We answer your customers in their own language and regional variety, in the tone that fits your brand, from your own procedures, around the clock. Every chat is measured on the service levels you already manage: FRT, AHT, FCR, CSAT and quality.
Five example chats. On the left is what the customer sees. On the right is what happens behind each reply. Click a step to see the message it shaped.
Compared with a human advisor, a typical chat costs a few cents of AI processing instead of minutes of paid handling time.
The agent responds to how the customer feels first, in their own dialect, then fixes the problem. It apologises once, takes ownership and never sounds like a script.
Every reply records which part of your knowledge base it came from. If an answer is ever wrong, you see exactly which section caused it.
You decide when a person takes over: refunds above a limit, legal threats, an upset customer after two attempts. The agent follows these rules on every chat.
FRT, ART, AHT, SLA, CSAT, DSAT, FCR and quality by day and by 30-minute interval, against recommended targets for 18 industries, from food delivery and banking to IT help desks and software.
It never pretends to be a person. If a customer sincerely asks, it says it is your AI assistant and offers a colleague. It never invents prices, policies or compensation.
What the agent does on its own, and what the platform does around it.
Writes each market's own variety: American or British spelling, Mexican or Spain Spanish, Brazilian or European Portuguese, Québec French, Traditional or Simplified Chinese, and the polite form each language expects (vous, usted, Sie, keigo). In Arabic it speaks ten country dialects and reads Arabizi.
Tracks how the customer feels on every message (calm, worried, confused, frustrated, angry, happy) and adapts. Reports show how many upset customers left calm.
Standard, friendly, professional, or an adaptive mix. The customer's mood always comes first: with an angry customer even the friendly tone turns calm and sincere, with no jokes, exclamation marks or emojis.
Your files are split into one section per topic. The agent works out the customer's exact situation and answers from that section only, so a card refund time is never given for a wallet refund.
If a question could match two topics with different answers, it asks one short question first instead of guessing.
When your knowledge base doesn't cover a question, or two sections contradict each other, it says it will confirm and hands over. It does not fill gaps with general knowledge.
Duplicated and contradicting content in your knowledge base is flagged automatically, with both versions side by side and the differing figures highlighted.
Questions the agent couldn't answer are grouped into missing topics, ranked by impact, with a draft article ready for you to approve.
Password resets with identity checks, email and VPN troubleshooting, access requests, ticket status and outage handling, with engineers brought in by priority and response time.
For clinics, salons, restaurants and service centres it checks the live calendar, offers real free times, confirms the details and books, moves or cancels.
Every finished chat is evaluated against your weighted scorecard, including customer, business and compliance critical items, with a coaching note.
Shows which scorecard behaviours move CSAT and FCR the most, and which contact reasons drive dissatisfaction, so you fix the right thing first.
Every chat is checked for wording, dialect and understanding mistakes while it's still going. Each mistake becomes a lesson the agent applies from the next message, and you can turn any lesson off.
Each answer is checked against the knowledge base section it used. If an amount, time or policy was wrong, the agent sends one short correction straight away and the chat is flagged for you.
Upload an order list or connect your order system, and the agent tells customers where their order is, after confirming the order is theirs with the phone number on it.
Answers on WhatsApp, your website, Facebook Messenger, Instagram, email and phone calls, with the same knowledge, rules and lessons everywhere.
Voice notes are turned into text and answered. Photos of a damaged item or an error screen are understood, and the originals are kept for your team.
Before a reply that uses your order or account data is sent, a second check compares every status, amount and date with what your system returned, and corrects it if needed.
Cancels orders, changes delivery addresses, requests refunds within your policy and opens tickets in Shopify, WooCommerce, Zendesk, Freshdesk or your own system. It always tells the customer what will happen and waits for a clear yes first.
Your agents take hand-overs in the Agent Desk with an AI summary, suggested replies and order lookup, or in your own help desk through webhooks and our API. Nobody asks the customer to repeat themselves.
Sends approved WhatsApp updates, like a delayed order, to customers before they contact you. If they reply, the agent knows what was sent.
Reads your FAQ, delivery, returns and contact pages and drafts a knowledge base for you to check, so a pilot can start the same day.
Forecasts hand-overs by half hour and works out how many people to schedule with Erlang C, and writes a monthly management report in English and Arabic.
Two-step login for admins, keys to your systems encrypted, an audit log of every change and nightly backups. Card numbers, bank account numbers and ID numbers are masked before chats are stored, conversations are deleted after the period you choose, and the agent never asks for PINs or passwords.
Pay per conversation, or only for chats the AI resolves without a hand-over. With the CSAT guarantee, part of the fee comes back if satisfaction misses the agreed level.
The Quality Monitor scores every chat, email and call from any system against your scorecard, written in plain words. Every fail quotes the exact words as evidence, scores are checked against CSAT, and people review only what's uncertain.
Your refund, cancellation and identity rules are tested and approved by you, then checked by our system before every action. Every language and channel gets the same decision.
Tags every chat with a contact reason, reusing your existing categories so trends stay comparable month to month.
| Service | What it does |
|---|---|
| AI Chat Agent | Handles customer chats on your website and WhatsApp in any language and regional variety, from your knowledge base, with booking, hand-over and full reporting.Includes the console, live wallboard, reports, quality scoring, knowledge base health and Excel exports. |
| CX Insights | Monthly subscription on top of the agent: CX drivers, quality-to-CX analysis, missing-topic scans with draft articles, and full CX analysis reports. |
| Agent Desk | One screen where your people answer chats from every channel, with queues, skills routing, live team performance and AI help. Works with the AI agent or on its own. |
| Quality Monitor | Scores 100% of your human team's chats, emails and calls from any system against your scorecard, with the evidence for every finding. |
No add-on fees for the basics an operation needs.
Time from the customer's first message to the first reply.
Average time to reply to each customer message.
First customer message to the end of the chat, finished chats only.
% of chats answered within the first response target.
% of ratings that are 4 or 5 out of 5.
% of ratings that are 1 out of 5.
% resolved with no hand-over, no drop and no repeat contact within 7 days.
% of chats resolved by the agent.
% of chats passed to a person, with the reason.
Weighted scorecard result. A failed critical item scores the chat 0%.
% of upset customers who end the chat calm or happy.
% of replies backed by a knowledge base section.
The 72-hour demo opens the full console: chat, live wallboard, quality, reports, CX insights and knowledge base health. Upload your own procedures and the agent answers from them straight away.