Customer.io vs Gumloop

Customer.io just shipped its largest release in years on April 8 — an AI 'agent' that can act inside the product, LLM actions for in-campaign personalization, expanded WhatsApp and LINE channels, and a UI overhaul. The weeks since have been quieter follow-on work: AI-driven style generation in Design Studio, multi-account switching, and small but useful campaign-workflow flexibility. The product is operating on two clearly distinct tracks at once.

The AI-agent push reads as a durable strategic bet rather than a one-off announcement — recent shipping keeps extending AI surfaces (Design Studio styles from a URL) and the underlying primitives that AI uses (journey attributes that LLM actions can write into). In parallel, the team is filling in long-standing workflow friction: changing campaign trigger type after creation, resetting message content without rebuilding blocks, juggling multiple workspace accounts. The shape of the roadmap looks like 'agent on top, workflow primitives underneath.'

Expect the agent's actuation scope to widen — more skills, more Routines for recurring tasks, deeper use of AI credits as a billing primitive — while the quieter QoL cadence keeps chipping at friction points marketers raise in support tickets.