Under the Hood

I design agent systems that
let CS teams ask questions
instead of building reports.

Most Customer Success (CS) automation is scripts that run on a schedule. I build conversational agents: systems that understand what you're asking, pull data from the right places, reason about it, and give you an answer. Here's how they work.

Philosophy

Automate the process.
Not the relationship.

I start with one question: what's keeping someone from doing their actual job? If a CSM is spending three hours building a report, that's three hours they're not talking to a customer. If a leader is waiting two days for a dashboard, they're making decisions without data.

The tools handle the mechanical work. People handle the conversations that actually drive retention and growth. That's how you scale a post-sale org without losing the part that makes it work.

Agent Architecture

How a question becomes an answer.

Pick a scenario to see how the agent processes it, from the initial question through data gathering, analysis, and response.

Pick a scenario above to see how the agent processes it.

Design Patterns

The building blocks
behind every agent.

The walkthrough above shows one complete flow. Every agent I build uses a combination of these core patterns, adapted to the specific problem.

🔗
Pattern

Multi-Source Data Correlation

Every interesting question requires data from more than one system. These agents query CRM, usage analytics, support history, and external sources in parallel, then cross-reference the results to find patterns no single dashboard would surface.

💬
Pattern

Conversational Interface

Instead of building another dashboard nobody checks, I build agents people can talk to. Ask a question in plain English, get an answer with the data behind it. The interface is the conversation. No training required.

🔄
Pattern

Autonomous Workflows

Some tasks need multiple steps executed in sequence: checking prerequisites, pulling data, transforming it, and delivering results. These agents plan the sequence, handle errors, and complete the workflow without human intervention at each step.

The experience behind the tools

20 years of CS taught me
what's worth automating.

The frameworks, the hard lessons, and why I build the way I do.