Barkha Saxena

Another chapter in our long-term strategic partnership with LSEG - Access to AI-ready financial data in customer workflows. Using a new MCP server, #LSEG customers will be able to connect with data, licensed through LSEG products like Workspace and Financial Analytics, to build #AIAgents in #CopilotStudio. #AI #Agents #MCP #FinancialServices #CopilotStudio #M365Copilot #CapitalMarkets #FutureofFinance

22

Well said. AI is no longer a technology investment but becomes part of the organizational strategy. It is forcing organizations to rethink its structure—not just tools. Integrated and multi-skilled teams are essential for speed of AI innovation, and new AI-driven roles will continue to blur traditional role boundaries.

1

𝗛𝗶𝗿𝗶𝗻𝗴 𝗔𝗹𝗲𝗿𝘁: AI in Risk is steadily moving from experimentation to operational use. While Generative AI is opening new possibilities, meaningful progress still depends on 𝗦𝘁𝗿𝗼𝗻𝗴 𝗔𝗱𝘃𝗮𝗻𝗰𝗲𝗱 𝗗𝗮𝘁𝗮 𝗔𝗻𝗮𝗹𝘆𝘁𝗶𝗰𝘀 foundations and 𝗵𝗮𝗻𝗱𝘀-𝗼𝗻 𝗽𝗿𝗮𝗰𝘁𝗶𝘁𝗶𝗼𝗻𝗲𝗿𝘀 who can translate data into decisions. As we continue building and scaling these capabilities — and gradually expanding into GenAI-driven use cases — we are looking to strengthen our team with 𝟮 𝗿𝗼𝗹𝗲𝘀 at different levels. 𝗥𝗶𝘀𝗸 𝗔𝗻𝗮𝗹𝘆𝘁𝗶𝗰𝘀 𝗟𝗲𝗮𝗱 – 𝗔𝗩𝗣 A senior role focused on shaping and delivering advanced analytics solutions that support risk identification, anomaly detection and decision support The role will involve working closely with stakeholders to translate analytical capabilities into practical outcomes and help scale analytics adoption within the organisation. 𝗥𝗶𝘀𝗸 𝗔𝗻𝗮𝗹𝘆𝘁𝗶𝗰𝘀 𝗔𝗻𝗮𝗹𝘆𝘀𝘁 (𝗗𝗲𝘃𝗲𝗹𝗼𝗽𝗲𝗿) A hands-on developer role focused on 𝗯𝘂𝗶𝗹𝗱𝗶𝗻𝗴 𝗮𝗻𝗮𝗹𝘆𝘁𝗶𝗰𝘀 𝘀𝗼𝗹𝘂𝘁𝗶𝗼𝗻𝘀 and models that support risk analytics use cases. The role will involve working with datasets, developing analytical models, contributing to analytics and emerging AI initiatives.  If this sounds interesting, feel free to reach out. Always happy to connect with people interested in building **AI and advanced analytics capabilities**. I will add the 𝗮𝗽𝗽𝗹𝗶𝗰𝗮𝘁𝗶𝗼𝗻 𝗹𝗶𝗻𝗸𝘀 in the comments below. #AdvancedAnalytics #RiskAnalytics #ArtificialIntelligence #Transformation #RiskManagement #Hiring

25

8 Comments

Snowflake + DataOS: The Stack That Actually Makes Enterprise AI Work How do we scale AI across the enterprise? Most answers point to models and compute. But that’s not where AI breaks. It breaks in the gap between data storage and data control. Snowflake solved one half of the problem. • Unified data platform • Elastic compute + storage • Governed system of record It became the foundation layer of enterprise data. But,nA system of record is not a system of action. Enter the missing layer: Data Control The Modern Data Company and its DataOS introduce a data control plane that sits above Snowflake. Not replacing it, Operationalizing it. Snowflake + DataOS = From Data Storage → Data Intelligence This combination turns enterprise data into a programmable system for AI: • Data → Products (discoverable, reusable, versioned) • Governance → Code (real-time, policy-driven) • Pipelines → Supply Chains (resilient, cross-domain) • AI → Production Systems (not experiments) The Market Reality We’re now looking at a multi-trillion dollar AI stack, but the dollars are unevenly distributed: 1. Data Platforms (Snowflake, etc.) - 150B+ market opportunity → Already well understood, heavily capitalized 2. AI / Applications / Agents -$500B–$1T+ projected over the next decade → Where most of the hype is focused 3. The Missing Layer: Data Control Planes - Emerging $50B–$100B+ category And it’s massively underbuilt. Why this layer is inevitable: At scale, every enterprise hits the same wall: • 100s of data sources • 10s of platforms • Fragmented governance • Broken pipelines • AI models starved for trusted, real-time data The Parallel is clear: • Kubernetes didn’t replace compute → it orchestrated it • Service meshes didn’t replace apps → they connected them • Control planes didn’t replace infrastructure → they made it usable at scale DataOS is bringing that same paradigm to enterprise data The Strategic Insight Snowflake owns where data lives. DataOS controls how data flows, is governed, and is activated. Together, they formnthe operating system for enterprise AI The next wave of value creation in AI won’t come from: • More compute or storage • Or bigger models It will come from: Coordinating data across the enterprise in real tiime The winners in enterprise AI will control the data plane. Not just the data. AI doesn’t scale on data lakes. It scales on data systems.

20

Another trend, not surprisingly, is that domain-specific embeddings consistently outperform generic models on specialized tasks. This doesn’t necessarily help with competing on broad benchmarks like MTEB, which may be one reason it doesn’t receive as much attention. However, in practical applications, the advantages of domain-focused training are clear and often substantial. https://lnkd.in/gBkuWs7E https://lnkd.in/gEGWuDS6 or https://lnkd.in/gyan-Ant https://lnkd.in/grZ6-Dpa.

25

Claude’s latest Fable 5 model is getting attention for good reason. But Anthropic made two key changes around the launch that have real operational impact: 1. No zero data retention for Fable 5 Anthropic’s support docs say prompts and outputs for Mythos-class models are retained for 30 days for trust and safety purposes, on every platform where these models are offered. For organizations using zero data retention configurations, that is a material change. In some enterprise environments, it may be a blocker for rollout. 2. Usage-based pricing after June 22 (bye-bye subsidies) Anthropic says Fable 5 is available to Pro, Max, Team, and seat-based Enterprise subscribers until June 22. On June 23, using it will require usage credits (aka API pricing) unless capacity allows an extension. Fable 5 is Anthropic's most expensive model yet, priced at $10 per million input tokens and $50 per million output tokens. So usage that feels “included” during the launch may become a meaningful usage-based cost very quickly. But the bigger point is that this probably will not be the last time this happens. Frontier models are getting more capable, more expensive to run, more capacity constrained due to compute shortage, and now they're becoming less straightforward from a procurement and governance perspective. We are moving into a world where teams cannot assume: - the best model is included in the subscription - the newest model has the same data policy - every model from the same provider has the same retention terms - experimentation costs will stay small - employees will naturally choose the right model for the job The takeaway here is that AI adoption and governance can no longer stop at the tool layer. It needs to move down to the model and workflow layer. The companies that handle this well will not default every task to the newest frontier model. They will build a model routing strategy: - Sensitive data goes only where the retention policy fits. - High-reasoning/planning work goes to frontier models when the value justifies the cost. - Repeatable execution routes to smaller, faster, cheaper models. - Usage gets monitored by workflow, not just by vendor. - Exceptions have an owner. You can already see the market moving this way. Model-routing gateways, cloud model catalogs, fallback logic, tracing, and spend controls are becoming part of the AI infrastructure layer (eg: Respan). That does not mean every company needs a complicated routing setup tomorrow. But it does mean model choice should be intentional at the organization level, not discovered later through a surprise usage bill. The question is not just “which AI tools do we allow?” It is “which model should handle this workflow, given the data, risk, quality requirement, and cost?” That is where AI governance is heading: less about approving tools once, and more about managing model usage as the models, prices, and policies keep changing.

5

4 Comments

Nasdaq's Bill Dague, Head of Data Product, explains how Databricks Marketplace, powered by OpenSharing, is expanding access to eVestment's trusted, AI-ready data. By bringing trusted investment data closer to customers' workflows, Nasdaq enables faster, more confident analytics and AI initiatives.

53

4 Comments