Discover how Tribe built an AI-powered research assistant that reduced outside-in research time from 7-10 days to under 1 day and reduced manual analyst hours by 40%.
The Challenge
As part of its investment diligence process, a global consulting firm accelerates “outside-in” research by analyzing public data to assess risks, opportunities, and trends affecting target companies.
Outside-in due diligence was constrained by manual research workflows. Analysts spent days or even weeks combing through public data sources—news articles, regulatory filings, job postings, and customer reviews—assembling fragmented insights into reports. This labor-intensive process risked missing signals, delayed insights, and created inconsistencies across projects. The firm sought to bring execution down to 3 days by leveraging an AI-driven solution to automate signal detection, accelerate research timelines, and ensure comprehensive coverage.
The Solution
Tribe AI partnered with the firm to build an AI-powered research assistant embedded within its proprietary due diligence platform. The system pulled aggregated data from public and proprietary data sources, extracting relevant signals such as financials, employee rosters, salary data, leadership changes, and hiring trends. By surfacing actionable insights in a centralized dashboard, the platform streamlined research workflows and reduced manual effort.
Key Features
The research assistant incorporated features designed to deliver timely, actionable intelligence while reducing analyst workload:
- Benchmarks for industry level comparisons of key metrics
- Automated signal extraction for key diligence indicators like revenue per headcount, third-party costs and percentage of headcount in best cost location
- Noise reduction and prioritization, filtering irrelevant data and highlighting high-impact signals.
- Searchable, curated dashboard with direct source citations for validation and follow-up.
- Excel and PowerPoint compatible outputs to go seamlessly from digital to consultant friendly tools
- Feedback loops to learn from user’s edits and corrections to model outputs
Together, these features gave consultants a real-time pulse on target companies while reducing manual information gathering.
How It Works
The AI research assistant orchestrates multiple natural language processing and data pipeline components:
- Information aggregation across disparate public data sources.
- Information extraction models parse unstructured content to identify financial statements, employee counts, entities, events, and relationships.
- Relevance ranking models prioritize signals based on factors like recency, source credibility, and materiality.
- Insights are aggregated into a continuously updated dashboard, complete with source links for traceability and verification.
This system empowers consultants to search, explore, and synthesize insights without starting from scratch with every project.
Tech Stack
- Cloud: Azure
- LLM: Azure OpenAI - GPT 4o (with structured outputs)
- Other cloud services: web search to get more vendor data; Azure app services for hosting
- Languages used: Streamlit
Impact
The AI-powered research assistant accelerated outside-in due diligence, enabling faster identification of risks and opportunities while improving consistency across projects. Analysts saved time on manual research and gained confidence in the breadth and depth of coverage.
Key results included:
- Reduced outside-in research time from 7-10 days to under 1 day.
- Increased signal coverage by 30%, capturing insights that would have been missed in manual workflows.
- Accelerated reporting by 2-3 days, enabling earlier decision-making.
- Reduced manual analyst hours by 40% per diligence project.
The Future
This initiative has established a platform for ongoing AI-driven innovation within the firm. Building on the success of automated outside-in insights, the firm plans to extend its AI research assistant to cover adjacent diligence areas, such as market sizing, competitor monitoring, and ESG signals. The next phase will focus on expanding data sources, improving signal accuracy, and delivering proactive alerts tailored to each deal team’s needs.
