Twilio makes Segment Conversation Memory Integration generally available

Closing the gap between marketing data and live conversations

Twilio describes a common problem: a customer moves from an email workflow to a live voice call, SMS thread or conversational AI agent, and their digital history disappears. The agent starts from zero and the customer has to repeat details. Bringing long-term CDP context into those moments has traditionally required engineering teams to build and maintain custom webhooks and middleware pipelines, Twilio says.

The company positions Segment CDP as the long-term system of record for identity, consent and cross-channel marketing automation. Live communications, it argues, have a different requirement: sub-300ms working memory. The integration connects Segment to Conversation Memory, which Twilio calls its high-speed conversational recall layer.

What the integration does

Twilio highlights three main benefits:

• Real-time context: Segment traits stream into live interactions so agents can skip redundant questions such as "What is your name and account number?" and move to validating known facts.
• Continuous cross-channel conversations: customer identity, preferences and dialogue context persist across web, voice, SMS and Flex touchpoints, so agents can pick up where the last interaction ended.
• Less custom engineering: the connector is available in Segment Unify and the Destinations Catalog, is click-to-configure, and, according to Twilio, streams profiles into live channels in minutes instead of months at no additional data ingestion charge.

Twilio also claims 10x to 15x lower LLM token costs. The reasoning is that teams sync only specific, relevant factual traits rather than "context stuffing" raw JSON customer histories into prompts, which Twilio says also increases reasoning accuracy.

How it works

The integration hydrates a Memory Store in one of two ways: Memory Sync for bulk profile imports, or Memory Destination, which targets specific user cohorts through Segment Audiences. When a call or chat connects, the agent queries the Memory Store through an API and retrieves only the traits needed for the task, in under 300ms, per Twilio.

Twilio's example payload includes fields such as first name, application status, missing documents and lead score, plus details of the last interaction. A sample Python function then turns those traits into a short system prompt for an AI voice agent, for instance telling it to validate known facts, skip discovery and offer to text a secure upload link, instead of passing pages of unformatted JSON to the model.

Use cases and customer examples

Twilio describes several scenarios. In warm start outbound calls, Segment traits such as budgets, intent and drop-off blockers are loaded into Conversation Memory before the call connects. Twilio says this cuts 30+ seconds of handle time per call and drives 1.6x faster lead conversions.

In live agent assist, traits reach Twilio Flex agents or AI co-pilots in under 300 milliseconds. Twilio says this removes 20 to 40 seconds of tab-switching and "dead air" and can power dynamic prompts, compliance guidance and next-best actions.

Two customers are named. CarFinance247 built an AI assistant called Carla that has handled approximately 288,000 customer conversations. According to Twilio, customers who engage with Carla convert to an approved lead roughly 1.6x faster than those who do not, producing a multi-million pound annual revenue uplift. Universidad UK deployed an autonomous outbound system called Jurgen that reads Conversation Memory enriched with Segment traits to run follow-up sequences across WhatsApp and voice calls for about 15% of leads.

What it means for Segment and Twilio users

For teams already invested in Segment, the release offers a way to reuse existing profile work in live channels without building a separate integration. Developers can treat Conversation Memory as a destination and configure it from within Segment. The performance and cost figures, as well as the customer results, come from Twilio itself and have not been independently verified.