New information moves prices, but in a stream of alerts, filings and commentary it took an investment firm's analysts hours to identify which events mattered. New Clarity built an agent that ingests news feeds, regulatory filings and analyst reports continuously, determines whether each item is the kind of event the firm's strategy acts on, estimates the probability of a price move within thirty days, and ranks the results for analysts by confidence, sector and relevance. Thirty days later the actual outcome is fed back and the model is updated, so accuracy is measured rather than assumed. Analysts decide what, if anything, to act on.
The firm needed to identify relevant events as they broke rather than hours later, and it needed a system that stayed accurate as market conditions changed instead of one calibrated once and left to drift.
A language model reads each item and classifies whether it is an event type in the firm's strategy, such as an acquisition, a contract win, a regulatory approval, a forecast change or a buyback, so irrelevant items are discarded before any analysis runs. A statistical model then estimates the probability of a price move within thirty days from market capitalization, trading volume and other variables. Thirty days after each prediction, the actual outcome and the market data for the period are fed back and the model is retrained, so its accuracy is tracked continuously against what happened. A dashboard ranks the events for the analysts by confidence, sector and relevance.
The agent recommends and people decide. Every flagged event carries its source, its score and the data behind it, so an analyst can see why it was raised. Outcomes are recorded against each prediction, so the firm can show how accurate the system has been, what it said, and when.
Within the first month the agent's signals were operating better than anticipated, and the measurement loop tracked continued improvement as new data arrived. The firm got more timely decisions and fewer missed opportunities, each with a record behind it. The same design, continuous monitoring, scoring, a human decision and a recorded outcome, applies to deal-flow screening, covenant and deadline monitoring, and news on portfolio holdings.