Intelligent Virtual Assistant (IVA)

The term comes from contact-center software, where it distinguished a system that understood free-form requests from a chatbot that matched keywords or an IVR that took keypresses. In practice today an IVA is the customer-facing agent that handles tier-one requests end to end on the phone and in chat, pulls from the company's systems to answer, and hands the rest to a person. The label is older than "AI agent" but describes the same role.

How it works. The IVA takes the caller's or chatter's utterance, runs natural language understanding to identify intent and extract entities (an order number, a date, an address), and decides whether it can complete the request itself. If it can, it calls the relevant system: an order status lookup, a calendar, a CRM record. If it needs more information, it asks one question at a time and keeps the context across turns. If it cannot resolve the request, it transfers with the gathered details. Modern IVAs use a language model for the understanding and dialogue steps rather than hand-built intent classifiers, which is why they cope with phrasing they have not seen before.

Why it matters for support and sales. The business case is containment without frustration. A menu-based system contains calls by forcing them into a few buckets and sends everything else to a queue. An IVA contains them by understanding them, which means the contained share grows as you add knowledge and tools, and the escalations that remain arrive with context. The cost is governance: the assistant must be grounded on approved content and bounded by rules, or its flexibility becomes a liability.

Example. A logistics company's assistant takes a call: "Hi, I've moved and my parcel for tomorrow needs to go to the new place." The assistant recognizes two intents, an address change and a delivery reschedule, asks for the tracking number, confirms the new address line by line, checks the carrier cutoff through an integration and updates the delivery. A menu system would have offered "press 2 for delivery changes" and failed on the compound request. See automating logistics delivery updates.

IVA vs chatbot. A chatbot in the narrow sense follows a scripted tree or matches keywords; it is predictable and cheap but brittle. An IVA understands intent and context and can act in systems. The line has blurred since language models arrived, so judge a product by what it does on a compound, badly phrased request rather than by the label on the box.

What to measure.

Botcadence agents handle compound intents, transfer with full context and run across phone, website chat, WhatsApp and other channels from one platform; see voice agents and conversational AI.