Conversive Intelligence

What it does

Intent Recognition determines the goal behind a customer’s message — not just the words they used. It classifies incoming text or speech into structured intents (for example, “book appointment,” “check order status,” or “request a refund”) so the platform can route the conversation to the right automation, agent, or journey step.

How it works

The engine runs a fine-tuned natural language understanding (NLU) model against every inbound message in real time. It evaluates the full message alongside recent conversation turns to resolve ambiguity — so a follow-up message like “actually, make it tomorrow” is understood in context rather than treated as a standalone utterance. Confidence scores are attached to each classification; low-confidence results can be configured to trigger a clarifying question or a human handoff rather than an incorrect action.

Example

A patient texts a healthcare provider: “I need to move my appointment.” Intent Recognition classifies this as reschedule_appointment with high confidence. The platform immediately launches the rescheduling flow — presenting available slots — without the patient needing to navigate a menu or speak to a receptionist.

What it does

Sentiment Monitoring tracks the emotional tone of a conversation turn by turn. It detects shifts toward frustration, urgency, confusion, or satisfaction and uses those signals to adapt the conversation — changing tone, offering a different response path, or escalating to a human agent before the customer disengages.

How it works

Each message is scored on a continuous sentiment scale (negative → neutral → positive) using a model trained on conversational data across multiple channels. The platform maintains a rolling sentiment window across the conversation, so a single terse reply does not immediately trigger escalation — but a sustained downward trend does. Thresholds are configurable per journey or automation: a support flow might escalate after two consecutive negative turns, while a sales flow might simply soften its language.

Example

A customer contacts a telecoms provider about a billing dispute. After two exchanges, their messages become shorter and more clipped: “This is ridiculous” and “I’ve already explained this.” Sentiment Monitoring detects the negative trend and automatically escalates the conversation to a senior billing agent — passing the full conversation history so the agent can pick up without asking the customer to repeat themselves.

What it does

Contextual Memory is the persistent store of everything Conversive knows about a customer: every conversation they have had, every preference they have expressed, every outcome that was reached, and every piece of data collected along the way. It makes that history available to every subsequent interaction — across channels, agents, and time.

How it works

When a conversation ends, a structured summary is written to the customer’s memory record alongside raw transcript data, extracted entities (dates, product names, account numbers), and outcome tags. When the same customer contacts again — on any channel — the platform retrieves their record and injects the relevant context into the active session before the first response is generated. Agents and automations query this context at runtime, so they can reference past interactions without requiring the customer to re-explain their situation.

Example

A car dealership customer called three weeks ago to enquire about a specific model and was told it was out of stock. When they text the dealership today, Contextual Memory surfaces that prior conversation. The AI agent opens with: “Hi Sarah — the Civic Hybrid you asked about in March is back in stock. Would you like to book a test drive?” The customer did not need to re-introduce themselves or repeat their interest.

What it does

Adaptive Routing decides, in real time, where a conversation should go next — which channel, which automation, which human agent, or which journey branch. It replaces static decision trees with dynamic routing logic that weighs customer data, live signals, and business rules simultaneously.

How it works

At each routing decision point, the engine evaluates a set of weighted factors: the customer’s detected intent, their sentiment score, their history (for example, VIP status or previous escalations), current agent availability, business hours, and any custom rules defined in the journey. The result is a ranked list of routing targets; the platform selects the highest-ranked option that is currently available. If the top choice is unavailable (for example, the preferred agent is busy), it falls through to the next best option rather than dropping the conversation into a generic queue.

Example

A high-value insurance customer submits a claim via webchat at 8 PM. Adaptive Routing detects that the customer has a premium policy (from Contextual Memory), that their sentiment is anxious (from Sentiment Monitoring), and that live agents are offline. Rather than sending a generic “we’ll get back to you” message, it routes the customer to a specialised after-hours claims automation that collects the claim details and schedules a callback with a senior claims handler for the next morning.

What it does

The Policy & Compliance Engine enforces your business rules, messaging regulations, and consent requirements automatically at every step of a conversation. It acts as a guardrail that prevents non-compliant messages from being sent, ensures opt-out requests are honoured immediately, and applies jurisdiction-specific rules without requiring manual review.

How it works

Policies are defined as structured rules in the intelligence layer — for example, “do not send marketing messages between 9 PM and 8 AM,” “always include an opt-out instruction in the first message of a campaign,” or “require explicit consent before sending health-related content.” Before any outbound message is dispatched, the engine evaluates it against the applicable rule set for that customer, channel, and message type. If a rule is violated, the message is blocked and the journey is redirected to a compliant alternative path. Consent state is tracked per customer and per channel, and opt-out keywords (STOP, UNSUBSCRIBE, etc.) are processed in real time regardless of which automation or agent sent the original message.

Example

A US-based retailer runs an SMS campaign. A customer replies “STOP” mid-journey. The Policy & Compliance Engine immediately updates the customer’s consent record, halts all pending messages in that journey for that customer, and logs the opt-out with a timestamp — all before the next scheduled message would have been sent. No manual intervention is required, and the retailer’s A2P 10DLC compliance record remains clean.

What it does

Outcome Analytics measures whether conversations achieved their intended business result — not just whether messages were delivered. It tracks conversion rates, time-to-outcome, drop-off points, sentiment trends, and engagement scores across every journey, automation, and channel.

How it works

Each journey or automation is configured with one or more outcome events — for example, “appointment booked,” “payment received,” or “lead qualified.” As customers move through the journey, the platform records which steps they completed, where they dropped off, how long each stage took, and what their sentiment was at each point. This data is aggregated into dashboards and exportable reports. Because the intelligence layer connects all channels, analytics reflect the full customer path — including cross-channel journeys where a customer started on SMS and completed on webchat.

Example

A recruitment agency runs a candidate outreach journey. Outcome Analytics shows that 68% of candidates who receive the initial SMS respond within 4 hours, but only 31% complete the screening questionnaire. Drop-off analysis pinpoints the third question as the abandonment point. The team shortens the questionnaire, re-runs the journey, and sees completion rates rise to 54% within two weeks.