How Multibagger AI analyses corporate announcements
An AI research brief for order wins and expansions, graded into verdict tiers, tracked against the market.
Multibagger AI reads corporate announcements about order wins and capacity expansions and asks one question: how large could this be relative to the business? It does not predict share prices.
Pipeline
- Extraction. A first model pass pulls the key figures out of the filing (order value, capacity change, execution period) and tags each number's role, so an order's value is never confused with the company's revenue or its cumulative order book. Every extracted number must be backed by a verbatim quote from the filing; unsupported numbers are dropped.
- Dossier. The announcement is combined with the company's point-in-time fundamentals (results, margins, balance sheet, promoter trend), its business profile and its peers.
- Brief. A second model pass writes a structured brief: the order's size relative to revenue, a revenue and EPS uplift scenario, bull, base and bear cases, and the key risks, ending in a verdict tier (Strong candidate, Possible, Unlikely).
What the track record shows
Every verdict is tracked against the market after publication. Measured on the first 392 matured analyses (held at least 30 days), the verdict tiers separated outcomes: Strong candidates beat the Nifty Smallcap 250 by about +11% on average (+2.4% median), while Unlikely picks lagged it by about −3% (−6% median). The continuous 0-100 score on its own carried little information (rank correlation with returns about 0.15). Individual calls fail in both directions. The best single performer in that sample was rated Unlikely. The tool is a research filter, not a buy list.
The models used are Google Gemini models on Vertex AI; the version is recorded with every analysis.
