Frequently asked questions about our solutions

Inbound support, customer experience surveys, appointment and reservation flows, reminder and confirmation calls, and HR pre-screening are the most common. Any repeated, structured conversation is a candidate for an AI voice agent.

Yes, and the transfer carries context: verified identity, a one-sentence summary of the request, relevant record numbers and the steps already tried. Without that package the customer explains everything twice.

The largest gain usually comes from a lexicon of organisation-specific terms: product names, campaign names, branch names and reference codes. Dual-channel call recording, which separates agent and customer, is the second.

Every call produces a transcript, a summary, topic labels and quality signals. These can be written back to your CRM, quality management tool or BI layer over API.

Yes. The AI call assistant checks real-time availability, proposes a slot, confirms details by reading them back, and can reschedule in the same call, which is where most of the value in no-show reduction sits.

Yes. The voicebot sits on your telephony layer rather than replacing it, so existing queues, recording and reporting continue to work.

Web, mobile apps, WhatsApp and internal portals. Channel rules differ, so the flow is adapted per channel while the answers come from one knowledge base.

The customer does not repeat themselves when they switch channel. Verified identity, the open transaction and a short summary travel with them, while verification status deliberately does not.

It classifies the request, completes missing information, searches the knowledge base and similar past tickets, closes what it can and routes the rest to the right queue with a clean summary.

Every document gets an owner, a version and a validity date, and expired content leaves the index. Questions the assistant could not answer become the content team’s work list.

Yes. The model, index and integrations can run entirely inside your network, with identity from LDAP or Active Directory and permission filtering applied before the search query runs.

When retrieval confidence falls below the threshold you set, when the answer is not supported by a source, or when the user asks. The threshold is a business decision, not a technical default.

They add analysis, search and moderation capability to products you already run, without changing your interface. You call them over REST and write the outputs wherever you need them.

Your documents, products or records are indexed with both lexical and vector search, and the permission filter is applied before the query. Your application then calls the search endpoint instead of its current one.

Time-stamped transcripts with speaker separation, summaries, topic labels, process compliance checks and risk signals, ready to be written into quality, CRM or BI systems.

Free-text reviews and messages are grouped by theme and scored for sentiment, urgency and product area, so specific labels can be routed automatically to the responsible team’s queue.

Communities, live support, social channels and in-app chat. Detection is contextual rather than list-based, and actions are graded per category rather than a single block-or-allow threshold.

An API key, the endpoint definitions and a decision on where the data will be processed. A test key and a pilot plan are usually enough to start.