Hatz AI vs Infobip

Hatz AI runs a fast, dense release train aimed at managed service providers reselling AI to their clients. The base layer keeps absorbing frontier models within days of release — Claude Opus 5, Kimi K3, GLM 5.2 Fast, Gemini 3.6 Flash and 3.5 Flash-Lite, and now DeepSeek V4 Flash 0731 — while platform work fills in around it: AutoTool v2 for automatic tool selection, artifact sharing and a fullscreen presentation mode, and a phone agent that has grown from a feature into a managed fleet. The partner layer arrived in force with Hatz Activate, a rollout console with its own Sales User role.

Two layers are being built at once, and they are diverging in kind. The lower one is model-agnostic plumbing where speed of adoption is the only differentiator, which explains the cadence and the sameness of those entries. The upper one is the partner layer, and it is now moving past acquisition into operations — phone agents can be paused, duplicated onto new numbers, and reassigned between tenants under the same MSP, which is fleet management rather than feature work, and credit usage became self-serve to answer instead of an admin request.

The self-reporting credit usage and per-number minute limits point toward margin control as the next area to thicken — an MSP reselling usage-priced phone agents needs to see consumption per client before it bills, and nothing in these entries closes that loop yet. Expect the phone agent to keep accumulating tenant-level administration rather than new conversational capability.