How AI is changing software development at JPMorgan Chase

Software engineers aren't being hired to write code anymore.

At least, that's how Gill Haus, CIO of JPMorgan Chase, frames the effect of AI on software development. As generative AI and coding assistants automate more of the coding process, Haus argues that the most valuable engineers are not the ones who can write the most code. Instead, they are the ones who understand customer problems, exercise sound judgment and know what to build in the first place. He also argues that AI is increasing -- not reducing -- the importance of software engineering fundamentals -- such as testing, architecture, security and governance.

In the following interview, Haus discusses how AI is changing the role of software engineers and why computer science fundamentals still matter.

Editor's note: The following transcript was edited for brevity and clarity.

How is AI changing what "valuable" means for a software engineer inside a company like Chase?

Gill Haus: We don't really hire engineers to write code -- we hire them to know what code to write. They must understand the problem and use technology to solve that problem. Before AI and the agentic world, engineers had to also write the code, install the software in your environment, make sure it's configured and write the tests for the code. That slows down product, feature and service delivery for our customers.

We don't really hire engineers to write code -- we hire them to know what code to write.

Enter this new technology: Engineers can focus more on knowing what code to write, versus the actual writing of the code. It doesn't change what the engineer needs to do, but changes how they do the work. It makes it more fun because you can focus on the customer problem rather than on why this thing won't compile, only to find out there was a typo.

If coding is becoming more automated, is there a future for software engineers, or will business leaders be the ones developing software?

Haus: There will be some degree of blend, because you have technical people in product roles or in other roles, and you have people in technology roles that have good product or design skills. But there will still be fundamental questions we need to answer.

First, we are going to have non-engineers who have access to these tools and are working hand in hand with engineers to solve business problems -- they absolutely should do that. But we all have that friend who has an app that they built and put in the App Store or on TestFlight. If you start asking questions such as:

• If you had a million customers on that, how would it work if there was a failure in one of your data centers?
• What happens if you have a new version of the application and you've made changes in the back end?
• How do you handle security?
• What do you do if somebody wants to delete their account?

Those things require a different degree of understanding of how the system behind the scenes works.

Code has transformed. Code is becoming English.

Where do you see today's engineers creating the most value?

Haus: Code has transformed. Code is becoming English. So, the impact is going to still be building those products, features and services. But now you can deliver more in a shorter time.

The value is going to be working through the entire backlog of things that we've wanted to deliver for our customers. We can only do so much because of the time it takes to build and test something, and there are many things that we just leave on the floor that we would love to get to. Now we can get to more of that and develop better solutions because we have more time to ideate.

To what extent should engineers be expected to understand the business rationale behind the features that they're building?

Haus: The important thing for anyone -- not just engineers -- is understanding the 'why' behind what you're doing. This helps you solve the problem appropriately. This doesn't really change at all when you have a technology like AI or agentic to build it. But when you're able to move more quickly, understanding the why is that much more important, particularly if you want to be thoughtful about how many tokens you are spending, etc.

Are computer science degrees and traditional engineering career paths still fit for purpose in an AI-driven environment?

Haus: It will change a bit, but the foundations of good software development and the software delivery lifecycle don't change. They've always been important, and they're much more important now. This includes not just building the code but also testing it in an automated fashion so you have confidence that what you have written will work in production.

When you're a human running code, you want to make sure it isn't broken. If you have a computer now writing code for you, there's a ton of testing that needs to be done. We can't keep up with that unless we automate it, so the good practices we teach in computer science for building software become even more important when you're engaging with AI.

However, there will still be areas that require specialists. What is the right architecture that we should have? You can ask AI what it recommends, but you also need to understand the architecture and have the judgment to say what it should be.

How do you balance encouraging engineers to adopt new AI tools with ensuring they still understand underlying systems and fundamentals?

Haus: We offer our teams guidance on what we want them to build, but the best way to learn a new technology is to play with it. At the same time, we provide the right training, guidance and career pathing around foundational engineering skills, so AI is one more capability we will be teaching our teams. The question isn't whether engineers are learning AI tools. It's how they are using those tools in the work they're already doing.

CIOs should make sure people are brought along on the journey. This technology is going to make life a lot easier, but it requires people to think differently. When you're writing in natural language rather than code, team structures may need to adjust to ensure newer team members still learn how systems work, so they know how to respond to production issues while also learning the tools.

It's still new. It's only been a few years since ChatGPT became a big thing. Claude Code didn't reach its current level of capability until late last year, so we're all still learning what this means for how people will work. That's why I come back to the core principles of building good software. That really hasn't changed. If you're training people in those fundamentals -- good security, architecture and software engineering practices -- AI becomes just another tool for applying them.

Where do you see the risk in over-relying on AI tools for software development inside a highly regulated institution like Chase?

Haus: Security is paramount, and so is privacy. Also, when we use AI in customer-facing contexts, we are very deliberate about keeping a human in the loop before anything is delivered. The guardrails are improving, and over time, we will move toward more agentic experiences. But today, human oversight remains essential so we can intervene if something is off.

Confidence in production comes from strong engineering practices -- automated testing, automated deployment and automated rollback.

In software development, confidence in production comes from strong engineering practices -- automated testing, automated deployment and automated rollback. With those controls in place, if an AI agent is writing code, we can detect issues before they reach production and respond quickly. We are also careful about internal use, ensuring appropriate access controls and governance.

As long as controls and practices scale with AI systems, you get security, privacy and all of that on the other side. It's when you have manual processes that problems emerge. We're focused on burning down those manual processes so we can keep pace as machines generate and run code.

How has adoption of AI tools been inside your organization? Are engineers excited, or is there resistance?

Haus: There's going to be trepidation with any new technology. There are those who are very excited about it, and that has its pros and its cons. The pro is a ton of energy and curiosity. The con is that this excitement can turn into frustration when rollout takes longer than they'd like. We must be controlled and thoughtful, and this can slow things down.

Conversely, some people are afraid of AI and what it means for their future role. CIOs must gauge if people are engaged or not engaged, and that's where we spend a lot of time communicating and making the tools available to our team and listening to whether things are working. We must also be very transparent about the fact that we know things will change, but that we will be there for our teams as we move.

I understand why people get scared of AI, but employees are the ones who still have the agency to decide what agents should do.

Tim Murphy is a site editor and writer for the IT Strategy team at TechTarget.