Ai4impact co-op | Fall 2023- Spring 2024 project
Assistive Buyers Engine
(ABE)
https://drive.google.com/file/d/1cV5e1-IJH2Z1tzama-9KKWqoo9Vp7-G9/view?usp=sharing
ABOUT THE PROJECT
Generative AI tool to enhance the procurement process for state buyers by enabling efficient navigation of procedures and regulations and offering support in understanding their application.
Impact: There are more than 100 statewide contracts available, and each one has unique instructions that cater to the contract. Balancing understanding these long and complex documents with numerous negotiations, regulations, and other responsibilities can make the procurement process overwhelming for buyers.
Partner: Massachusetts Operational Services Division
Status: The tool is functional and is currently being tested and readied for production. This term the team is adding functionality to allow for the tool to compare legal requirements against bids. This is a legal and time-consuming process where lawyers must sit and redline contracts comparing bids against protocols. The new tool features being implemented will review, markup and summarize these for lawyers saving time and energy.
Partners

Executive Summary
Massachusetts’ procurement process is complex, demanding precise adherence to detailed regulations and guidelines, leading to potential delays, compliance risks, and inefficiencies. The Assistive Buyers Engine (ABE) addresses this challenge through a generative AI solution developed to streamline and simplify procurement workflows. ABE integrates a comprehensive knowledge base, including user guides, statewide contracts, and procurement handbooks, with advanced AI capabilities to provide context-aware guidance, precise conflict detection, and enhanced search functionality. The anticipated impact includes improved procurement efficiency, enhanced compliance, and reduced operational delays, benefiting procurement professionals and the Commonwealth of Massachusetts. This project was a collaborative effort involving the Massachusetts Operational Services Division (OSD), the Burnes Center for Social Change at Northeastern University.
Problem Context
State procurement processes involve complex regulations that buyers must navigate carefully. Manual interpretation of procurement guidelines and resolution of procedural conflicts result in significant delays and compliance risks. With increasing contract volumes, the need for a scalable, automated solution has become critical. ABE addresses these issues by leveraging generative AI to offer instant, precise, and contextually relevant guidance directly to procurement officials.
Innovation Process
The development of ABE followed an Agile and Human-Centered Design approach, emphasizing iterative improvement based on real user feedback. Initial phases involved extensive discovery sessions with procurement teams, identifying pain points, procedural bottlenecks, and common challenges. This feedback directly informed the design of ABE’s interfaces and functionality. Procurement experts at the OSD participated actively in testing and refining the tool, ensuring its practicality and alignment with user needs. ABE’s knowledge base is sourced from official procurement documents, guidelines, and statewide contracts to ensure accuracy and compliance.
AI Solution Overview
ABE is a comprehensive AI-powered assistant structured around intuitive user interfaces:
- Homepage: Enables users to submit queries via text or audio, maintain session history, and access additional resources.
- Data Dashboard: Provides administrators with functionalities to manage the knowledge base through uploads, synchronization, and deletion, ensuring data accuracy.
- User Feedback Page: Allows continuous refinement by enabling admins to monitor, review, and act on user feedback efficiently.
- LLM Evaluation Dashboard: Allows administrators to conduct automated testing, upload test datasets, evaluate ABE’s responses, and review clear summaries and performance metrics.


Outcomes & Impact
ABE has recently been deployed into production. While comprehensive outcomes are still emerging, anticipated impacts include:
- Potential reduction in procurement delays through instant access to accurate information.
- Enhanced compliance by ensuring procurement decisions align closely with Commonwealth regulations.
- Improved efficiency and satisfaction in procurement workflows, enabling procurement professionals to focus on strategic tasks.
Early evaluations indicate strong potential for substantial improvements in procurement efficiency across Massachusetts.
In collaboration with the Operational Services Division who handles nearly $3 billion per year in state procurement contracts under the Executive Office of Administration and Finance, students built two AI tools to enhance the procurement process and to accelerate legal review of vendor submissions
Lessons Learned
ABE’s iterative, user-focused development process highlighted the critical importance of early and frequent stakeholder engagement. Continuous feedback loops allowed rapid refinement of AI functionalities, ensuring that the tool was both effective and user-friendly. Challenges faced included balancing the complexity of procurement regulations with user-friendly design, underscoring the need for ongoing refinement and adaptive learning capabilities within the tool.
Future Roadmap
ABE has successfully transitioned into production. Future efforts will focus on:
- Continuous monitoring by maintainers
- Regular updates based on real-world feedback and performance data
- Ongoing user engagement to refine usability and functionality
Acknowledgments
The development of ABE was a collaborative effort by Prasoon Raj, Rui Ge, Ritik Bompilwar, and Divya Hegde, under the AI4Impact initiative at the Burnes Center for Social Change, Northeastern University. We extend our gratitude to everyone involved at the Operational Services Division (OSD), Commonwealth of Massachusetts, for their invaluable support and collaboration throughout this project.
Link to Source Code
PROJECT TEAM
- Prasoon Raj
- Rui Ge