What happens when AI moves beyond answering questions and starts managing workflows, connecting ideas, and building software?
At our latest Enrich AI Show & Tell, members gave us a firsthand look at what they're building with AI—and how they're applying it to solve real problems.
Unlike a traditional presentation, AI Show & Tell is an opportunity for members to demonstrate projects they're actively experimenting with, exchange ideas, challenge one another's approaches, and learn directly from peers.
This session featured Enrich members Gavin Roy and Mona Chitnis, who shared three fascinating applications of agentic AI, from a smarter email assistant to a personal knowledge system and an increasingly autonomous software development lifecycle.
Here's a look at what they shared, and the ideas that sparked discussion.
1. An AI email assistant that handles the noise for you
What if your inbox could identify what deserves your attention, handle unwanted messages, and complete administrative tasks without you having to ask?
Gavin Roy demonstrated an AI-powered email application he built to do exactly that.
What started as an experiment in automatically handling unwanted sales emails evolved into a much more sophisticated system. His application can:
Prioritize important messages using AI-driven attention rules.
Handle unwanted outreach by automatically responding and filtering follow-ups.
Consolidate repetitive notifications, such as service status updates, into summaries.
Connect related conversations across separate email threads.
Coordinate with other AI agents to automate tasks like collecting invoices and preparing monthly expenses.
One particularly interesting example was Gavin's expense-reporting workflow. Instead of manually tracking down invoices across email and vendor portals, his agents gather the necessary information and prepare the expense spreadsheet—giving him back roughly two hours each month.
Gavin also built in configurable approval steps and the option to use local AI models, giving users more control over automation, privacy, and costs.
2. A personal "second brain" that connects the dots for you
We're surrounded by information: meeting notes, articles, newsletters, podcasts, research papers, and ideas we capture throughout the day.
But how often do we actually return to those notes—or remember the right insight when we need it?
Mona Chitnis tackled that challenge by building her own AI-powered second brain.
Her premise was simple: capturing information is no longer the hard part. The real challenge is synthesizing it, remembering what's important, and retrieving the right insight at the right moment.
Using tools including Claude, Obsidian, and Granola, Mona developed a system that collects information from multiple sources and continuously distills it into useful knowledge.
Her system:
Consolidates meeting notes, articles, videos, podcasts, and voice memos.
Produces personalized daily briefings highlighting relevant developments.
Identifies emerging themes and alternative perspectives worth exploring.
Synthesizes daily insights into weekly and monthly learnings.
Uses accumulated context to help with research and meeting preparation.
Rather than becoming another archive of information, the system is designed to develop a deeper understanding of Mona's interests and priorities over time.
One especially interesting feature is its ability to highlight why a particular research paper or industry development matters to her, based on what she's previously been exploring.
3. An engineering team moving toward autonomous software development
For his second demonstration, Gavin took the group inside the AI-powered engineering systems his team has been building at AWeber.
Their ambition? Move toward a software development lifecycle where AI agents can independently identify issues, create tasks, write code, review changes, and support deployment.
Gavin walked through an interconnected ecosystem of agents that work across Slack, Jira, GitHub, and internal engineering systems.
For example, when a customer reports an issue, an agent can pick up the task, investigate it, develop a fix, and move it through the pull request process.
The results Gavin shared were striking:
100% of customer-reported issues were being resolved within SLA at the time of the presentation.
Approximately 80% of recent agent-generated pull requests were accepted.
Only around 35–40% of pull requests were coming from developers working interactively with Claude Code, rather than more autonomous workflows.
The team reported significant improvements in release frequency and a reduction in customer-facing incidents caused by internal changes.
The team is now working toward a unified platform where agents can collaborate, manage workflows, track costs, and involve humans when decisions or intervention are needed.
The conversations were just as interesting as the demos
What made this session particularly valuable wasn't just seeing what members had built. It was hearing the questions and ideas that followed.
Members explored how to balance agent autonomy with appropriate safeguards, whether local AI models could reduce operating costs, how to limit agents' access to sensitive information, and how to measure the actual return on AI investments.
They also noticed an interesting connection between Gavin's email assistant and Mona's second brain: one helps eliminate immediate distractions, while the other helps preserve and synthesize knowledge for the future.
Could the two systems eventually work together?
That's exactly the kind of unexpected connection that happens when experienced technology leaders share what they're building and invite others to challenge and expand on their ideas.
Curious what else Enrich members are building and want to bring your own ideas and questions into the conversation? Apply to join Enrich.
