This changes everything: Customer.io's Spring 2026 release 

What if your AI agent could do anything in your workspace that you can, and it got smarter about your business every time you used it?

That's the idea behind this release, and this replay walks through the thinking, the live demo, and every feature that shipped alongside it.

• How a single AI agent builds full campaigns, audits underperforming ones, and remembers your business context and preferences across sessions
• How LLM actions score and personalize content for every individual profile that moves through a campaign
• How goals connect every campaign and broadcast to the revenue or lifecycle metric you're actually trying to move
• How WhatsApp and Line plug into your existing campaign builder as two new channels, no separate tool required
• Where the expanded MCP server, an upcoming CLI, and an early-stage Slack integration fit into Customer.io's AI roadmap
• A handful of smaller releases built to make everyday work faster, from a new homepage to a workspace performance dashboard

Some platforms are shipping a copywriting agent, a deliverability agent, and a segmentation agent as separate tools. Customer.io built one agent instead, with access to everything you can already do through the product. The reasoning: a marketer building a campaign doesn't hand off deliverability questions to a different specialist mid-task, so an agent split into narrow tools just adds handoffs without adding value.

The agent runs on a business context hub, where you add your brand voice, tone, and background so its output sounds like you instead of a generic assistant. It also carries memory across sessions, so you can pick up a conversation about a campaign you were building yesterday without repeating yourself. That same intelligence follows you outside Customer.io too, through an MCP server that connects to tools like Claude, Cursor, and ChatGPT, with an agent CLI and a Slack integration both in progress.

Three features from this release, the agent, LLM actions, and goals, are live in beta. Customer.io is actively shaping all three based on customer feedback.

AI agent A single agent with full access to your workspace: building campaigns and segments from a plain-language prompt, auditing existing campaigns for issues like missing conversion goals, and remembering your business context and preferences across conversations. Every edit or deletion the agent makes still requires your approval, individually or through a broad approval setting.

LLM actions A campaign step that runs an LLM against every individual profile passing through it, using the fields and instructions you define. One use case demoed live: scoring free-plan users on likelihood to upgrade, then generating a personalized upgrade email for anyone who scores above a set threshold. You can assign a lighter model, like Gemini 2.5 Flash, to high-volume steps and a stronger model to more selective ones. Paid plans get a sample allotment of 100,000 AI credits to start experimenting with LLM actions under the new credit model.

Goals A way to tie every campaign and broadcast back to a single metric you care about, like credit pack purchases or a trial-to-paid conversion, instead of tracking opens and clicks in isolation. Goals break down attribution by campaign, by individual message, and by channel, including email, in-app, SMS, WhatsApp, and Line, so you can set your own rules for what counts as a conversion on each channel.

WhatsApp Native, self-service setup: connect your Meta account, sync your approved templates, and send from the same workflow builder you already use for email and SMS. WhatsApp has more than 3 billion monthly active users across 180 countries, with open rates north of 90%.

Line The dominant messaging app in Japan, Thailand, Taiwan, and Indonesia, now available alongside your other channels in the same campaign builder, with the same drag-and-drop and Liquid personalization support.

Expanded MCP server Customer.io's MCP server previously supported read-only access. This release adds the ability to create and edit campaigns, segments, and broadcasts directly through MCP clients, plus access to your agent's stored memory and preferences, and support for routines, scheduled tasks like a weekly campaign performance report.

Agent CLI (coming soon) A command-line interface built for how LLMs actually work best: discovering available commands and schemas on the fly instead of guessing at a REST API payload. The agent itself will use this CLI internally to make workspace changes, and Customer.io plans to ship a set of skills alongside it so an agent can look up best practices, like creating a segment before a campaign, before acting.

Slack integration (early exploration) An early-stage concept for a Slack bot that can answer questions about your customer data directly in a channel or thread, such as why a specific message didn't send to a specific person.

New homepage Surfaces what you were recently working on and, if you choose, what your teammates are working on, so you can jump back into a campaign or broadcast without hunting for it.

Workspace performance dashboard A workspace health overview with specific cleanup suggestions, like flagging unused segments to archive.

Universal search Search across campaigns, broadcasts, and more from one search bar, ordered to surface what you touched most recently.

Profile search (coming soon) Look up a profile by email address directly from universal search.

Design Studio dark mode preview Preview how an email renders in forced dark mode for Outlook on web and for Gmail, including Android apps, before you send it.

Can the AI agent do anything in my Customer.io workspace? The agent can do anything a current user can do in the workspace, with one exception: every edit or deletion requires your approval first, either one at a time or through a broader approval setting you control.

Will my data be used to train AI models? No.

Can I turn off AI for my Customer.io account? Yes. An account admin can disable AI features through account settings.

Are the agent, LLM actions, and goals fully released or still in beta? All three are in beta. Customer.io is actively iterating on them based on customer feedback and bug reports.

How much does it cost to try LLM actions? Every paid plan received a sample allotment of 100,000 AI credits to experiment with LLM actions under the new credit model.

Do I need a separate tool to send WhatsApp or Line messages? No. Once WhatsApp and Line are connected, both channels appear in the same workflow builder you already use for email, SMS, push, and in-app messages, with the same drag-and-drop building and Liquid personalization.