Notes on running conversations at scale

WhatsApp Business API: A Practical Guide for Growing Teams

What the WhatsApp Business API actually is, how it differs from the free Business App, what it costs per conversation, and how to get approved without wasting a month.

AI Agent vs Chatbot: What Actually Changed

Rule-based chatbots failed because they could only match keywords. Here is what LLM-based agents do differently, where they still fail, and how to tell the two apart when buying.

How to Cut First Response Time Without Hiring

First response time predicts conversion better than almost any other support metric. Here is what drives it, and five changes that move it without adding headcount.

Getting WhatsApp Message Templates Approved First Time

Template rejections are predictable and almost always avoidable. The category rules, the formatting requirements, and the phrasing that gets you rejected.

Abandoned Cart Recovery on WhatsApp: What Works

WhatsApp cart recovery converts several times better than email, and it is much easier to get wrong. Timing, message count, and when a discount destroys margin.

Shared Inbox Best Practices for Messaging Teams

Ownership rules, routing, internal notes and escalation paths that stop two agents answering the same customer and stop threads falling through the gaps.

WhatsApp Broadcast Compliance: Staying Off the Blocklist

Opt-in requirements, quality ratings, messaging tiers and the sending behaviour that gets numbers restricted. What to do before your first campaign.

Running Multilingual Support Without Hiring Per Language

Detection, tone consistency across languages, where machine translation still fails, and how to decide which languages justify a native speaker.

The Conversation Metrics That Actually Predict Revenue

Deflection rate and CSAT are easy to game. These are the metrics that correlate with retention and revenue, and how to instrument them properly.

Guardrails: Keeping an AI Agent From Saying the Wrong Thing

Scope control, grounding, confidence thresholds and audit trails. The controls that make an AI agent safe to put in front of customers.