One support agent,
every messenger your customers already use
Customers do not wait for business hours anymore, and they do not want to repeat themselves across three different apps before someone helps them. This agent reads your documentation, order and account data, and answers in the channel the customer picked. It hands off with full context the moment a question needs a person.
Customers do not wait for business hours
Most small and mid-size teams answer customers through whatever app the customer happened to pick. The same question gets typed into Telegram, WhatsApp and email by three different people, at three different speeds. Nights and weekends go unanswered until someone opens a laptop. The person answering often has to dig through order history or a policy document mid-conversation while the customer waits. None of this is anyone’s fault. It is just what happens when support runs on whoever is online.
What the agent answers
The agent reads your product documentation, return policy, shipping rules and FAQ, and answers from that material rather than guessing. It works in Telegram, WhatsApp, LINE and Instagram Direct, matching the channel and language the customer used. For anything account-specific, order status or subscription state or a past ticket, it looks up the answer in your CRM, helpdesk or order system. The customer never has to repeat details they already gave once.
It remembers the thread. A customer who messaged yesterday about a delayed order and writes back today does not have to explain the situation again. Each channel runs under its own reply-rate limits, respecting each platform’s own rules and published policies. Some questions need real judgment: a refund outside policy, a complaint, a request the knowledge base does not cover. For those, the agent stops, summarizes what it found, and hands the conversation to a person with that summary attached, instead of making something up or going quiet.
Over time, the agent’s answer patterns become data your team can act on. A spike in questions about one shipping policy or one product defect shows up as a cluster rather than scattered tickets. That is usually the first real signal a policy page needs rewriting, or a product issue needs escalating beyond support. The same memory also helps review: a manager checking a flagged conversation sees the entire relationship, not just the message that triggered the review.
What your team still owns
Anything involving money outside standard policy, a customer who is upset, or a question the knowledge base genuinely does not answer goes to a person. Your team owns the knowledge base itself: what is accurate, what is out of date. Your team also reviews a sample of agent answers each week, to catch drift before it becomes a pattern.
How every answer stays traceable
Every conversation is logged with the sources the agent used to answer, so a manager can see exactly why it said what it said. Reply limits per channel are hard caps, not suggestions, to protect your accounts from platform bans. A new knowledge base or policy change runs in a review mode against recent real conversations before it goes live. A single switch takes the agent fully offline if something looks wrong. The agent introduces itself as an AI assistant at the start of every conversation.
Price and timeline
| Option | Price | Timeline |
|---|---|---|
| Agency runs it | from $2,500 + support plan | 3 to 5 weeks |
| Full control, handover-ready | from $3,500 | 5 to 7 weeks |
“Agency runs it” keeps the agent on our infrastructure with a monthly support plan covering tuning and monitoring. “Full control, handover-ready” delivers the agent on your own servers and accounts, with documentation, source and credentials, so your team can run and change it without us. It costs more up front because the handover package, the deployment on your infrastructure and the internal documentation are built in from day one.
Related
See the AI agents overview for how we build and guard these systems, and Routine takeover for the services that often pair with this one. Inside the agents catalogue: FAQ and knowledge base agent, Helpdesk triage agent, Multilingual support agent. For a narrower, single-process version of this work, see Zendesk and Intercom ticket agent or Support ticket triage. For how the same guard patterns held up on real systems: AI sales agent for seven channels.
Want this running for your team? Get in touch and tell us where the messages pile up.
FAQ
How much does a customer support AI agent cost?
From $2,500 for one knowledge base and up to four messenger channels, live in 3 to 5 weeks. Cost rises with the number of systems it needs to read from, like order status or subscription data.
How long before it is live?
3 to 5 weeks for the first channel, once we have your documentation and a sample of real past conversations to test against.
Which channels and tools does it work with?
Telegram, WhatsApp, LINE, Instagram Direct and your website chat widget on the customer side. Zendesk, Intercom, your CRM or a plain spreadsheet on the system side, whichever you already run.
What if the agent gets an answer wrong?
It can only answer from the knowledge base and data you give it. Every factual claim is checked against your systems rather than guessed. Anything outside its confidence or scope gets handed to a person with the conversation attached, instead of answered anyway.
Is customer data safe?
Conversations and account data stay in your own messenger and CRM accounts. The agent reads and writes through your existing API keys, and every exchange is logged so you can audit it.