Insights

Field notes from shipping AI in production.

Architecture decisions, regulatory traps, prompt patterns, and the hard-won lessons from real engagements.

How to ship voice AI inside a regulated perimeter (without lying to your compliance officer).

The reference architecture, the consent patterns, and the audit logging shape we use for regulated voice deployments — distilled from the Hayah engagement.

Moussa Sangare · 12 min read · Mar 15, 2026

↳Read the article

Architecture · 8 min

Why we chose Ultravox over the obvious vendors.

Latency, language support, and the data residency story — what mattered, what didn't.

Prompt Design · 6 min

Prompts that survive contact with users.

The patterns we use to make agents robust to messy real-world inputs.

Founder Track · 10 min

If you're shipping your first AI SaaS, read this first.

The five mistakes we see founders make in the first 30 days — and how to skip them.

Stack · 7 min

n8n vs custom code for AI workflows.

When to reach for low-code, when to build, and when the answer is "neither, do it manually for now."

Compliance · 9 min

SOC 2, ISO 27001, DIFC: the consultant's cheat sheet.

What each audit actually checks for in an AI system, and what's just noise.

Process · 5 min

Why our default engagement is 6 weeks, not 6 months.

What changes about scope, deliverables, and accountability when you compress the timeline.