Developer Experience and AI in GovTech with Tracy Bannon from MITRE

About Tracy Bannon

Tracy “Trac” Bannon is a seasoned software architect, engineer, and researcher leading MITRE’s ArchAITecture initiative on AI in the Software Development Lifecycle (SDLC). She brings decades of experience driving secure, AI-augmented software practices across government and industry. A DevEx advocate and podcaster, Trac amplifies human/machine teaming, trust, and diversity in tech through talks, workshops, and Real Technologists.

AI Will Change Software Delivery Life Cycle in the Next Six Months

In this episode of the HangarDX podcast Tracy Bannon shares her bold predictions for how AI agents will radically change jobs of software engineers, her vision of what the collaboration of humans and AI will look like in the near future, and explains the tech side of government.

What is MITRE?

MITRE is a federally funded research and development center (FFRDC), meaning it’s chartered by Congress to serve the public interest. MITRE doesn’t compete with private companies in the industry; it bridges the gap between government, industry, and academia, helping them collaborate more effectively on complex challenges.

What does software engineering look like in the government?

Technology touches nearly every domain you can imagine - defense, health, education, transportation, and more. If there is a kind of software in the private sector, there's likely a version of it happening in government, too. But the way software is acquired and developed is fundamentally different.

The government often doesn't build software directly. Instead, it partners with industry. Sometimes this means a company delivers a completed solution. Other times, there's tight collaboration with continuous feedback loops. There are also government-run “software factories” staffed by a mix of industry partners, civilian government workers, and even military personnel like sailors, airmen, or corpsmen, who also help to design, develop, and deliver software.

When we think about government tech, we think of legacy systems and an outdated tech stack. Is that a true picture?

The government moves at the pace it can. Yes, legacy systems exist, but the same is true in the private sector. Modernization is ongoing, and increasingly, it's continuous rather than one-off.

The government is slow, in part, because it has to be. Government software often handles critical services like taxes or defense, where errors are unacceptable.

Risk aversion is baked in and appropriate. That said, the culture is shifting. Early-stage experimentation is more common, and agile practices are gaining traction.

How is AI changing software engineering?

Today, individual tasks, like drafting requirements or writing code, are supported by AI tools. Every task in the software development lifecycle is still owned by the same human.But soon, even in six months, it will be dramatically different. Entire workflows will be restructured around teams of AI agents. We're already starting to see agents being brought in. At first, it's humans pair programming with AI kind of mentality. Soon, the software development life cycle will change to adapt for teams of AI agents. That’s already happening!

Developers will have to become software engineers

Those who are self-taught developers will have to get additional education and become software engineers. Our job will be to manage this digital platform orchestrating AI agents. Every part of the SDLC is being reimagined. I expect that in the near future, software engineers will be able to explain in natural language what they want to build, and that the “digital platform” will be able to accomplish whatever is needed to put software into production. Today, tools are fragmented, soon it will all be under one umbrella.

Think of it as one big collaborative session between humans and AI, spanning planning, coding, testing, deployment, and monitoring. And all of the AI technologies are plugged into it.

Human in the loop vs human on the loop

In that changed SDLC scenario, we need to be even more focused on developer experience and how software engineers interact with all that technology. We need to focus on humans in the loop, ot better said, humans on the loop.

Human in the loop is what we have now -  the human is using the tool and owns the outcome, the result that comes out of the tool. When I use the code completion tool, for example,  I'm the human in the loop. I am responsible for the quality of the software, if tests are written, or for what happens downstream.

But, as more trust builds with AI tools, I won’t have to be in the loop in the same way. Instead, I am on the loop. I'm just looking for where I must get involved, as opposed to being constantly involved.

Humans on the loop will just be there to make sure there is a safety net, that we can stop something if it’s necessary.

Is the government a good place for software engineers to work?

For people who are passionate about technology, the government sector is a great place to be. It's not hundreds of years behind. It's not even decades behind. It works at a different pace, but it also impacts at a different scale.