Capture signals
AI detects preferences, intent, usual services, and important changes during calls or chats.
Kelvia separates preferences, history, and consent from unfinished work. Once identity is verified, voice, WhatsApp, webchat, and email can continue with the right context.
Sofia Martin
Recognized profile
Called to move a booking
voice - 3 days ago
Confirmed the new time
WhatsApp - 2 days ago
Accepted reminders
consent - active
Offer an afternoon follow-up check-up and confirm it over WhatsApp.
Kelvia does not store noise. It structures facts, events, and consent that can improve the next interaction.
AI detects preferences, intent, usual services, and important changes during calls or chats.
Data becomes memory facts, events, and a customer profile tied to the right tenant.
The next conversation starts with context: favorite channel, recent actions, and recommendations.
Kelvia separates customer preferences from work state: what they were trying to do, what is still missing, and whether the task is active, paused, or closed.
The customer begins a quote, order, booking, or support request and leaves some information for later.
They can leave that task unfinished, discuss another topic, and keep each piece of work separate.
In a later verified interaction, Kelvia restores the summary and continues from the missing detail.
A task is explicitly completed or cancelled; temporarily changing topic does not close it.
Resuming across sessions or channels requires a verified customer binding with customer-profile access. Voice, WhatsApp, webchat, or email can continue the task only after resolving that person.
It can keep isolated context while the call is active, but it does not create or retrieve persistent task memory across sessions.
If two quotes, bookings, or support cases could match, Kelvia presents the choices and asks the customer to select one; it does not guess which to resume.
Detailed context remains for the configured continuity window. Afterwards, sensitive notes and free-text fields are redacted, leaving only the minimum needed for safe recovery.
Customer Memory is designed to be useful without becoming a black box.
Context stays inside the business tenant.
Private customer actions are tied to a temporary verification, not only to a phone number typed into chat.
Channels and purposes respect customer permissions.
Relevant memory changes stay traceable for the team.
Customer Memory turns scattered conversations into a continuous relationship.