity over outcome. If you build AI features without considering how they will (or won't) add value for users, you may end up wasting resources.
One strategy I've used is to pick a high-impact project in our current roadmap where we can apply AI. This ensures that AI isn't a distraction from core business goals, but rather a tool to help us solve meaningful problems.
2. Embrace creativity, but start small
While AI solutions need to be sustainable and align with business goals, it's equally important to nurture creativity on your team.
Avoid stifling the flow of new ideas. Even if they're not totally viable, they may help to shift your thinking in helpful ways or inspire actionable projects. I want developers to feel completely free during the ideation phase.
Once you have an idea that you want to test, start small. This is where engineering leaders need to step in and make time for experimentation. Typically, I assemble a very small team and de-prioritize some non-critical work tasks. This creates time for developers to pilot the solution and determine its feasibility and effectiveness.
The pilot team then works together to develop a comprehensive plan, from concept to release. Release doesn't just mean technology! It's important to plan for all supporting functions, including product marketing, customer success, and feedback loops.
Finally, we establish the metrics we'll use to measure impact.
3. Ensure compliance early
AI is evolving so rapidly that it can be difficult to stay on top of legal and ethical considerations.
As an engineering leader, it's my job to equip developers with this information early. Without it, the team may build a prototype that fails to uphold data privacy or other requirements. This can have serious consequences for your budget and morale. Now the company has wasted resources, and developers may feel blindsided if they have to overhaul or scrap their work.
This tip is especially salient for AI, but it holds true for any software development project. Prototypes are most successful when developers view them as part of a whole, rather than one small piece of technology delivered in a vacuum.
It's easy for developers to get lost in the weeds of building something. Engineering leaders can help developers bridge the gap between technology and business implications. As a result, your team's work will add more value to the business.
4. Gather feedback and iterate
As you roll out your pilot project in a controlled environment, gather feedback using the metrics established during planning. Here's what this looked like in practice when we piloted an AI framework in our engineering workflow.
There had been a lot of excitement around Copilot and its potential to speed up development. As a test, I enabled Copilot for developers at every level of the organization, for a limited period of a few months. I planned to gather feedback from developers at different levels (e.g., junior mid-level, staff, etc.) to determine who found Copilot most valuable, and why.
The results:
Junior mid-level developers said that Copilot enhanced their work and made them less reliant on senior team members to answer questions. Staff engineers found that it took more time to correct AI-generated code than to write their own. Copilot helped all developers automate writing test cases, so engineering managers could spend less time supervising this step.
This controlled implementation and feedback loop helped us understand Copilot's strengths and weaknesses in the context of our organization. The data informed our decision to adopt Copilot, and gave us a more nuanced understanding of how to leverage AI in different roles and tasks.
5. Build a culture of learning
If we want to use AI to enhance our workflows and products, we need to help developers build these skills (and keep them sharp).
Learning sticks best when you apply it. For this reason, I find it useful to keep a running list of problems that our team wants to solve. Instead of building generic AI knowledge, developers can narrow the scope of learning to a concrete use case.
With a clear task in front of them, developers can gain practical experience by answering questions like:
How can AI help me solve this problem?
What frameworks and fundamentals do I need to build a solution?
Using this problem-based model, I give developers the freedom to choose whatever learning resources work best for them. Critically, I also work with teams to set aside dedicated learning time. In the past, we've had book club discussions, pair programming, and "community of practice" sessions to work on problems together.
Because learning is so valuable to the business, I never expect my teams to build new skills outside of their regular work. A healthy engineering culture integrates learning into daily activities. Technically, this means developers spend less time building products — but learning is an investment in your team that improves speed and quality in the long run.
So, if you find yourself overwhelmed by the pressure to adopt AI, I hope that my practical approach encourages you to take a breath. AI is a revolutionary technology, but so was the internet!
It's not sustainable to push out AI features for the sake of "keeping up." Instead of delivering products that may not be valuable or compliant, engineering leaders can peel back the onion to understand our core business problems. From there, we can empower developers to learn the AI skills they need to build truly innovative solutions.
