Success stories • Oil & Gas
Productivity, efficiency and intelligence from end to end
A combination of solutions enabled this major integrated energy company to optimize its upstream, its drilling activities, and its bid development and HR processes.
About the client
Leading Argentine integrated energy company that develops activities in the oil industry, including upstream, midstream, downstream, power generation and participation in the renewable energy sector.
It is the second largest oil and natural gas producer in the country and operates in the main basins: Golfo San Jorge, Noroeste, Neuquina and Marina Austral. It contributes 16% of the hydrocarbons produced in the country and employs more than 20,000 people, including its own personnel and contractors.
Upstream optimization with the power of data
Client’s need
• Exploit information from diverse databases centrally without requiring managers to have knowledge of the underlying data models or SQL.
• Increase efficiency of drilling engineers.
Solution
Nubiral developed a solution based on Generative AI and serverless integrations to data, that allows:
• Information request in natural language through a conversational interface.
• Request transformation into a SQL query that is automatically executed in the corresponding database engine, by using a language model (LLM).
• With the result of the query, an LLM is again used to transform the response (typically numeric) into natural language. The user receives the natural language data, the generated SQL query and an explanation of how the SQL query was created.
Maximum productivity for drilling engineers
Client’s need
• Exploit information from various sources (databases, business rules, mining simulation software) in a centralized manner.
• Simplify drilling engineers’ tasks: minimize the number of tools and data sources they have to interact with for analysis or design.
Solution
Nubiral developed an end-to-end solution that allows drilling engineers to interact with data in a more agile way. Using Amazon Bedrock and AWS native integration services, current processes were optimized and the power of AI was leveraged to streamline tasks such as finding drilling data, simulation and comparisons for different extraction methods (electro submersible, gas lift, etc.).
Improved HR expertise and processes
Client’s need
• To have a conversational assistant that allows its 20,000 employees to access HR and process information through a conversational interface.
• To have a single chat to interact with the AI and, depending on the user’s query, identify the case/agent and data to refer to in order to query and build the response.
Solution
Nubiral developed a Generative AI-based conversational assistant capable of having a conversation with end users through a web interface, allowing them to ask questions about HR issues and processes in natural language.
Using an LLM-based orchestrator and agents accessing various information sources (RAG with vector databases, prompt engineering and APIs to retrieve information from databases) the question is processed, information is collected and an appropriate response is generated.
Speed and efficiency in the generation of requests
Client’s need
For contracting products and services, the company uses RFPs (request for proposals, RFP) in which it describes the product or service required and the conditions. Based on this, third-party companies submit bids that are compared to select the one that best suits the request. The manual generation of these specifications is time-consuming. The objective was to implement a tool that would allow them to be generated faster, with less effort and standardized.
Solution
– Implementation of a conversational wizard based on LLM models, through which users can be guided through the process of gathering information to assemble the specifications.
– The user interacts through a chatbot interface.
– The bot prompts the user to describe the type of solicitation needed to identify the type of RFP.
– The bot uses the RFP model to ask the user for all the information needed to generate the RFP. For each response, the bot checks if it adequately addresses the question. If not, it asks follow-up questions.
– With all questions answered and details from the database, the application starts the RFP generation work.
