Insights
Proof & field notes.
How we build AI systems that do real work — case studies from live client engagements and practical guides from the things we ship.
Case studies
Shipped, in production
Zoho CRM follow-up agent
A human-in-the-loop AI agent that drafts context-aware email and LinkedIn follow-ups from Zoho CRM — researched, reviewed, and sent on approval. No rip-and-replace.
WhatsApp + voice lead agent
NudgeFlow orchestrates follow-ups across WhatsApp and AI voice calls — policy-driven, idempotent, with human handoff. Built for ClickPe (YC S23).
Distributor catalog pipeline
A distributor was retyping supplier price-list PDFs by hand. We built an AI pipeline that extracts, structures, and matches them into a clean catalog — in minutes, not days.
X DM outreach engine
An automated X/Twitter DM system built to stay unbanned: humanized timing, office hours, daily caps, reply detection, and A/B sequencing with a management dashboard.
AI cold-email on AWS SES
A cold-email system you own: AI personalizes copy, AWS SES sends one message per recipient, and suppression + bounce/complaint handling protect the sending domain.
Guides
Field notes
SES tenant-level suppression
Amazon SES added per-tenant suppression lists in June 2026. Here is how account, configuration-set, and tenant-level suppression fit together — and when to use each.
FTC disclosure for AI UGC
Running AI-generated UGC ads in 2026? The FTC now expects two disclosures — commercial relationship and AI involvement. Here is the practical checklist and placement rules.
WhatsApp + voice follow-up guide
How to orchestrate lead follow-ups across WhatsApp and AI voice calls without spamming — policies, idempotency, human handoff, and the funnel metrics that matter.
Price list PDF → Excel
Stop retyping supplier price lists. Here is how an AI extraction pipeline turns messy distributor PDFs into a clean, matched Excel catalog — and where to put a human check.
Automate X DMs safely
The safe-outreach playbook for automated X/Twitter DMs: human-paced sends, office hours, daily caps, reply detection, and the patterns that trip detection.