AI Thinking · From the Field

Insights on
Building AI That Works

Product teardowns, agentic AI deep-dives, and honest lessons from shipping intelligent systems in production. No hype — just what actually works.

Latest Writing

AI Explorations
& Reflections

How I'd Fix Confluence's Broken Search Experience

Confluence stores your company's brain — but its search is so broken that people ask Slack instead. This is a deep dive into why the current system fails, and a concrete semantic + AI redesign proposal that would actually work. Includes a proposed RAG architecture, UX wireframes in words, and why embedding-based search changes everything for enterprise knowledge management.

OnboardIQ — AI Onboarding Assistant for B2B SaaS

B2B SaaS companies lose 30% of customers in the first 90 days — not because the product is bad, but because onboarding is generic. Here's how an LLM co-pilot changes that.

What I Learned Building AI Tools Inside a Large Enterprise

Shipping an LLM product in a large enterprise is nothing like building in a startup. Honest lessons from shipping a RAG-based tool for 1,000+ internal users — what worked, what nearly killed it.

The Agentic AI Stack: What to Use and When

LangGraph vs CrewAI vs AutoGen — a practical comparison of agent frameworks for production use. When each framework shines, and what they all get wrong.

Why Most RAG Systems Fail (And How to Fix Yours)

Chunking strategy, retrieval quality, prompt design, and hallucination mitigation — the four places where enterprise RAG deployments fall apart, with concrete fixes for each.

Next Article
Currently writing about responsible AI guardrails in production systems.
Coming Soon
What We Cover

Topics We Write About

Agentic AI Systems RAG Architecture LLM Product Design Workflow Automation AI Adoption in Enterprise Responsible AI Product Teardowns AI PM Lessons 0→1 AI Products Multi-Agent Design