C1 Insights | Correlation One Blog

How Asset Managers Are Using AI to Unlock the Analyst Knowledge Trapped in Company Calls | Correlation One

Written by Correlation One | May 18, 2026

Every year, the analysts at a typical asset manager attend hundreds of company calls. They capture signals competitors miss—shifts in management tone, a CFO's hesitation before a forecast, an early read on a supply chain disruption. Most of that insight never leaves their notebook. A small group of leading firms have deployed AI to change that. Here's what the playbook looks like—and why every asset manager should be building it now.

By the Numbers
86% of hedge fund professionals already have access to generative AI tools (AIMA, 2024)
25–40% potential reduction in asset manager cost base through AI-driven efficiencies
95% of wealth and asset managers have scaled GenAI to multiple use cases (EY, 2025)
400K earnings call transcripts in BlackRock's fine-tuned investment model (BlackRock.com)

What Is the Analyst Knowledge Problem in Asset Management?

Asset managers have a knowledge distribution problem that has persisted for decades. Analysts are the firm's forward sensors—they attend company calls, management meetings, and industry conferences, and they develop nuanced, contextual views on the companies they cover. But that intelligence has always been local. It lives in one analyst's head, one team's morning notes, or buried in a CRM comment that no one reads twice.

The consequence is a systematic tax on investment performance. Portfolio managers make decisions with an incomplete picture. Junior analysts reinvent research their colleagues already completed six months ago. The firm's institutional memory resets every time a senior analyst walks out the door.

The problem is not that firms lack information—it's that the information doesn't flow. AI changes that equation entirely.

"In the past, [a research question] would have required reading a great deal of research and other material, as well as talking to analysts and attending company meetings. Now, we have internal tools that can look across all the research, earnings call transcripts, and public commentary, and answer very specific queries in great detail—often within hours."

— Brian Barbetta, Global Industry Analyst, Wellington Management (November 2025)

Three Firms That Have Built the Knowledge Infrastructure

The firms getting this right share a common architecture: they treat every company call as a data event, not just an analyst task. Here is what that looks like in practice.

Case Study 01
Wellington Management
Source: Wellington Management, November 2025

Wellington Management is a $1 trillion+ active asset manager with deep sector coverage across equities, fixed income, and alternatives. With analysts attending company meetings and calls spanning dozens of industries globally, the challenge of routing that accumulated intelligence to the right portfolio manager at the right time is immense.

In a November 2025 article published on their institutional insights platform, Wellington Global Industry Analyst Brian Barbetta described how the firm is solving it: internal AI tools that can query across the entirety of the firm's research archive, earnings call transcripts, and public commentary—returning detailed answers to specific questions in hours rather than days. Barbetta gave a concrete example: asking the AI what every company in health care is saying about AI. Previously, that would have required reading vast amounts of sell-side research, talking to analysts, and attending company meetings. The internal tool collapses that process into a single, rapidly returned query.

The implication is significant: individual portfolio managers can now access the collective intelligence of an analyst bench far beyond their own sector coverage, synthesized on demand. This is the knowledge democratization problem solved—not by hiring more analysts, but by building the infrastructure that makes every analyst's insights available to every PM.

Hours Time to answer complex cross-sector research queries, down from days
$1T+ Assets under management at Wellington Management
Case Study 02
BlackRock
Source: BlackRock.com, Systematic Active Equity Team

BlackRock's Systematic Active Equity team took the knowledge problem and solved it at a scale most firms can only aspire to. Rather than relying on individual analysts to surface call insights manually, they built and fine-tuned a proprietary large language model trained on over 400,000 earnings call transcripts spanning more than 20 years of market data.

The model is specifically designed to forecast how markets will react to corporate earnings—not as a general-purpose chatbot, but as a purpose-built investment tool. BlackRock publishes a direct benchmark comparison on their website, showing that their earnings call model demonstrated measurably stronger accuracy than general-purpose AI models like GPT, precisely because it was trained on investment-specific inputs.

The practical output for portfolio managers is a Thematic Robot that uses this transcript intelligence to help construct investment theses in minutes. BlackRock's published example describes a PM using the tool to build an investment basket around the GLP-1 pharmaceutical theme—the AI surfaced a comprehensive view of companies with direct and indirect exposure across the entire corpus of earnings call data, with full transparency into its reasoning. Overlaid on this is Aladdin Copilot, a generative AI layer that allows portfolio managers to query the firm's accumulated research in natural language. What previously required a senior analyst working across multiple systems for days can now be completed in a fraction of the time.

400K+ Earnings call transcripts in BlackRock's fine-tuned investment model
20+ yrs Market data used to train the earnings call model
Case Study 03
Balyasny Asset Management
Source: OpenAI Case Study, March 2026 / Bloomberg Tech Disruptors, April 2024

Balyasny Asset Management—a global multi-strategy firm running approximately 180 investment teams across asset classes and geographies—confronted a particularly acute version of the knowledge problem: with that many desks operating in parallel, insights generated in one corner of the firm rarely reached another. Their answer was to build a centralized AI research platform from the ground up.

In late 2022, Balyasny established a dedicated Applied AI team of 20 researchers, engineers, and domain experts, led by Chief AI Officer Charlie Flanagan, recruited from Google. Their mandate was to build AI-native tools that embed directly into team-level workflows—capable of synthesizing earnings materials, broker research, filings, and expert call notes across tens of thousands of documents simultaneously.

As of a March 2026 OpenAI case study, approximately 95% of Balyasny's investment teams actively use the platform. The impact is measurable and specific: a Central Bank Speech Analyst agent reduced the time required for macroeconomic scenario analysis from two days to approximately 30 minutes. A Merger Arbitrage Superforecaster agent continuously monitors deal probabilities in real time as new filings and press releases arrive, replacing manual spreadsheet tracking. Critically, Balyasny built the system to be proactive—rather than waiting for portfolio managers to query it, agents push relevant alerts on breaking developments, filing discrepancies, and emerging risks before anyone asks.

~95% Of Balyasny's 180 investment teams actively using the AI platform
2 days → 30 min Time for macroeconomic scenario analysis, per OpenAI case study

What the Implementation Playbook Actually Looks Like

Across these examples, the architecture follows a consistent pattern. Firms that get this right are not buying a single AI tool—they are making a deliberate sequence of infrastructure and workforce investments.

  1. Data foundation first. AI cannot extract insights from knowledge that has never been captured. The first investment is always in the infrastructure to ingest, transcribe, and structure call data—whether through a proprietary data lake or a third-party ingestion platform. The firms above invested in data pipes and governance infrastructure before they built the AI layer on top.
  2. Fine-tune for investment context. General-purpose AI models produce general-purpose outputs. BlackRock publishes a direct comparison showing their earnings-specific model outperforms general-purpose AI on the task of predicting post-earnings market reactions. The firms generating the most differentiated intelligence are those training or fine-tuning models on financial language, management communication patterns, and their own historical research archives.
  3. Build for portfolio manager workflows, not just analyst workflows. The goal is democratization. The system needs to be queryable and useful for the portfolio manager who was not on the call—not just a slightly faster tool for the analyst who was. Balyasny inverted this further: instead of PMs querying the system, the system proactively pushes insights to PMs.
  4. Develop the human layer. Technology does not adopt itself. Balyasny's ~95% adoption rate did not happen by accident—it required deliberate organizational change management. The most sophisticated AI implementations in asset management have been accompanied by programs to build AI fluency across investment teams, so portfolio managers know what to ask and analysts know what to feed in.

Where most firms are still falling short

The EY 2025 Wealth and Asset Management GenAI Survey found that 95% of firms have scaled GenAI adoption to multiple use cases—yet only about one in four executives reported substantial business impact. The gap between deployment and value is almost always a workforce problem, not a technology problem. Firms have access to the tools. They lack the trained practitioners who know how to deploy them against the right problems, evaluate their outputs critically, and build the feedback loops that make models progressively more accurate over time.

Frequently Asked Questions

What AI tools are asset managers using to analyze company calls?
Leading firms use a combination of proprietary large language models fine-tuned on earnings transcripts (as BlackRock has described publicly), retrieval-augmented generation (RAG) systems that query internal research archives, and generative AI platforms embedded in existing investment workflows. The most effective implementations are fine-tuned on firm-specific financial data rather than relying on off-the-shelf general-purpose models.
How does AI help portfolio managers access analyst insights from company calls?
AI systems can transcribe and index company call recordings, extract key themes and sentiment signals, and make that analysis queryable by any portfolio manager in the firm via natural language. Wellington Management has described this publicly: internal tools allow PMs to surface what every company in a given sector is saying about a topic—returning detailed synthesis in hours rather than requiring days of analyst interviews and reading.
What is the ROI of AI-powered knowledge management for asset managers?
Reported outcomes vary by application. Balyasny Asset Management reported in an OpenAI case study (March 2026) that macroeconomic scenario analysis that previously took two days now takes approximately 30 minutes. BlackRock has published data showing their earnings call model outperforms general-purpose AI in predicting post-earnings market reactions—with the implication that even marginal improvements in forecasting accuracy compound significantly over time. Industry analysts estimate AI can reduce an asset manager's cost base by 25–40% while expanding the coverage universe any given team can monitor.
What workforce capabilities do asset managers need to get value from AI?
The critical capabilities span several roles: prompt engineering and AI workflow design for analysts; AI literacy and critical evaluation of model outputs for portfolio managers; data governance and model oversight for risk and compliance teams; and AI strategy and vendor evaluation for investment leadership. The technology gap has largely closed. The skills gap is where most firms are still losing value.
How long does it take to build an AI knowledge management system for an asset manager?
It varies significantly by ambition. Simple implementations—capturing and indexing call transcripts with existing tools—can be operational within weeks. Firm-specific AI platforms with proprietary fine-tuned models and compliance guardrails (like what Balyasny has built) are multi-year programs. Balyasny established their Applied AI team in late 2022 and achieved ~95% team adoption by early 2026. The data governance infrastructure is typically the longest-lead component.

Why This Is No Longer Optional

The firms described above are not early adopters. They are setting the new baseline. The AIMA's 2024 survey of hedge fund managers found that 86% of professionals at those firms already have access to generative AI tools. The question is no longer whether AI will change how investment knowledge flows across an organization. It is whether your firm will be building that infrastructure proactively—or catching up to competitors who already have it running.

The firms that will compound the most advantage over the next decade are not necessarily the ones with the biggest teams or the most data—they are the ones whose people know how to translate raw AI capability into investment edge. That is an organizational capability. It is built deliberately, through training, workflow design, and a genuine commitment to changing how knowledge moves through the firm.

BlackRock has trained their model on 400,000 transcripts. Balyasny has ~95% of their investment teams actively using AI. Wellington Management's analysts are getting answers that used to require days in a matter of hours. Every asset manager sitting on years of company call recordings and analyst notes should be asking the same question: what is the cost of continuing to leave that knowledge in a silo?

Correlation One — AI Enablement

Ready to build the AI workforce behind your knowledge strategy?

Correlation One works with leading financial institutions to develop the AI skills their investment teams need—from portfolio manager AI literacy to advanced workflow design for analysts. We've trained 500,000+ professionals across 500+ employers globally. Let's talk about what your firm needs.

Speak with our team →

Sources: Wellington Management, "Is AI taking over? Portfolio and productivity insights for asset allocators," November 2025 (wellington.com). BlackRock Systematic Active Equity Team, "Reimagining alpha with data and AI," blackrock.com. BlackRock, "How AI is Transforming Investing," blackrock.com. OpenAI, "How Balyasny Asset Management built an AI research engine," March 2026 (openai.com). Bloomberg Tech Disruptors podcast, "Balyasny's Use of Generative AI in Investing," April 2024. EY, "GenAI in Wealth & Asset Management Survey," 2025. AIMA, "Getting in Pole Position: How Hedge Funds Are Leveraging Gen AI to Get Ahead," 2024.