AIM 2026: Your Digital Double: A Hands-On Agent-Building Workshop

Your Digital Double: A Hands-On Agent-Building Workshop

Kristi Fickert of Curie and Justin Jones of SmartRent, co-founders of the AI Exchange, challenged attendees to move beyond prompting and begin thinking about AI as a workforce. Instead of giving an LLM individual instructions each time, an agent receives an objective, context, guardrails, and tools, then reasons through how to achieve the desired outcome.

Their advice was intentionally practical: start small and boring. Rather than attempting to automate an entire department, identify a repetitive task you already understand well. Predictable work is easier to evaluate, safer to experiment with, and provides a clear benchmark for whether the agent is performing correctly.

The speakers introduced a simple framework for building an agent: Role, Rules, Routine. Define who the agent is and the outcome it owns; establish guardrails around compliance, accuracy, tone, and human review; then define what it should consistently produce or accomplish.

 During the session, they demonstrated a multifamily marketing agent capable of turning a property brief into listing copy, social captions, prospect email content, and compliance flags. The agent was provided brand guidelines and Fair Housing information as context, allowing it to evaluate its own output and flag content requiring human review. 

More advanced applications involve multiple specialized sub-agents—such as researchers, strategists, critics, and writers—working together. For lean multifamily marketing teams, the speakers suggested asking a simple question: If you had the budget to hire anyone, what roles would you add? Those needs can become a roadmap for future agents.

The ultimate goal isn't simply saving time. It's expanding what an individual or team is capable of accomplishing.

Key Takeaways

• Think outcomes, not prompts. Tell agents what success looks like rather than scripting every step.
• Start with boring work. Repetitive, predictable tasks make the safest and easiest first agents.
• Remember Role, Rules, Routine. Define the job, guardrails, and repeatable workflow.
• Give agents context. Brand standards, policies, documentation, and examples improve output quality.
• Keep humans in the loop. Especially in regulated areas such as Fair Housing, people remain accountable for what AI produces.
• Measure outcomes, not novelty. Don't report that AI wrote captions; show that an agent produced content for 10 properties in five minutes or created measurable capacity for higher-value work.