Business AI agent pricing
Estimated cost
We agree the quote based on the size of the knowledge base and integrations. AI model usage and hosting are charged separately; support from the second month is optional. A RAG assistant for employees
An internal assistant is useful when instructions, price lists and procedures are spread across different documents. An employee asks a question in ordinary language, and the agent finds the relevant passage and can show its source. It can, for example, help a sales manager clarify the order approval process or prepare the required documents. Only approved materials enter the knowledge base, and outdated versions are excluded. Access to internal information is role-based: a document for accounting staff must not become a general answer available to everyone. A private interface, authentication and complex permissions are quoted separately according to the work processes involved.
An AI assistant for a website with several services
Website visitors often know what they need to achieve but do not know which service to choose. The AI assistant asks for details and finds a description of the relevant service in company materials. It can, for example, explain the difference between equipment repairs and scheduled maintenance, and show the terms and required documents. If a specialist is needed to prepare a quote, the agent collects the agreed information and passes the enquiry to a staff member. Customers do not have to search several pages for the answer, and the sales manager continues the conversation with its context. The base project can start with one group of services.
An AI agent for customer support
Existing customers ask about setup, documents, payments and the service process. The AI agent finds an appropriate instruction in the approved knowledge base and explains the next step in response to the question. It can, for example, help users find the right section of a manual or prepare information for a support request. Disputed cases and questions without sufficient source material are passed to a staff member with the conversation history. After launch, the conversations reveal which explanations are missing from the documentation. Specific account data is available only through a separately connected system that checks user permissions.
AI product recommendations based on specifications
When a catalogue contains similar products, buyers find it difficult to choose by name and price alone. The AI agent asks about the intended use, important specifications and limitations, then compares suitable items using the agreed data. It can, for example, explain equipment configurations or consumable compatibility. Current prices and availability can be retrieved through a separate catalogue integration. If there is no exact match, the request is passed to a sales manager. Before development, we define the required parameters and comparison rules. AI helps customers choose; the final order follows the store's usual process.
An AI agent for checking order status
Customers can ask whether an order has been accepted, a repair is complete or a delivery has reached a particular stage. For this use case, the AI agent connects to the business system and checks whether the user is authorised to see the information. The answer uses the current status, while requests to change an order or resolve a dispute are passed to the responsible staff member. Without an integration, the agent can explain the general process but cannot report the status of a specific order. API integration, authentication and access rules are quoted separately. We agree the data fields with the company before launch.
How an AI agent with RAG works
Approved materials
I prepare service descriptions, price lists, FAQs and instructions. We identify up-to-date sources and exclude information that must not appear in customer answers.
Knowledge base search
RAG finds passages related to the question and provides them to the model to prepare an answer. Company rules come from the connected materials.
Clarification and handover to a staff member
I configure follow-up questions and the conditions for passing an enquiry to a person. The staff member receives the agreed data and conversation context so they can continue the discussion.
Testing and updates
Before launch, we test common and complex questions. After launch, we review conversations and update materials when services, prices or rules change.
Questions about business AI agent development
What is included in an AI agent from $400?
Preparation of a small knowledge base of an agreed size using your materials, one chat on an existing website, configuration of answers and question handover to a staff member, conversation testing and launch. One month of support and reasonable minor updates is included free. A new website, CRM, other channels and complex features are not included in this price. A 30% deposit is due before work starts. The balance is payable after you accept the agreed result.
What materials are needed for the knowledge base?
Service descriptions, current prices, FAQs, working terms, instructions and company pages. We start with one clear task and an agreed set of sources. Large archives, complex tables, scans and contradictory materials require separate preparation. We agree the size and content of the knowledge base before development.
How does a RAG agent differ from a standard chatbot?
A scripted bot guides users through predefined buttons and questions. A RAG agent searches the knowledge base for relevant passages and prepares an answer to a question written in the user's own words. Retrieving an order status or performing an action becomes possible only after the relevant system and access rules are connected.
Can an AI agent make mistakes?
Yes, a generative model can give an inaccurate answer. That is why we agree the sources, limits and handover conditions, and test common and complex questions. After launch, conversations help us identify weak points and refine the knowledge base. A staff member remains responsible for decisions about disputed terms, individual commitments and non-standard situations.
Can I add Telegram, MAX and CRM transfer?
An additional selected Telegram or MAX channel starts at $200; enquiry transfer to amoCRM or Bitrix24 starts at $300. Before quoting, I check service capabilities, accounts and access. Channels can share a knowledge base, but the interface, authentication and conversation handover are configured separately.
How does the agent retrieve order or account data?
Through a separately connected API from $400. We define the source, available fields and user permission checks. A document containing general terms alone cannot reveal the status of a specific order. A complex account area, roles and changes to data in the business system are quoted according to the task.
What costs are there besides development?
AI model usage, hosting and subscriptions for connected services. Costs depend on the number of conversations, message volume and selected integrations. Before work begins, we agree the project scope, development cost and timeline, and list external costs separately. Ongoing support is optional.
What is included in the free month of support?
Minor changes to agreed answers, sources and settings, and help with the agent's operation after launch. Significant knowledge base expansion, new use cases and integrations are quoted separately. I hand over the code and agreed access credentials. Support from the second month starts at $100 per month if needed.
AI CATALOG
An AI assistant on your website
A customer asks a question, the agent looks for the answer in company materials, and any enquiry that needs a staff member is passed on with its context.
from $400SPECIAL OFFER
One month of support and minor updates after launch included free 30% upfront. The balance is due after you approve the work. 
