Industry scanning, target identification and post-investment tracking all ran on manual effort, with internal and external data siloed apart. The fund built an AI investment intelligence system on MOI: multi-source data is collected and structured automatically, linked into industry-chain, company and technology knowledge graphs, with reports and alerts generated on top.
An industry fund focused on advanced manufacturing, new energy, semiconductors and AI, whose research team works across a large volume of internal research reports, deal sourcing data and real-time external market intelligence.
The private-market investment cycle — industry scanning, target identification, post-investment tracking — remains a largely manual, labour-intensive process. Every sector the fund covers sits on a supply chain of hundreds of companies across tiers.
Information about those companies is fragmented across thousands of sources: annual reports, patent filings, news, industry associations, analyst research, social media — and it changes constantly.
So analysts spend most of their time collecting and organising data rather than analysing and deciding. That is a structural mismatch: the fund pays for judgement and consumes information-handling hours.
Siloed internal and external data made it worse. Internal research and deal sourcing records lived in one system and external intelligence in another, with no way to cross-check between them, so analysts stitched the two together in their heads.
The result was that emerging companies and technology shifts were missed, or found too late — by the time something reaches mainstream view, the valuation reflects it. Post-investment was the same story: tracking portfolio dynamics and sector shifts at scale is not sustainable by hand.
The fund built an AI investment intelligence system on MOI, covering three core workflows: industry landscape scanning, target profiling and post-investment monitoring.
The data layer runs as an AI-driven ETL workflow: more than 60 external sources are collected and processed automatically, deduplicated, cleaned and tagged, then landed alongside internal research reports and deal sourcing data in one store — which is what removes the internal/external split and makes cross-checking possible at all.
On top of that sits the knowledge graph: entity recognition across industries, companies and products, relation extraction across competition, partnership and supply-chain links, plus event mining and sentiment analysis — producing multi-dimensional graphs of industry chains, company profiles, patent landscapes and event timelines. Its value is not what it stores but that it connects scattered facts into relationships, which no spreadsheet replicates.
The application layer generates intelligence reports and alerts, supports natural-language questions via NL2SQL and RAG over the private knowledge base, and produces charts and reports automatically.
Analysts get 80% of their time back, moving from information gathering to judgement, analysis and relationship-building — the work the fund actually pays for.
Intelligence shifts entirely from static reports to live updates. A sector report used to start going stale the day it was finished; the graph now refreshes as its sources do.
More than 60 external sources are mined and updated automatically, giving continuous coverage of the trends, emerging companies and technology shifts that manual research would miss or find too late.
The system went live in two weeks, showing that this capability does not require a long data-platform build as a precondition.