Product Update: Building Better Control, Clarity, and Continuity in AI-Led CX

As AI systems become more deeply embedded in customer experience, the work shifts from getting an agent live to operating it well.

That means visibility into how decisions are made. It means easier control over brand guidelines. It means continuity across conversations, not isolated interactions. And it means giving customers simple ways to take value with them when they leave a session.

Over the past few weeks, we’ve shipped a set of updates that all point in the same direction: more control for teams, and more coherence for customers.

Here’s a look at what’s new.

Conversation Debugger Agent

Understand how your AI thinks, not just what it says

One of the most common questions we hear from CX and digital teams is simple: Why did the agent answer this way?

The new Conversation Debugger Agent is designed to answer that question clearly.

It allows teams to inspect agent performance at the conversation level, including:

• The sources the agent used to generate a response
• How different inputs and guidelines influenced the outcome
• Where answers are coming from when multiple data sources are involved

This makes debugging faster, but more importantly, it makes improvement deliberate. Instead of guessing which rule, document, or integration caused an issue, teams can see it directly and adjust with confidence.

For brands operating in regulated or high-accuracy environments, this also adds an extra layer of reassurance and auditability.

New Integrations, Now Native to the Dashboard

Less setup friction. More usable data.

We’ve expanded our integration ecosystem and made it significantly easier to access and manage directly from the dashboard.

Newly available integrations include:

• Magento
• Ship24
• mParticle
• Salesforce Commerce Cloud
• Narvar

What’s changed isn’t just the list. It’s the experience.

These integrations are now discoverable and configurable from the dashboard itself, reducing reliance on manual setup or back-and-forth with implementation teams. The result is faster time-to-value and cleaner data flows into the agent.

For commerce and CX teams, this means the AI has better context across orders, shipments, profiles, and events without added operational overhead.

AI-Powered Dashboard Navigation & Guideline Editing

Reducing the most painful part of AI administration

Writing and maintaining guidelines is one of the most important parts of deploying a reliable AI agent. It’s also one of the most time-consuming.

Our internal helpdesk assistant can now:

• Navigate admins to the correct dashboard page based on intent
• Create new guidelines through natural language
• Edit existing guidelines without manual searching or rewriting

Instead of learning the structure of the system, admins can simply describe what they want to change. The assistant handles the navigation and the update.

This significantly lowers the effort required to keep agents accurate, on-brand, and up to date. Especially for teams managing large catalogs or evolving policies.

User Memory System

Personalization that persists, without becoming stale

We’ve begun a gradual rollout of our User Memory System, designed to give AI agents continuity across conversations.

The agent can now:

• Remember returning users across sessions
• Extract and store preferences, profile details, and relevant past context
• Use a decision cycle to update or discard memories as things change

For customers, this means conversations feel less repetitive and more helpful. The agent already knows preferences, past issues, and context, so interactions start further along instead of from zero.

For brands, this enables personalization that feels earned and useful, not intrusive or brittle. Memory evolves as the customer does.

Send Transcript Agent

A small feature that removes real friction

We’ve also launched a new Send Transcript Agent, enabled by default for all customers.

If a shopper says, “Send this chat to my email,” the agent can now do exactly that.

No forms. No follow-ups. No copying and pasting.

This is especially useful for:

• Product recommendations
• Routine or regimen guidance
• Order explanations
• Support conversations customers want to reference later

It’s a simple capability, but one that respects how customers actually use chat - as a place to think, decide, and then continue later.

A Common Theme

Individually, these updates solve specific problems. Together, they reflect a broader focus.

AI in customer experience is no longer about novelty. It’s about operability, trust, and continuity.

These features are designed to help teams understand their agents better, manage them with less effort, and deliver experiences that improve over time rather than reset every session.

More to come soon.