The rapid growth of artificial intelligence (AI) technologies has made coding more accessible and allowed individuals with little or no software experience to integrate it into their workflows.
Engineering Management and Systems Engineering (EMSE) Professor John Helveston and EMSE PhD candidate Pingfan Hu recently hosted a workshop titled “Agentic Workflows with Claude Code” to address this shift and empower members of the GW community to code responsibly using AI agents. This event was held in collaboration with the GW Trustworthy AI Initiative and GW Coders.
“The goal of the workshop was to cover the workflow Pingfan and I have been using for almost a year now, and to help people get set up in Claude Code and see how it can work for them and their needs,” Prof. Helveston explained.
Participants were asked to install a few applications and agentic tools prior to the workshop to enable them to follow hands-on. Helveston and Hu noted that Claude Code, their agent of choice for the workshop, requires a paid subscription, but participants were also welcome to use Google Gemini CLI, which is free with a Google account and works similarly.
The workshop began with a presentation about the basics of AI agents, including the critical distinction between AI agents and AI chatbots. A chatbot is best suited for one-off questions, while an agent is more appropriate for longer-term projects and automating workflows, as Helveston and Hu demonstrated.
Throughout the workshop, participants had several opportunities to practice building their own website with Claude Code. While a chatbot may only provide you with the necessary files to build a website, which you would then need to download and deploy yourself, an agent goes a step further by actually building the website directly on your computer. For example, it created Helveston’s example website showcasing Tom Hanks’ career in under four minutes.
Helveston and Hu stressed the significance of careful stewardship of the agents to ensure that you don’t inadvertently give them permission to private files and that their outputs are correct and clear.
“We want people to understand that agents can do almost anything with their files, like write code, write and summarize documents, automate tasks, build HTML, generate images, analyze data, and more,” Hu said.
But Hu emphasized the importance of not relying too heavily on agents. “You should think of yourself as the pilot and the agent as your copilot. You are guiding the flight. Even though AI is extremely powerful, you should not give up on thinking and coming up with your own ideas,” he concluded. Anthropic’s own research has shown that users with greater domain expertise in their field have more success with agentic tools like Claude Code.
The workshop had over 180 participants. Participants were highly engaged, and even helped one another troubleshoot issues in the comments.
“We were thrilled to have such an active audience, with many people trying out these tools for the first time alongside us and asking a lot of questions. We look forward to future workshops where we can continue to share how we are exploring the capabilities and limits of AI,” Helveston shared.
As Google and Alphabet CEO Sundar Pichai noted in a 2025 interview about the effects of AI on coding: “Things are getting more approachable, it’s getting exciting again, and the amazing thing is, it’s only going to get better.”
A recording of the workshop is available on YouTube.