Slide 1: How to Optimise Financial Data for AI: Tools, Techniques, and Use Cases
LSEG Academy James Perkins Director of Customer & Solutions Learning Global Data, Feeds & Analytics How to Optimise Financial Data for AI: Tools, Techniques, and Use Cases
Slide 2: Disclaimer
LSEG LSEG Academy Academy Disclaimer The content in this presentation is provided by LSEG for your personal information only and is not intended for trading purposes or deciding to carry out a transaction or trade. Nor does it provide any form of advice, or make any recommendation regarding financia
Slide 3: Resources
LSEG Resources Academy LSEG Academy lseg.com/en/training/learning-centre Data Catalogue lseg.com/en/data-analytics/financial-data Data Discovery datadiscovery.lseg.com 24x7 Customer Support New support.lseg.com myaccount.lseg.com Developer Community developers.lseg.com Learn More About Financial AI
Slide 4: Agenda
LSEG Agenda Academy 1. Financial Data & Analytics AI Landscape 2. Financial AI – Hype vs. Reality 3. Optimising Financial Data for AI 4. AI Data and Technology Requirements LSEG Academy LSEG | LSEG
Slide 5: Financial Data & Analytics AI Landscape
Financial Data & Analytics AI Landscape
Slide 6: Defining Artificial Intelligence (AI)
LSEG Defining Artificial Intelligence (AI) Academy LLM GenAI Agent Agentic ”AI” has a range of definitions LSEG Academy LSEG
Slide 7: Financial Data Catalogue
LSEG Financial Data Catalogue Academy An Industry View Instrument Types Data Types • Bonds • Date Time • Loans • Integer Data Categories • Currencies • Percentage • Commodities • Float Macroeconomic • Indices • String Pricing & Reference • Economic • Equities Unstructured Text • Funds Company • Futu
Slide 8: Focus on Data and Global Presence
LSEG Academy Focus on Data + Global Presence 40k customers, 400k users, 170 countries Partnerships and Brands Cloud Collaborative AI Collaborative Managed services, consulting, data and technology SMEs LSEG LSEG Academy
Slide 9: Data & Analytics Trends
LSEG Data & Analytics Trends Academy Where are the shifts? • Cloud migration • Liquidity fragmentation • Algo trading • Sustainable finance • Crypto & digital assets • Risk-free rates (RFRs) • Regulatory mandates (MiFID, Basel 3.1 / FRTB, DORA, SFDR, 2a-5) Source: IDC & Seagate Data Age 2025 • EU AI
Slide 10: Financial AI Hype vs. Reality
Financial AI Hype vs. Reality
Slide 11: Financial AI+ Themes
LSEG Financial AI+ Themes Academy How AI Is Growing Fast On Wall Street JP Morgan Investment Banking & Risk • LLMs automating tasks: email, research, creation. Risk Management • AI models forecast exposure to interest rate, credit, and liquidity risks, aiding compliance with stress- testing regulati
Slide 12: Financial AI- Themes: What's not being highlighted
LSEG Financial AI- Themes Academy What’s not being highlighted • Up to 85% of AI projects in finance fail—mainly due to data issues, talent gaps, and misaligned strategies. Source: ProPair/RedPoint Research, Forbes • Challenges Stalling AI Implementation In The Financial Sector. Source: Forbes • 30%
Slide 13: AI is only as good as the data behind it
LSEG Financial AI - Data Academy “without the right data, even the best algorithms can deliver mediocre, or worse, misinformed results.” David Schwimmer Chief Executive Officer, LSEG World Economic Forum AI is only as good as the data behind it LSEG Academy LSEG
Slide 14: Top AI Data Sources Model Challenges
LSEG Top AI Data Sources Academy Model Challenges • Low signal-to-noise ratio • Echo chambers • Informal language • Licensing • Context fragmentation • User consent • Demographic skew • Redundancy Source: statista LSEG Academy LSEG
Slide 15: Financial Data - Expertise
LSEG Financial Data - Expertise Academy Across equity and fixed income are the implications of corporate actions understood? Are nuances like trade qualifiers being captured in trading models? For futures and options are contract rollover adjustments being made? Does the model consider regional mark
Slide 16: How to Optimise Financial Data for AI
How to Optimise Financial Data for AI: Tools, Techniques, and Use Cases
Slide 17: Optimising Macroeconomic Data for AI
Optimising Macroeconomic Data for AI
Slide 18: Macroeconomic Data
LSEG Macroeconomic Data Academy Understanding the global economy + Real-Time ü Real-time Indicators ü Economic forecasting ü Economics point-in-time ü Emerging markets ü Consumer and business sentiment Historical ü National and international ü Comparable economics Source: LSEG Datastream, LSEG Works
Slide 19: Optimizing Macroeconomic Data for AI
LSEG Optimizing Macroeconomic Data for AI Academy Considerations • Useful in macro forecasting models or signal enrichment for trading. • Low-frequency, lagged, often revised data. • Requires feature engineering to handle sparsity and noise (rolling averages, lagged features, statistical methods) Ri
Slide 20: Optimising Pricing Data for AI
Optimising Pricing Data for AI
Slide 21: Pricing Data
LSEG Pricing Data Academy Valuing securities Timeframe ü Real-Time (streaming, delayed) ü Historical Source ü Exchange-traded instruments (ETI) ü Over the counter (OTC) Type ü Contributed pricing ü Composite pricing ü Evaluated pricing “Pricing” can refer to absolute 90 million price, or to spread,
Slide 22: Optimizing Pricing Data for AI
LSEG Optimizing Pricing Data for AI Academy Considerations • Supports high-frequency, real-time models, valuations. • Requires trade qualifier filtering, outlier detection, tick aggregation (e.g., VWAP). Risks: • Noise amplification, anomalies • Look-ahead bias • Overfitting to historical prices. Gu
Slide 23: Optimising Reference Data for AI
Optimising Reference Data for AI
Slide 24: Reference Data
LSEG Academy Reference Data Security master ü Security details ü Entity details ü Descriptive information ü Identifiers ü Fixed Income Terms and conditions: ü Issuer Info ü Maturity Date ü Amount Issued ü Amount Outstanding ü Coupon ü Schedules ü Ratings ü Prospectus… Source: LSEG Workspace 80 milli
Slide 25: Optimising Reference Data for AI
LSEG Optimising Reference Data for AI Academy Considerations • Often used across platforms (trading, risk, compliance). Data Lineage • Some fields are static (e.g., ISIN, issue date), others change (e.g., credit rating, status). • Ensure consistent identifiers and definitions across systems to avoid
Slide 26: Optimising Symbology for AI
Optimising Symbology for Al LSEG
Slide 27: Symbology
LSEG Symbology Academy Data mapping and stitching • Combining datasets using common identifiers. • 100+ different market identifiers which can change: SEDOL, ISIN, CUSIP, RIC… PermID: open, permanent, machine-readable identifier spanning organizations, instruments, funds, issuers, and people Source:
Slide 28: Optimising Symbology for AI
LSEG Academy Optimising Symbology for AI Considerations API Query API Response • Symbology is essential for all securities { "from": [ { workflows. { "data": [ "identifierTypes": [ { • Necessary for linking across datasets (i.e., "LEI" "input": [ ], { pricing, reference, analytics). "values": [ "val
Slide 29: Optimising Unstructured Text for AI
Optimising Unstructured Text for AI
Slide 30: Unstructured Text: News, Research, Filings, Transcripts
LSEG Unstructured Text Academy News, Research, Filings, Transcripts ü Financial markets ü Capital markets ü Commodities ü Companies and industries ü Sentiment ü Breaking news ü Research reports ü Transcripts ü Filings Source: LSEG Workspace, MarketPsych, Machine Readable News 2-3 million news articl
Slide 31: Optimising News & Research for AI
LSEG Optimising News & Research for AI Academy Considerations: • LLMs used for summary, classification, and sentiment. • Reuters Super Summaries Unstructured Data • LSEG MarketPsych NLP Engine • NLP must handle jargon, sarcasm, and entity disambiguation. • Requires temporal alignment with market dat
Slide 32: Optimising Transcripts for AI
LSEG Optimising Transcripts for AI Academy MarketPsych Transcript Analytics • Original text • Detailed JSON output Doc • All aggregated levels • UI and API Section Speaker Sentence Detailed word level NLP output Aggregation levels Exploring AI Mentions in Earnings Calls and Building Thematic Portfol
Slide 33: Optimising Company Data for AI
Optimising Company Data for AI
Slide 34: Company Data
LSEG Company Data Academy Understanding and valuing companies • Unstructured ü News & Research ü Filings ü Transcripts • Structured ü Pricing ü Reference Data • Company events • Corporate actions • Deals data and league tables • ESG • Fundamentals • Estimates • Organisation entity data • Ownership,
Slide 35: Optimising Company Data for AI
LSEG Optimising Company Data for AI Academy Considerations • Both structured (revenue..) and unstructured data (footnotes.). • Companies report on different fiscal calendars. • Data may come from filings (10-K, 10-Q), vendor feeds, or company websites. Risks • Data gaps and missing values. (especial
Slide 36: Optimising Risk Intelligence Data for AI
Optimising Risk Intelligence Data for AI
Slide 37: Risk Intelligence Data
LSEG Risk Intelligence Data Academy Tracking global risk and regulation ü Sanctions ü Law and regulatory enforcement ü Politically exposed persons (PEPs) ü State-owned enterprises ü Adverse media ü Ultimate Beneficial Owner ü Account Verification ü Identity Verification ü Regulatory Data ü KYC (know
Slide 38: Optimising Risk Intelligence Data for AI
LSEG Optimising Risk Intelligence Data for AI Academy Considerations • Use cases include document verification, fraud detection, and/or spotting political/legal risk. • Variations in how names, addresses, or IDs are recorded can confuse models. • Data is often governed by strict rules (e.g., GDPR).
Slide 39: Optimising Analytics for AI
Optimising Analytics for Al LSEG
Slide 40: Analytics
LSEG Academy Analytics Calculation engines ü Valuations ü Hedging ü Spreads ü Yields ü Volatilities ü Risk metrics ü Greenium ü Modelling ü Statistics ü Term Structures Source: LSEG Analytics API, LSEG Workspace, LSEG Yield Book Analytics are derivative data and are limitless LSEG Academy LSEG
Slide 41: Optimising Analytics for AI
LSEG Academy Optimising Analytics for AI Considerations • Pervasive throughout the financial ecosystem. • Use cases across valuations, risk, trading. • Only as good as the input data behind them. Risks • Potential for black box – challenge to explain to clients and regulators. • Regulatory complianc
Slide 42: AI Data and Technology Requirements
AI Data and Technology Requirements
Slide 43: AI Data and Technology Requirements
LSEG Academy AI Data and Technology Requirements 7 Key Questions Latency 1. What frequency of data does your AI model require? Volume 2. What are your AI data consumption volumes? Transfer 3. Where does your AI model reside? Delivery 4. What’s the best delivery for your AI? Access 5. Who needs acces
Slide 44: Data Latency
LSEG Academy Data Latency What frequency of data does your AI model require? Real-Time (RT) Intraday End of Day (EOD)+ Pico (ps) 10-12 Nano (ns) 10-9 Micro (µs) 10-6 Milli (ms) 10-3 Intraday Daily+ Imaging Algo & Systemic Trading Real-Time P&L Strategy and Research Computing High Frequency Trading M
Slide 45: Data Volume
LSEG Academy Data Volume What are the AI data consumption volumes? latency 1. Security “Watch List” Size (unique instruments / datapoints) Data Requirement Examples 4ms 1s 15m EOD 2. Coverage (regional, global, segment) 20 Historical Economic Series 20 3. Timeframe (# Days, Years, Min, Sec…) 100 FX
Slide 46: Data Transfer
LSEG Academy Data Transfer Where does your AI model reside? 1. Infrastructure Web Deployed Hybrid Cloud D.R. 2. Connectivity Bandwidth, PoP, Speed Desktop API HTTP 3. Delivery Method Excel SQL Digital Files Website SFTP RSS Feed LSEG Academy LSEG
Slide 47: Data Delivery
LSEG Academy Data Delivery What’s the best delivery for your AI? Desktop Website API SFTP / FTP SQL Single user Single or multi- Supports real-time Bulk Data File Direct, complex, access user. data + high traffic Delivery high-performance Desktop + Mobile API Decision Tree database operations Perfor
Slide 48: Data Access
LSEG Academy Data Access Who needs access to the AI data? Single User Multi-User Group Firmwide Internal = Distribution Desktop Enterprise External = Redistribution Technical, Workflow, and licensing considerations * Data can as well be accessed through Partners solutions and applications MCP AI Mod
Slide 49: Licensing
LSEG Licensing Academy Are there any AI restrictions on your data? Raw Data Timeliness of Data • As received: can only be used in the format as Real-time data in most cases has different duties – supplied by the Data Provider fees, obligations and restrictions – from delayed, • Normalized: data can
Slide 50: Maintenance
LSEG Academy Maintenance How are you managing AI data changes? • Data is constantly changing: 1. Market events 2. Data provider changes 3. Regulations • Ongoing data and system updates are needed LSEG + Industry Changes Data Notifications | MyAccount (lseg.com) Notifii API LSEG Academy LSEG
Slide 51: Recap - How to Optimise Financial Data for AI
Recap - How to Optimise Financial Data for AI
Slide 52: Key Points
Key Points 1. Financial Data & Analytics AI Landscape • Defining AI: Data Science vs. LLMs • Regulatory Drivers, Liquidity Challenges, and Financial Data 2. Financial AI – Hype vs. Reality • 85% of AI projects in finance fail—due to data and skills issues… • AI is only as good as the data behind it
Slide 53: LSEG Resources
LSEG LSEG Resources Academy 24x7 Data Catalogue lseg.com/en/data-analytics/financial-data Data Discovery datadiscovery.lseg.com Academy lseg.com/en/training/learning-centre Customer Support New support.lseg.com myaccount.lseg.com Developer Community developers.lseg.com Learn More About Financial AI
Slide 54: Thank You
LSEG Academy Thank You LSEG Academy LSEG