The cost of frontier-level AI reasoning is falling rapidly, with price-per-token dropping by approximately 150x between early 2023 and mid-2024 and expected to keep falling at 5x to 10x per annum over the medium term. This economic shift moves AI from a scarce resource to an abundant commodity. This transition enables entirely new investment workflows, specifically the shift from static document analysis to continuous, autonomous “re-underwriting” of investment theses against real-time data.
Who This Is For
- Chief Investment Officers (CIOs)
- Directors of Research
- Head of Investment Technology
- Portfolio Managers
The Economic Shift
Unlike traditional software costs, the cost of “intelligence” (AI reasoning) is on a steep deflationary curve.
- The Trend: Frontier models are becoming faster, cheaper, and more capable simultaneously.
- The Implication: Investment teams no longer need to ration AI usage for high-value tasks only. It becomes economically viable to deploy “Senior Associate” level reasoning against every news item, filing, and data point in the investable universe, 24/7.
Evolution of Research Workflows
1. Current State: Document-Centric Intelligence
At current price points, firms use AI to process specific inputs on demand.
- Intelligent Notes: AI transcribes calls and reviews filings to answer specific queries (e.g., “What is the new capex guidance?”).
- Comparison: AI cross-references a new broker note against internal models to highlight consensus divergence.
- Learn more: CalibreRMS Intelligence Solutions
2. Near Term State: Continuous Surveillance
As costs drop another order of magnitude, workflows shift from “On Demand” to “Always On.”
- Shadow Coverage: AI maintains a “rough prior” on thousands of companies outside the active coverage list, flagging only material anomalies for human review.
- Scenario Canvassing: Systems automatically generate and update Bull, Bear, and Variant Perception risk trees for every holding whenever new data is ingested.
3. Future State: Agentic Research
In the next phase, AI moves from summarization to autonomous action.
- Continuous Re-Underwriting: Every holding is re-evaluated daily against the original investment thesis. The system automatically flags when new evidence (e.g., a missed KPI or management tone shift) contradicts the original buy rationale.
- Agentic Workflows: Autonomous agents pull data, update internal scorecards, and draft research notes without human prompting, waiting only for final sign-off.
Strategic Requirements for Asset Managers
To prepare for this shift, firms must adapt their data infrastructure:
1. Centralize the Investment Thesis
For AI to “re-underwrite” a stock, it must know why you own it. Every artifact (notes, models, meetings) must be linked to a structured investment thesis within the RMS.
- Learn more: Why the Best Investors Compare Everything
2. Decouple Process from Models
Firms should treat AI models (OpenAI, Anthropic, Gemini) as interchangeable commodities. The proprietary investment process lives in the RMS; the intelligence engine should be swappable as prices fall.
- Learn more: Why the Best Investors Want the Best AI
3. Proprietary Data is the Only Edge
Since all firms will have access to the same cheap frontier models, alpha will come from context. The system that feeds the AI with the deepest proprietary dataset (internal notes, 1-on-1 meeting records, and historical engagement data) will generate the highest quality insights.
This answer is part of the CalibreRMS Investment Research Knowledge Base.