How Agents Work Together

​

Overview

Nexudus AI is powered by two complementary systems that work together to deliver a complete communication experience:

• Conversational AI — responds to customers in real time across chat, email, and voice.
• Proactive Agents — monitor your space for patterns and opportunities, then propose outbound actions for your review.

These two systems are not separate silos. They share the same intelligence, the same tools, and the same conversation history. When a proactive agent sends an email or places a voice call, the customer’s reply flows back into a conversational session — creating a seamless loop between proactive outreach and two-way dialogue. This page explains how these systems connect, how handoffs work, and what the full journey looks like from detection to resolution.

---

​

The Two Systems at a Glance

---

​

How They Connect

The diagram below shows the complete flow from pattern detection to customer conversation:

---

​

The Proactive-to-Conversational Journey

Here’s what happens, step by step, when a proactive agent identifies an opportunity and the customer responds:

​

Step 1: Detection

Proactive agents run automatically in the background, scanning your location’s data for patterns you’ve configured. For example:

• An invoice has been overdue for 5 days
• A contract expires in 30 days and hasn’t been renewed
• A lead inquired about workspace last week but never followed up
• A new member hasn’t made their first booking yet

Each agent has its own criteria and thresholds that determine when to act. See Agent Types for what each agent watches for.

​

Step 2: Message Composition

When a pattern is detected, the AI composes a personalized draft message using:

• Your location’s tone and personality — the same voice used in conversational AI
• Customer context — their name, history, activity patterns, and relevant data (like invoice amounts or contract dates)
• Channel-appropriate formatting — email messages include subject lines and structured content; voice scripts are conversational and concise

The draft is tailored to the specific situation — not a generic template. A payment reminder for a member with perfect payment history will read differently from one for a member who frequently pays late.

​

Step 3: Your Review

The proposed action appears in your AI Inbox where you can:

• Approve — send the message exactly as drafted
• Edit & Approve — personalize the message before sending
• Reject — dismiss the action if it’s not appropriate right now
• Snooze — defer it for later review

If you’ve enabled auto-execute for a specific agent, high-confidence actions skip the inbox and send automatically.

​

Step 4: Delivery

Once approved, the message is delivered through the configured channel:

​

Step 5: The Conversation Begins

This is where the handoff happens. The moment the customer replies (to an email) or speaks back (on a voice call), they’re in a conversational AI session — the same system that powers the Members Portal chat. The key difference: this conversation was initiated by the proactive agent, not the customer. But from this point on, it behaves exactly like any other AI conversation:

• The AI has access to all the same tools (bookings, plans, resources, FAQ)
• It knows the full context of why the outreach happened
• It can answer follow-up questions, take actions, or escalate to your team
• The entire exchange is logged in AI Conversations with a link back to the original proposed action

---

​

How Replies Are Handled

When a customer responds to a proactive outreach, the reply routes back through the same channel and is handled by the conversational AI:

​

Email Replies

When a customer replies to an email sent by a proactive agent:

1. The reply arrives at a special address that encodes the conversation session.
2. The system identifies which session the reply belongs to and loads the full context — including why the outreach happened, the original message, and all prior exchanges.
3. The conversational AI processes the reply with that context and composes a response.
4. The response is sent back via email, continuing the thread naturally.

The customer sees a normal email conversation. Behind the scenes, the AI knows this started as a proactive outreach about (for example) an overdue invoice, and can address the customer’s response accordingly — whether they’re asking for a payment link, disputing the charge, or requesting more time.

​

Voice Replies

Voice works differently because the conversation happens in real time:

1. The proactive agent places an outbound call and delivers its message.
2. The customer can immediately ask questions or take actions during the same call.
3. The voice AI uses the same tools available in chat — it can check availability, create bookings, provide pricing, or look up information.
4. If the customer needs something the AI can’t handle, it offers to connect them with your team (see Escalation below).
5. Any links (like payment pages or booking confirmations) are sent via SMS during the call instead of being read aloud.

---

​

Escalation: Connecting to Your Team

Both conversational and proactive AI share the same escalation capability. When the AI reaches the limits of what it can handle — whether during a proactive outreach or a regular customer conversation — it seamlessly connects the customer with your team.

​

When Escalation Happens

The AI will offer to connect a customer with your team when:

• The customer explicitly asks to speak with a person
• The question falls outside what the AI can answer (not covered in FAQ or its tools)
• The customer is frustrated or the topic is sensitive (billing disputes, complaints)
• The customer needs something that requires human judgment

​

What Gets Transferred

When escalation occurs, a support ticket is created in your Help Desk with:

• A summary of the conversation and what the AI tried
• The full exchange history so your team doesn’t ask the customer to repeat themselves
• The channel the conversation happened on (email, voice, or chat)
• The customer’s identity and relevant context (invoices, bookings, contract details)

​

How Your Reply Gets Back to the Customer

Once a conversation is escalated, the Help Desk takes over and all further communication happens via email. This is true regardless of which channel the original conversation was on: Escalation is an intentional handoff from real-time AI to your team’s asynchronous workflow. The customer receives an email from your Help Desk with the operator’s reply — just like any other support ticket.

---

​

Real-World Examples

​

Example 1: Overdue Invoice → Payment

​

Example 2: Lead Follow-up → Booking

​

Example 3: Contract Renewal → Voice Call → Escalation

---

​

What Makes This Different from Automation Rules

You might wonder how proactive agents differ from traditional automation (like automated email sequences or reminder rules). Here are the key differences:

---

​

Shared Capabilities

Both conversational and proactive AI share the same underlying capabilities:

​

Tools

The AI can take actions on behalf of your customers during conversations — whether those conversations started from the Members Portal chat, a proactive email, or a voice call:

• Check resource availability and create bookings
• Look up plan and pricing information
• Search your FAQ and knowledge base
• Provide location details and operating hours
• Generate portal links for payments, bookings, and profiles
• Create day pass purchases

​

Knowledge Base

Both systems draw from the same knowledge sources:

• Your published FAQ articles
• Location details and operating information
• Plan and pricing configurations
• Resource availability data

​

Business Persona

Whether a customer chats in the portal or receives a proactive email, the AI uses the same tone, personality, and communication style configured for your location. This ensures a consistent experience regardless of which system initiated the conversation.

---

​

Privacy and Control

​

Key safeguards

• Opt-out respected — Customers who have opted out of communications are excluded from all proactive outreach.
• Rate limiting — Each agent has a configurable cooldown period and daily maximum. The same customer won’t receive multiple messages in rapid succession.
• Deduplication — If an agent has already proposed an action for a specific situation (e.g., this specific overdue invoice), it won’t propose it again.
• Expiry — Proposed actions expire if not reviewed within a configurable window, preventing stale outreach.
• Auto-execute is opt-in — Every agent starts in manual review mode. You choose which agents (if any) to promote to automatic sending.
• Business hours — Actions configured with business hours enforcement will only be delivered during your operating hours.

---

​

Getting Started