Development of a chatbot and AI assistant for business with a knowledge base
RESTART develops corporate chatbots and AI assistants for business: the assistant responds to employees and clients using the company’s knowledge base - documents, regulations, contracts - and attaches a link to the source. access rights, logs and the data environment are customized to the customer’s requirements, and the assistant works where the work is already going on: on the portal, in Telegram, mail, Service Desk and by phone.
Corporate AI assistant instead of an answering bot
RESTART Group of Companies is the development of corporate chatbots and AI assistants with a knowledge base for companies: from a pilot in one process to production operations in the customer’s environment.
| Criterion | Autoresponder bot | Enterprise AI assistant |
|---|---|---|
| Where does the answer come from? | Pre-written scripts and templates | Company knowledge base: the answer is collected from found document fragments with a link to the source |
| If there is no answer | Pattern phrase or incorrect answer | The assistant reports that there is no reason and records a request to update the knowledge base |
| access rights | Same for everyone | Roles and rights: the response is based only on documents to which the user has access |
| Magazines | Not maintained or unavailable | Logs of requests, downloads and settings changes |
| Data outline | External cloud service | On-prem, private cloud or hybrid - according to customer requirements |
| Channels | One window | Portal, Telegram, mail, Service Desk, phone, API |
What AI assistants are we developing?
Knowledge Base Assistant
Answers on regulations and policies, project documentation, contracts and procurement materials, assistance to newcomers during adaptation. AI assistant with knowledge base for the company
Helpdesk chatbot
Classification of requests, hints to the operator, answers from the knowledge base and SLA control; channels - portal, web, mail, Telegram, internal bots. Chatbot for customer support
Voice assistant
Receiving requests by phone and via instant messengers, speech recognition, clarifying questions, generating and routing applications. Voice AI assistant VoiceHelp
Meeting assistant
Transcription, division of speech by participants, protocol with decisions and tasks, search through past meetings. Meeting Hub
Ragify assistant platform
Bots and knowledge bases, uploading documents with OCR, roles and access card, reply with a link to the document, web chat, Telegram, voice, public link and API. Ragify
Ragify is developed by Invent, a company of the RESTART group; VoiceHelp and Meeting Hub are part of the CleverHub line. If the process requires its own script, the assistant is designed on the kernel Restart AI Enterprise Platform, and where he must not only respond, but also act in systems, this is already AI agent.
Channels and integrations
- Where the assistant answers: web chat on a portal or website, Telegram, mail, voice mode and telephony, public link for external audiences, API for embedding in your own interfaces.
- Where does the knowledge come from: manual document uploads, question-answer pairs, corporate repositories like SharePoint, Confluence, portals and file storages.
- What systems does it work with: Service Desk and ITSM (Jira, Naumen, ServiceNow, 1C ITIL), CRM, 1C-Bitrix, ERP and 1C, EDMS.
How is the development of an AI assistant going?
Scenario selection
Define the business task, process owner, document sources, information security constraints, and success criteria.
Loop assembly
We set up the environment, upload documents, configure knowledge bases, assistants, roles and channels.
Check on real questions
We run through user questions, check answers with sources, and fine-tune the search and prompts. Format - RAG pilot on a corporate basis: 4 or 6 weeks.
production launch
We are formalizing the architecture, policies, access rights, training, and the roadmap for expansion to adjacent processes.
Security and Data environment
Roles separate users, administrators, downloaders and data owners; logs of requests, downloads and changes are maintained; sensitive data is masked, work with personal data is based on Federal Law No. 152-FZ. Hosting - on-prem, private cloud or hybrid; production data is not transferred to external AI services without the written permission of the customer. environment architecture - on page corporate AI environment.
Experience
- AI platform and RAG agents for a bank from the top 5 of Uzbekistan: the solution has passed the pilot and is under maintenance and development.
- RAG assistant for Spina Bifida for the foundation: answers based on documents and materials of the fund based on sources.
- The chatbot on this site is the RESTART assistant: it responds to visitors in Russian and English on the public pages of restart.re and provides links to them.
The entire implementation path is on the page implementation of AI in business.
Frequently asked questions
How much does it cost to develop a chatbot for business?
The cost of industrial development depends on the scenario, the volume of the knowledge base, channels and integrations. The first stage is fixed: RAG pilot on a corporate basis - from 700K – 3M rubles excluding VAT in the “Light” and “Full” editions. Licenses, computing resources and use of external models are considered separately and are shown before the contract.
How is a corporate AI assistant different from a regular chatbot?
A regular bot responds according to pre-defined scripts. The corporate assistant responds to the company's knowledge base with a link to the document, takes into account user access rights, keeps logs and works in the data loop that the customer chooses.
How does an artificial intelligence chatbot respond to a company's knowledge base?
Documents are indexed, relevant fragments are found for the question, and the answer is collected only from them - with a link to the source document. If there is no basis in the documents, the assistant reports so, and the request is recorded to refine the knowledge base.
Is it possible to connect an AI assistant to Telegram, mail and Service Desk?
Yes. The assistant works in web chat, Telegram, mail, voice and via API, and in support it works on top of Jira, Naumen, ServiceNow, 1C ITIL and internal ticket systems.
Is it possible to deploy an AI assistant without transferring data to external services?
Yes. The deployment option is selected according to customer requirements: on-prem, private cloud or hybrid. Production data is not transferred to external AI services without the written permission of the customer.
Where to start developing an AI assistant for a company?
From one process and a limited document base: the best pilot is 200–1000 documents of one subject area and a clear group of users, for example, support, legal department, project office, HR or information security.
Is it possible to make an AI chatbot for a website?
Yes. The assistant can also work for an external audience - in public mode using open company materials. An example is a chatbot on this site that responds to public pages of restart.re.
Let's discuss your environment
Describe the task, current systems, constraints, and expected results. We will offer a practical first step: diagnostics, pilot, audit, roadmap or project team.
