Building a Copilot Agent — From Problem to Proof
Featuring special guest speaker Mrs. Gin Jespersen
The Concho Valley Technology Alliance, in partnership with the ASU-SBDC, hosted Building a Copilot Agent: From Problem to Proof, an interactive workshop focused on moving beyond basic AI prompting and exploring how AI agents can solve real-world problems.
Attendees learned that while tools like ChatGPT and Microsoft Copilot can answer individual questions, AI agents can be designed with a specific purpose, instructions, knowledge, and guardrails to perform repeatable tasks.
Start With the Problem, Not the AI
A major theme of the evening was to start with the problem, not the AI. Participants were encouraged to identify repetitive, manual, or time-consuming tasks and consider how an agent could help. The presenters shared examples of agents being used across HR, sales, marketing, recruiting, finance, IT, education, customer service, real estate, and research. One real-world case study demonstrated how connected AI agents transformed a complex HR and compensation process that previously required multiple people and hours of work.
During the hands-on portion, attendees saw agents built in Microsoft Copilot, including a real estate assistant that could answer questions about property listings and a customer complaint assistant that could reference policies and draft responses. The group also explored how agents can search information, adjust tone, translate content, summarize data, and complete recurring tasks—all while operating within boundaries established by the person building the agent.
The evening concluded by looking at what happens when multiple AI agents work together as a team. Specialized agents can take on roles such as researcher, writer, designer, or quality assurance, while another agent coordinates their work. The biggest takeaway for those who couldn’t attend: don’t start by asking what you can build with AI. Start with a problem worth solving, clearly define what success looks like, and build from there.