Introducing Proactive Agents: User Memory and Outbound Voice built for the AI Concierge

Today, we’re announcing Proactive Agents, our latest platform launch that brings together user memory and outbound voice so AI agents can initiate conversations, remember customers across interactions, and help you build relationships that grow stronger over time.

Customers are the foundation of business growth, yet even brands with the strongest intentions to put customers first have been constrained by systems built for transactions, not relationships. And while AI agents have made it possible to deliver high-quality interactions at scale, until now, they’ve been reactive. It’s still not enough.

The best customer experiences have always been built on memory, anticipation, and timing. A concierge knows who you are, recalls your preferences, and reaches out before you even need to ask. This should be the standard for AI concierge, too. It doesn’t just answer questions, but shapes experiences that feel personal, timely, and genuinely helpful.

Alongside Proactive Agents, we’re also introducing Agent Workbench, a new way to debug and iterate on agent performance with speed and confidence. Check out deeper dives into each of the launches on our launch page. 

User memory: Personalization and context that carries across every interaction

User memory gives AI agents the ability to remember customers across conversations and channels. Conversational history persists across sessions, so interactions pick up right where they left off. A customer can troubleshoot an issue, step away for days, and return to an agent that already knows the full story.

Beyond remembering conversation context, user memory captures key signals like feature requests, sentiment, and stated preferences as structured metadata. For example, an agent recalls a customer's shirt size during checkout, applies a birthday discount automatically, or proactively reaches out when a long-awaited product launches. These small acts of recognition turn transactions into trust and loyalty.

For teams, user memory helps surface the full arc of a customer's journey in one place. The user view consolidates conversation history, outreach attempts, and custom signals into a single profile, making it easy to spot trends and act on opportunities. With full API access, expiration controls, and built-in redaction policies, teams stay in control of what's stored and for how long.

Read more in our deep dive on user memory.

Outbound voice: Proactive engagement that meets customers at every touch point

Outbound voice enables AI agents to initiate calls at the right moment, with the right context. It's built for scenarios where speed and scale matter, like appointment coordination, informational calls with follow-up actions, or high-volume outreach where the goal is consistent across every interaction. Reliable outbound calling requires navigating real-world conditions like detecting voicemail, handling robocall screeners, and staying compliant with calling regulations.

We’ve fine-tuned our voice capabilities to handle all of these automatically, so teams can reach customers reliably without manual overhead. Outbound AI logic is powered by Agent Operating Procedures (AOPs), giving non-technical teams the ability to design workflows using natural language prompts. Branded caller IDs increase trust and answer rates, while low-latency turn detection keeps conversations feeling natural from the first greeting.

Every call captures preferences and intent that get stored in a customer’s profile, helping inform future interactions and reasons to call. By automating repetitive outreach, teams can focus their energy on interactions that require human interaction.

Read more in our deep dive on outbound voice.

Agent Workbench: Debug and optimize agents without engineering support

Behind every proactive experience is the ability for teams to maintain and improve agents over time. Agent Workbench is a built-in AI chat assistant that helps you identify, understand, and resolve issues with your agent as well as find opportunities to improve workflows. When something goes wrong, teams can reduce reliance on engineering tickets or vendor support. When you ask Agent Workbench a question, it references the agent’s logs, reasoning, traces, and latency events, acting as your debugging partner to surface actionable suggestions. Pre-built sample prompts make it easy to investigate common issues.

Instead of guessing, teams get clear answers grounded in context and end-to-end autonomy in keeping agents running smoothly.

Read more in our deep dive on Agent Workbench.

Shaping the future of concierge customer experiences

We believe every customer deserves an AI concierge. With Proactive Agents, we’re taking the next step toward making that standard achievable at scale.

When brands deliver trusted, context-rich experiences, customers willingly share more, engage more deeply, and build lasting loyalty. That flywheel of trust and personalization is what will define the next generation of customer experiences, and we’re excited to shape it alongside our customers and partners. To learn more, book a demo with our team.