UA EN
SUDUS·TEAM web studio · Ukraine Closed
Call Telegram Get a quote
AI Agents · Customer Support

AI agent for
customer support

A first line that never sleeps and never hands in its notice. The agent answers repetitive questions around the clock — from your own knowledge base, not from whatever the model imagines. Anything complex goes to a person on your team together with the conversation context. Launch in 14–25 days, fixed price in the contract.

24/7
Answers with no weekends and no overnight gaps
14–25days
From brief to launch
900$
Starter package, price fixed in the contract
When you need it / 01

When support
becomes a bottleneck

There is one tell-tale sign: enquiries grow faster than your team, and most of them are the same twenty questions. If you recognise at least three of the points below, the agent will pay for itself.

The same questions every day
"How much is it?", "Is it in stock?", "How do I pay?", "Where is my order?" Someone on your team spends the whole day pasting the same answers.
Nights and weekends go quiet
A customer writes at 10 pm and gets a reply the next morning — by which time they have gone to a competitor. In markets where people decide on the spot, that is revenue lost outright.
Enquiries in five different places
Website, Instagram, WhatsApp, Messenger, email. Something always slips through, because nobody keeps every tab open at once.
Seasonal peaks
In high season enquiries triple, and you either hire people for two months or live with the queue and the bad reviews that come with it.
New starters take ages to get up to speed
Every new hire spends a month asking colleagues the same things. If that knowledge has to be gathered in one place anyway, the agent might as well use it too.
Your sales team has no time to sell
Your most expensive people spend half the day working as an autoresponder instead of closing the customers who are already ready to buy.
What the agent does / 02

Not a "smart chat",
but real work

01
Answers questions properly
Reads the question however it is phrased, finds the answer in your knowledge base and writes it in plain, human language. Copes with follow-ups and typos.
02
Checks order status
Connects to your CRM, online store or shipping provider and answers "where is my order?" with real data, not a stock reply.
03
Hands over with context
When a question goes beyond its remit, it does not cut the conversation off — it passes it to a person with a short summary: who, what about, and what has already been established.
04
Qualifies enquiries
Tells a ready buyer from someone who is just browsing, and both from a complaint. Buyers are passed on straight away; complaints are flagged as a priority.
05
Creates a record in your CRM
Collects the name, contact details and the gist of the request and puts a record into your system — no copying things over from the chat by hand.
06
Works in several languages
Replies in the language the customer wrote in. For businesses selling across borders, that removes a whole line of costs.
07
Shows you what is going on
A dashboard with conversations, ratings and errors. You can see which questions the agent struggled with — and what to add to the knowledge base.
08
Stays silent where it should
Guardrails: off-limits topics, and no promises of discounts, deadlines or compensation. Better "let me get a colleague" than a promise you then have to keep.
How it knows the answers / 03

Four steps
from question to answer

The main reason rollouts fail is that the agent goes live without a knowledge base and starts making things up. That is why we spend most of the time on the first step.

Step 01
We build the knowledge base
Manuals, the price list, shipping and returns terms, warranty, FAQs, the history of real customer conversations. What already exists as files we take as it is; what lives in someone's head has to be written down first — and that is a normal stage of the work, not a delay.
Step 02
We connect RAG
The agent does not rely on the model's general knowledge. For every question it first searches your documents and answers only on that basis. If nothing turns up in the knowledge base, you can see it — and the agent does not improvise.
Step 03
We set boundaries and a persona
We define the tone of voice, off-limits topics and what the agent is not allowed to promise. This is also where the handover rule is set: under what conditions a conversation goes to a person, and how that person receives the context.
Step 04
We launch and read the conversations
For the first weeks you see every conversation in the dashboard. Wherever the agent struggled, we add to the knowledge base. This is not "fixes after delivery" but part of the job — for the first two months after launch we watch closely and keep improving it.
Where it works / 04

One agent,
every channel

One knowledge base, many ways in. Customers do not need to know where to write — the answer is the same everywhere.

Website widget
WhatsApp
Facebook Messenger
Instagram Direct
Telegram
Email
Inside your own system · via API
What it costs / 05

The price depends
on complexity, not buzzwords

PackageWhat is includedFrom
AI STARTER One agent for support: RAG knowledge base of up to 100 pages, integration with 1–2 external systems, basic usage dashboard. Timeline 14–25 days. $900
AI PROFESSIONAL Multi-agent orchestration, extended RAG of up to 1,000 pages, MCP integrations with CRM, ERP and accounting systems, an evaluation harness on your real cases, full analytics dashboard. Timeline 35–60 days. $3,000

Model usage is billed separately — typically $40 to $400 per month, depending on how many conversations the agent handles, which is why the brief asks about your expected volume. Every package comes with a 6-month warranty. The full package comparison is on the AI agents page, and all other services are in the price list.

Limits / 06

What the agent
does not do

This list matters just as much as the previous one. Better to hear "no" now than in week three.

It does not replace your team on complex, disputed or emotional cases.
It does not make up answers: if the knowledge base has nothing, it calls a person.
It does not promise discounts, deadlines or compensation without your approval.
It does not take payments, and it does not complete refunds or returns without a person signing off.
It does not give legal, medical or financial advice.
It does not launch in a week if the knowledge base does not exist yet.
Questions / 07

What we are asked
most often

How is an AI agent different from an ordinary chatbot? +
A chatbot follows a script: button, answer. One step off the path and it is lost. An AI agent reads the question the way a person actually wrote it, looks for the answer in your knowledge base and puts it into its own words. It copes with follow-up questions, rephrasing and typos. If there is no answer, it does not make one up — it hands the conversation to someone on your team, together with the context.
Where does the agent get its answers from? +
From your knowledge base: manuals, the price list, shipping and returns terms, FAQs, the history of past conversations. We turn this into a structured knowledge base and connect it through RAG — the agent searches your documents and answers only on the basis of what it finds there. This is the main reason some rollouts work and others produce made-up answers: without a knowledge base, the agent has nothing to stand on.
What if it tells a customer something wrong? +
There are three safeguards. First, the agent answers only from your knowledge base, not from the model's general knowledge. Second, guardrails: off-limits topics, and no promises of discounts, deadlines or compensation without your sign-off. Third, a dashboard where you can see every conversation and its rating: for the first weeks you read them and we adjust the agent. We do not put an agent in front of your customers without this stage.
How long does it take to launch? +
STARTER takes 14–25 days. Most of that time goes not on code but on putting the knowledge base together: if your manuals and FAQs already exist as files, it is quick; if everything lives in one person's head, it has to be written down first. That is why at the first meeting we ask about your documents, not about design.
Will it replace our support team? +
No, and we do not promise that. It takes over the first line — the repetitive questions that come up every day, and the messages that arrive overnight. Complex, disputed and emotional cases stay with people, but now with the context already prepared: who the customer is, what they asked, what the agent has already found out. The real effect is that your team stops working as an autoresponder, not that it disappears.
What happens when we change our prices or terms? +
You update the document in the knowledge base, and the agent starts answering the new way. No model retraining is needed. That is the fundamental difference from bots built on rigid scripts, where every change in your terms means rewriting the dialogue tree.
What does it cost to run after launch? +
There are two parts: our support and the model itself. The first two months after launch, when we read real conversations and keep improving the knowledge base, are part of the work, and every package comes with a 6-month warranty on what we deliver. Ongoing support after that is agreed separately, depending on how many changes you plan. Model usage depends on the number of conversations and is billed separately — typically $40 to $400 per month. That is exactly why the brief asks about your expected volume.
Other jobs for AI / 08

Support is
just one scenario

Voice agent for call centres →
When customers call instead of typing. It picks up, answers by voice and transcribes the call.
All AI agents →
Types of agents, how a production agent is built, 12 use cases, and a full comparison of packages and timelines.
AI agent for sales →
When the customer is still choosing: it qualifies the lead, finds out budget and timeline, and passes it to your sales team.

Want a quote for an agent
for your support?

Send a request → Contact us directly
Request

Let's price an agent for your support

In the message, tell us roughly how many enquiries you get a month, which channels they come through, and whether you have manuals and FAQs as files. That is enough for an estimate in numbers. The first call is 30 minutes, free.

  • We reply within 2 working hours
  • Fixed price and timeline in the contract
  • We say no to anything we cannot deliver
Not a fan of forms? (066) 650-43-33
Telegram @sudus_team · info@sudus.team