Back to case studies
Corporate AI Assistant and a Single Document Base for an Energy Holding
- Overview
- Challenges
- Services
- Functional Modules
- Technologies
- Our Achievements
Client
A European energy holding that develops solar and wind power plants for businesses and supplies electricity.
Technologies
Java / Python / Claude / Amazon Web Services / PostgreSQL / OpenSearch / Google Drive / OneDrive / Telegram
The company's contracts, projects and reports were scattered across Google Drive, OneDrive and employees' computers. Finding the right file was hard, and new employees took a long time to get up to speed on projects. We built a platform that gathers all documents in one place, sorts them on its own and answers employees' questions about company documents.
Challenges
Collect documents from Google Drive, OneDrive and employees' computers in one place.
Bring order to the files and automatically sort them into contracts, projects, proposals and reports.
See the key points of a contract at a glance: the client, price, deadlines and terms.
Give employees an AI assistant that answers questions about company documents.
Speed up the preparation of commercial proposals for clients.
Keep the AI's work and its costs under control.
Services
CRM and AI-Powered Document Processing
A single document base with an AI assistant. The system finds the client, price and deadlines in contracts on its own, and an employee asks a question and gets an answer based on company documents.
Read more
Business Process Audit and Solution Design
We reviewed which documents the company has, where they are stored and how employees work with them. We designed the platform on that basis.
Read more
End-to-End Software Development
We built a web platform with an AI assistant, a file manager, user management and document processing settings.
Read more
AI Automation and Integrations
We connected Google Drive, OneDrive and folders on computers, and files move into the platform on their own. Employees can also ask the AI a question in Telegram without opening the platform.
Read more
Agile Development, Accelerated by AI
We worked in short stages and showed the client a finished part of the platform after each one. Tasks and documentation were kept in Jira and Confluence, so the client could always see the progress.
Read more
DevOps and Support
The platform runs in the Amazon cloud and starts processing by itself as soon as a new file appears. It is easy to scale when the number of documents and employees grows.
Read more
Functional Modules Delivered
Document Collection
Files from Google Drive, OneDrive and employees' computers move on their own into a single secure storage.
The history shows what was transferred and when.
As soon as a new file appears in the storage, the platform starts processing it on its own.
Automatic Sorting
The AI reads a file and determines what it is: a contract, a technical document, a proposal, a project or a report.
The administrator confirms the AI's choice, and the file goes into the right section.
The platform reads PDF, Word and Excel files, as well as scans and photos.
Key Contract Terms
The AI finds the client, price, dates, term and main conditions in a contract.
Instead of 50 pages, the employee reads a short list of the key points.
Contracts from all sources are collected in one place instead of being scattered across folders and computers.
AI Assistant
Employees ask about contracts, clients and processes on the platform or in Telegram.
The AI understands the meaning of a question and finds the answer, even when the document words it differently.
New employees get up to speed without distracting their colleagues.
Comparison and Proposals
The system compares projects, proposals from different companies or versions of a document and shows the differences.
In a conversation with a client, the AI selects solar panels, inverters and energy storage.
The client immediately sees the project cost and how much they will save or earn on electricity.
AI Control and Costs
For each document, you can see what stage it is at and whether there were any errors.
Every employee question and every AI answer is saved in a log.
The cost of processing is known in advance, and daily limits protect against unnecessary spending.
Technologies used
Platform:
Java and Python. File transfer and the AI's work are separated, so one does not get in the way of the other.
AI:
Claude by Anthropic through Amazon Bedrock. Company data is not used to train the AI.
Semantic Search:
OpenSearch finds a document by the meaning of the question, not by an exact match of words.
Document Reading:
The platform reads PDF, Word and Excel files, as well as scans and photos.
Cloud:
Amazon Web Services. Processing starts on its own as soon as a new file appears.
Integrations:
Google Drive, OneDrive and Telegram are connected to the platform.
Security:
Sign-in with Google, role-based access and a log of document processing and AI answers.
Our achievements
All documents in one place
Contracts, projects and reports from all sources are collected and sorted into sections.
An answer instead of a search
An employee asks a question and gets an answer based on documents, with no digging through folders.
Faster start for new employees
Newcomers find answers on their own and get involved in projects sooner.
Proposals during the client conversation
The AI selects equipment and calculates the client's benefit right in the conversation.
Predictable AI costs
The cost of processing is known in advance, and limits protect the budget.
AI under control
The administrator confirms categories, and every AI answer is saved in a log.
Let’s talk
Have any questions or feedback? Send us a message!
Our contacts
