Agent One x UCD: an AI Agent that Knows Your Customer

Most AI agents treat every customer like a stranger – your AI agent greets a logged-in Gold-tier member by asking for their name. It recommends a brand the customer has bought from for years, as if for the first time. It has no idea which campaign, product page, or offer brought the person into the conversation.

This is context blindness, and it’s the default state of almost every conversational AI platform on the market.

The cost runs both ways. The agent doesn’t know the customer, so the experience feels generic. And the rich signals every conversation generates, like intent, product interest, and satisfaction, simply vanish the moment the chat ends. Marketing teams lose data they should own, and customers repeat themselves to a system that should already know them.

Agent One™ closes that gap from both directions.

Agent One is now part of your customer database, not just connected to it

With this release, Agent One becomes a full participant in the Insider One ecosystem. It does two things that no point-solution chatbot can do natively:

It feeds the Unified Customer Database (UCD) with behavioral data from every conversation. 

And it reads from UCD to personalize each conversation in real time.

That bi-directional flow is the difference between a chatbot that integrates with your customer data and an agent that is genuinely part of it. Integrations break. Architecture doesn’t.

Feeding UCD: every conversation becomes a data asset

The first part of this release is the data Agent One sends back. Agent One now sends nine structured events directly into the Unified Customer Database, available for segmentation and journey orchestration across every Insider One channel:

• Conversation Lifecycle: conversation started, conversation ended, handover to live agent
• Feedback: conversation rating (CSAT / NPS)
• AI and Knowledge Base: action triggered, knowledge base response, recommendation shown, recommendation clicked
• Live Agent Operations: tag assignment

Agent One is now an official UCD feeding source. On WhatsApp, profiles are matched or created automatically via phone number (set as an identifier). On Web Chat, events flow once the Insider One Tag is configured on the site.

The practical result: conversational behavior becomes an audience. “Users who clicked a product recommendation in chat” or “users who rated their last conversation below three” stop being invisible session logs and become actionable segments you can target through email, push, and onsite.

Reading UCD: personalization without prompt engineering

The second part of this release is Dynamic Instructions, a new field on the Agent personality page. In simple terms, instead of writing one static script for the agent, you can pull in customer profile attributes that are automatically populated when a conversation starts.

Think of it like mail merge for conversations. You write the instruction once, leaving placeholders for details such as a customer’s name or loyalty tier, and Agent One fills them in automatically. So rather than a generic “How can I help you today?”, a Gold-tier member is greeted as a Gold-tier member—without requiring separate instructions for every customer.

The third part of this launch is Audience Instructions: segment-based instruction sets that let you tailor the agent’s behavior for different audiences, such as VIP members, cart abandoners, or brand loyalists. Instead of personalizing for each individual customer, you’re personalizing for entire customer segments—all from a single configuration. If a profile attribute is unavailable, fallback values ensure conversations still feel natural.

Audience Instructions and Dynamic Instructions are available across both Web Chat and WhatsApp (WhatsApp requires a phone number as the identifier).

The result is personalization at both the customer and segment level, with no prompt-engineering overhead.

What this looks like in practice

VIP and loyalty recognition 

A Gold-tier member opens the chat. The agent reads their loyalty tier from UCD and opens with a personalized acknowledgment, no name-asking friction, no generic greeting.

Brand-loyalist recommendations 

A customer with a history of Samsung purchases asks for a phone recommendation. The agent leads with the relevant lineup rather than first asking which brand they prefer.

Purchase-eligible upsell 

A customer flagged as upgrade-eligible in UCD starts a conversation. The agent surfaces trade-in offers and current promotions early, with mild urgency and no pressure.

Segmentation from conversational events 

A user clicks a product recommendation in chat. The event fires to UCD, and the CRM team builds a segment of customers who engaged with chat recommendations, then targets them with a follow-up email journey.

CSAT-based recovery journeys 

A user rates a conversation one out of five. The rating event fires to UCD, and an automated Insider One journey triggers a recovery communication within 24 hours.

Why is this different from every other AI agent

Most conversational AI platforms are closed systems. They don’t write back to a customer data platform, and they don’t read from one. Personalization, where it exists, is limited to session context or a handful of static variables, and event data stays locked inside the chatbot tool.

Agent One is the only AI agent natively embedded in a full-stack customer data platform and cross-channel engagement suite. It enriches the customer profile with every interaction, and those enriched profiles feed personalization on every other channel.

What sets Agent One apart:

• No middleware or third-party connectors required
• Bi-directional data flow (read and write), versus competitors’ one-way or no-flow model
• Segment-scoped instructions (Audience Instructions), versus a single static system prompt
• Real-time attribute resolution at conversation start, versus session-level memory only
• Nine structured events with a defined schema, versus unstructured logs

Every conversation teaches the platform. Every profile makes the conversation smarter.

Who this is built for

This release is built for digital and CRM leaders at mid-to-enterprise retail, telco, lifestyle, e-commerce, and financial services brands already invested in Insider One’s CDP and segmentation, and who want to close the loop between Agentic AI and their broader customer data strategy.

Ready to get started?

UCD x Agent One is available now within the Insider One platform. 

👉 Talk to your account team to enable UCD events and configure your first Audience Instructions. 

👉 Request a personalized demo of Insider One today to see Agent One and UCD working together.

Frequently Asked Questions

What is the UCD x Agent One integration? 

It is a native, bi-directional connection between Insider One’s Agent One and the Unified Customer Database. Agent One feeds conversational behavioral data into UCD and reads customer profile data from UCD to personalize conversations in real time.

What data does Agent One send to UCD? 

Nine structured events: conversation started, conversation ended, handover to live agent, conversation rating, action triggered, knowledge base response, recommendation shown, recommendation clicked, and tag assignment.

What are Dynamic and Audience Instructions? 

Dynamic Instructions let you inject UCD profile attributes into an agent’s instructions using token syntax, resolved in real time at conversation start. Audience Instructions are segment-scoped instruction sets, so different audiences receive different agent behavior, with up to five traits per agent and fallback values when an attribute is unavailable.

Which channels are supported? 

Web Chat and WhatsApp. WhatsApp requires a phone number set as an identifier; Web Chat requires the Insider One Tag configured on the website.

Do I need custom integration or middleware? 

No. Agent One feeds and reads from UCD natively within the Insider One platform, with no third-party connectors required.