MNRE Issues EoI for Market Assessment of AI-Based Retrieval Augmented Generation Systems

The Ministry of New and Renewable Energy (MNRE) has issued an Expression of Interest (EoI) to conduct a market assessment of AI-based Retrieval Augmented Generation (RAG) systems available in India.

The exercise is aimed at understanding the capabilities of advanced AI platforms that can improve document search, retrieval, and knowledge management across the ministry’s expanding repository of policies, scheme guidelines, technical standards, reports, circulars, and inter-ministerial communications.

According to the ministry, existing keyword-based and manual document retrieval systems are becoming inadequate due to the growing scale and complexity of information handled by MNRE. Through this exercise, the ministry intends to evaluate how AI-powered RAG systems can provide contextual, accurate, and source-backed responses by combining semantic search with large language model (LLM)-based generation technologies.

MNRE clarified that the EoI is only a market assessment and not a procurement process. The ministry stated that participation in the exercise will not create any commercial relationship, financial commitment, or preferential advantage in future tenders. Any future procurement, if undertaken, will be conducted separately through an open competitive process in accordance with General Financial Rules (GFR) 2017 and other Government of India guidelines.

The evaluation process has been divided into three stages. In the first stage, interested entities will submit technical and capability documents, which will be assessed based on predefined eligibility and scoring criteria. Companies and LLPs registered in India with demonstrable AI/ML or NLP-based document query systems are eligible to participate. The ministry will shortlist up to 15 entities based on factors including prior project experience, document handling scale, multilingual capability, supported file formats, and deployment flexibility.

Under the scoring framework, the highest weightage of 30 marks has been assigned to prior experience, especially projects executed for Central or State Government departments or public sector undertakings. Additional marks will be awarded for systems capable of handling more than 10,000 documents, supporting complex formats such as scanned PDFs, spreadsheets, and presentations, and offering multilingual functionality beyond English and Hindi. Systems deployable on-premise or through hybrid government cloud infrastructure will also receive higher scores.

In the second stage, shortlisted companies will provide MNRE with independent access to a working version of their RAG platform. The ministry’s testing committee will then conduct real-time testing using 250–500 MNRE documents, including scheme guidelines, circulars, annual reports, and press releases. The document corpus and query set will not be shared with vendors beforehand, and testing will be conducted without vendor participation or monitoring.

The testing committee will evaluate systems across several operational dimensions, including factual accuracy, hallucination control, source citation quality, natural language query understanding, multi-document synthesis, multilingual performance, and handling of scanned and structured files. The ministry will also assess whether systems can automatically adapt response formats such as narrative explanations, tables, summaries, and step-by-step outputs depending on query requirements.

MNRE stated that no formal scoring or ranking will be generated during the testing phase. Instead, the committee will prepare internal observation notes to support future policy and procurement decisions. At the conclusion of testing, participating entities will receive only a standard acknowledgement of participation.

Based on the observations from Stage 2, MNRE may optionally invite selected entities for a market dialogue to discuss system architecture, deployment strategy, scalability, and long-term roadmap. The ministry clarified that these discussions would be exploratory in nature and would not constitute negotiations, presentations, or procurement-related evaluations.

The EoI also places strong emphasis on compliance, data localisation, and governance requirements. Participants must disclose whether their systems store or process data within India, identify infrastructure providers and hosting locations, and clarify whether any component involves overseas data transfer. Companies must also specify whether user documents or interaction data are used for AI model training or fine-tuning.

Additionally, respondents are required to provide details on compliance with Government of India standards such as GIGW 3.0, MeitY AI guidelines, and Government Community Cloud (GCC) requirements. The ministry has also sought information on encryption practices, role-based access controls, audit logging mechanisms, and future compliance roadmaps as government AI standards evolve.

The last date for submission of responses is 27 May 2026, with shortlisted entities expected to provide system access within five working days of notification. MNRE said the exercise is intended to help the ministry better understand the current maturity of AI-based RAG technologies and shape the technical specifications of any future AI procurement initiatives.