Technology
Enhancing Sell-Side Equity Research with Custom Small Language Models

In the fast-paced world of Wall Street, sell-side equity research teams are under constant pressure to produce high-quality, timely analyses of public companies. Their clients—hedge funds, mutual funds, and other institutional investors—demand actionable insights based on large volumes of data, from SEC filings and financial statements to earnings call transcripts and breaking news. A large Wall Street firm found its research arm struggling to keep up with these demands. Analysts were stretched thin by manual data gathering, repetitive modeling tasks, and the need to cover an ever-growing roster of companies.
This firm recognized that artificial intelligence (AI) could help them streamline operations and maintain their competitive edge, but concerns about data security, accuracy, and compliance made leadership wary of typical off-the-shelf large language models. In response, they engaged 1ConsultingSolution to develop a suite of custom small language models. The resulting transformation allowed the firm to save time, improve accuracy, and expand its coverage universe—without compromising on data privacy or regulatory standards.
Background and Challenges
The Wall Street firm had a storied reputation for thorough, high-quality sell-side equity research. Its analysts prided themselves on providing thoughtful commentary on quarterly results, running detailed valuation models, and producing sector-wide comparisons. However, they were increasingly bogged down by the sheer volume of information: corporate filings, market data, analyst consensus forecasts, CEO interviews, and more.
Key challenges included:
-
Manual Data Collection: Analysts spent hours combing through 10-Ks, 10-Qs, and transcripts to find relevant figures or management commentary. Each new quarterly cycle brought another wave of manual compilation.
-
Time-Consuming Updates: Initiating coverage on a new company or updating an existing model often involved duplicative tasks, from rewriting summaries to gathering similar data points across multiple sources.
-
Risk of Errors and Inconsistencies: With so much manual effort, errors could slip in—leading to inconsistencies in financial models or rating reports. Even small mistakes risked tarnishing the firm’s reputation with clients.
-
Difficulty Scaling Coverage: Leadership wanted to cover more companies, but analysts had limited bandwidth. Adding headcount was an option, but it was expensive and still didn’t address the underlying inefficiencies.
Regulatory constraints posed another concern. Because the firm’s research arm was a critical part of its market-facing identity, it had to meet compliance standards set by the SEC, FINRA, and internal legal teams. Leadership realized that a carefully crafted, secure AI solution could provide a way forward—but they needed a partner who understood both technology and the intricacies of financial research.
1ConsultingSolution’s Approach
1ConsultingSolution conducted a top-to-bottom assessment of the equity research process, talking to analysts, compliance officers, and IT staff. They identified the key points where automation could genuinely free up analysts’ time without replacing the human insight that made the research reports valuable.
Key Focus Areas
-
Targeted Use Cases
The team zeroed in on high-impact tasks: summarizing earnings calls, extracting key figures from filings, creating initial draft reports for new coverage, and updating valuation models with fresh data. By breaking these down into micro-use cases, 1ConsultingSolution could develop small language models that fit neatly into existing workflows rather than forcing a wholesale redesign. -
Custom Small Language Models
Instead of using a single, massive language model, 1ConsultingSolution developed multiple specialized models, each trained on curated data sets relevant to a particular set of tasks or industries. For instance, one model specialized in analyzing energy company filings, while another focused on tech sector earnings calls. This approach allowed for more accurate and secure handling of sensitive financial data, as well as faster inference times. -
Data Security and Compliance
All models were designed to run on the firm’s private cloud or on-premise infrastructure, ensuring that proprietary research and client data never left the organization’s controlled environment. Access control protocols were integrated so that analysts could leverage the models based on their permission levels, satisfying compliance mandates and internal policies. -
Seamless Integration
1ConsultingSolution built custom APIs that allowed analysts to query the small language models directly from their existing research platforms and spreadsheets. This tight integration minimized friction, letting analysts remain within the familiar tools they used every day.
Implementation Highlights
After finalizing the design, 1ConsultingSolution introduced a phased rollout across the research division. The energy and technology sectors served as pilot programs:
-
Earnings Call Summaries
For each earnings call transcript, the relevant small language model automatically generated a concise summary, capturing key metrics (e.g., revenue growth, EPS, guidance changes) and any notable management commentary. Analysts used these summaries as a quick reference, drastically reducing the time spent manually combing through transcripts. -
Automated Data Extraction from Filings
When a new SEC filing—like a 10-Q—was released, the models parsed the document, extracted essential line items and footnotes, and populated a draft update to the analyst’s existing valuation model. Analysts simply reviewed and validated the figures, ensuring a faster turnaround and reduced risk of manual entry errors. -
Draft Coverage Reports
When initiating coverage on a new company, analysts started with a model-generated draft that included key background information, competitor comparisons, and a summary of relevant financial ratios. This blueprint could be further refined and augmented with an analyst’s qualitative insights, cutting initial coverage time in half. -
Quality Assurance and Continuous Training
As analysts used the models, they provided feedback on errors or incomplete summaries. This iterative data was fed back into the system to continually improve performance. Compliance officers reviewed the outputs to ensure no inadvertent disclaimers were omitted or misleading statements introduced.
Results and Impact
In less than six months, the Wall Street firm witnessed tangible benefits across its sell-side equity research teams:
-
Time Savings: Analysts reported a 40% reduction in the time spent on routine data gathering and preliminary analysis. This freed them to focus on deeper financial modeling, client interactions, and thought leadership.
-
Improved Accuracy: Error rates in financial models and summaries dropped significantly due to the automated extraction and cross-referencing capabilities of the small language models. Fewer mistakes enhanced the firm’s credibility with institutional clients.
-
Expanded Coverage: With greater efficiency and reliable AI-driven support, the firm was able to increase the number of companies under coverage by 20%. This broader coverage appealed to a wider range of clients, generating additional revenue streams.
-
Enhanced Analyst Satisfaction: By offloading repetitive tasks to the AI models, analysts could devote more energy to high-value activities—such as investor calls, thought pieces, and advanced modeling. This shift improved morale and reduced turnover.
-
Compliance Confidence: Because all data processing happened internally, the firm met strict regulatory and compliance standards. Frequent audits of the small language models’ outputs demonstrated adherence to legal guidelines and internal best practices.
Conclusion
By developing custom small language models for a large Wall Street firm’s sell-side equity research division, 1ConsultingSolution ushered in a new era of efficiency and coverage expansion. The firm’s analysts saw immediate gains in time savings and accuracy, enabling them to service more clients and provide deeper insights. Moreover, carefully crafted compliance safeguards ensured that the technology enhanced—not jeopardized—the firm’s standing in a heavily regulated industry.
This case study highlights how a focused, small-model approach to AI can yield significant returns in the complex landscape of financial research. For organizations looking to remain competitive, improve productivity, and scale coverage in an ever-evolving market, custom small language models offer a secure and effective path forward.