In large and medium-sized companies (banks, insurance, leasing, and factoring companies, etc.), employee work is governed by internal regulatory documents, namely regulations, policies, procedures, and instructions. These are typically lengthy documents that regulate employee behavior in various situations. However, it is difficult for employees to read them in full. Furthermore, these documents are periodically updated, and it can be hard to keep track of changes. This is a problem because employees, not knowing how to act in accordance with internal documents, may make mistakes or spend a lot of time searching for answers to their questions in these documents.
Moreover, company lawyers sometimes receive questions, the answers to which are contained in the company's local regulatory acts.
This problem can be solved by:
-
A process navigator;
-
A chatbot that answers questions based on the content of the company's internal document.
The process navigator will guide employees through the process, advising them on what and how to do at each stage. It also helps employees: generates necessary document forms, performs automated calculations based on received data, sends documents and data via email.
This is a service whose algorithms accurately reflect the algorithms of business processes described in the company's local documents (regulations, policies, procedures, instructions, etc.).
If business processes change for any reason, the process navigator algorithms must also be adjusted.
Thus, all company employees, regardless of their department, will work according to uniform and up-to-date algorithms.
To turn local regulatory documents into algorithms, it is necessary to study them thoroughly and build an algorithmic tree based on the business processes described in these documents.
At Botman.one, you can create graphical algorithms based on processes described in local documents, which are immediately transformed into a working service – a business process navigator. Such a navigator can be installed on your company's corporate portal for employee use or connected to messengers as a chatbot (Telegram / MAX/VK).
For example, this is what a fragment of a process navigator algorithm for generating contracts with teachers for RANEPA employees looks like.
A chatbot consulting on user questions is created based on one of the Large Language Models (LLMs).
To create such a chatbot, a special system prompt is set, instructing the AI on the form in which to provide an answer to the user's question. The company's internal document (in PDF format), on which the chatbot should consult, is also fed into the neural network.
We have looked at how to make it easier for employees to work with standard business processes or get answers to questions about the content of local company acts. But is it possible to create a service that will advise employees on legal issues instead of lawyers?
Yes, that is also possible.
Let's consider creating such a service using the example of a chatbot that consults on constitutional law based on the text of the Russian Constitution. You can try the result here – on the website and here – in Telegram.
And a demo video of this service is available here.
Creating it on the Botman.one low-code platform takes about 20 minutes.
It is necessary to create a field for the user to enter a question, then add a file with the text of the Constitution, create a system prompt (which will also include the user's question), create an action to send the Constitution file and the system prompt along with the user's question to the neural network, and a message into which the neural network's response will be inserted.
This is what this chatbot looks like inside the Botman.one platform.
On the platform, you can choose from many available language models you want to use: DeepSeek, ChatGPT, Grok, etc. On our platform, all these neural networks work through Proxy API and Open Router. Therefore, to use them, you will need to pay for a Proxy API key, which then needs to be inserted into a special field when creating the action to send data to the neural network. Getting a key on the Proxy API website is very simple; you just need to create an account there and top it up in RUB from any Russian bank.
This is what creating an action in Botman.one to send a request to a neural network looks like (the order of entering information is marked with red numbers).
As you can see, it's not complicated, as everything happens in a simple visual environment on the low-code platform.