Latest news with #ArtyomKeydunov
Forbes
08-07-2025
- Business
- Forbes
Aligning On One Source Of Truth: Why It's Hard—And How To Get It Right
Artyom Keydunov, CEO, Cube. I've written extensively about the universal semantic layer, an emerging (and sometimes misunderstood) concept. A semantic layer abstracts a corporation's business metrics, or standard units of measurement, from its data. Adopting a universal semantic layer isn't just a technical deployment—it's a cultural shift. Getting every team in a large enterprise to pull from the same data definitions requires more than installing new software. It demands structured change management, deliberate training and a commitment to cross-functional alignment. Making The Semantic Layer Work In practice, success starts with executive sponsorship. Teams often fall back into siloed habits without senior leaders reinforcing the importance of shared definitions and unified metrics. A cross-functional governance committee should be formed, ideally including finance, HR, sales, operations and IT stakeholders. This group defines, refines and owns the business logic embedded in the semantic layer. Next comes training—not just on how to use the tools but also on why the change matters. Data consumers must understand that consistent metric definitions (like "revenue" or "active customer") are essential for trustworthy insights. Hands-on workshops, role-based dashboards and embedded documentation all help reinforce these principles. But expect friction. Teams may resist relinquishing control of their own metrics. To address this, create a clear escalation path for reconciling disputes and showcase early wins demonstrating improved efficiency and decision-making from unified data. One of the most overlooked challenges is integrating the semantic layer into daily workflows. Implementing it is not enough—it must be embedded into the tools teams already use, like BI dashboards, spreadsheets or CRM platforms. The path isn't easy, but enterprises can achieve data alignment that scales by prioritizing executive buy-in, cross-team governance, contextual training and workflow integration. Benefits Of A Universal Semantic Layer Strategy A well-implemented universal semantic layer helps deliver measurable business value by unifying definitions and enabling trustworthy, real-time insights across departments. Here's how a successful semantic layer strategy can be realized by different teams: • Finance And Accounting: Finance teams often operate in fragmented environments with legacy OLAP systems and conflicting reports. Ensuring everyone is pulling consistent metrics—like revenue recognition or margin analysis—from a single, auditable source helps to centralize financial definitions, reduce reconciliation time and improve regulatory confidence. • Human Resources (HR): HR decisions suffer when data is spread across disconnected systems. The ability to consolidate performance, engagement and compliance data into one coherent view empowers more proactive, strategic workforce planning. • Sales: Sales leaders frequently lack a unified view of pipelines, customer behaviors and KPIs. With standardized definitions across CRM, emails and BI tools, teams unlock better forecasting, lead prioritization and coaching decisions based on facts, not guesswork. • Marketing: Disparate sources of campaign data lead to missed insights and wasted spending. By aligning attribution and ROI metrics across platforms, marketing teams can execute and adapt with real-time clarity. • Operations And Supply Chain: Fragmented supply chain data undermines efficiency and responsiveness. A consistent lens across ERP, logistics and inventory systems helps drive better planning, cost control and risk management. • Customer Support: As customer-facing teams often struggle with slow, inconsistent service due to scattered data, a semantic layer provides unified access to service histories and communication logs for faster, more accurate responses. Alignment First, Technology Second The universal semantic layer's promise is powerful, but it is not automatic. The technology itself can only take organizations so far. Success depends on how effectively teams align on definitions, processes and responsibilities. Creating a shared understanding of key business metrics across departments is a profoundly human challenge. However, when companies invest in organizational scaffolding—governance, communication and education—the semantic layer becomes more than a data architecture. It becomes a strategic advantage. Those who get it right aren't just managing data better—they're making better decisions, faster and with greater confidence. That's a competitive edge every organization should aim for. Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?
Forbes
03-04-2025
- Business
- Forbes
How Businesses Can Get The Most Out Of AI Agents
Artyom Keydunov, CEO, Cube. Experts predict that 2025 will be the year of AI agents, as evidenced by the emergence of AI agent-based products to automate workloads across nearly every function. It's no wonder we're entering the era of the agentic workforce: The business benefits are manifold, including increased productivity and lower expenses. The productivity and cost savings apply to every AI-driven workspace transformation, from HR to marketing. Before AI agents, organizations had to increase headcount to increase productivity. Today, AI agents can help businesses achieve the same productivity boost without increasing headcount. In 2025, many white-collar roles, including data and analytics professionals, will have an AI teammate. AI agents are ideal for automating many of their functions. For instance, AI agents can help data teams automate day-to-day operations, such as building pipelines, data models, reports and dashboards, optimizing query performance and cost and monitoring data issues. The data and analytics workflow is similar to the software engineering workflow, where AI-driven tools have proven that AI agents can tackle complex software engineering problems. Although these tools can't be directly translated for use in the data domain due to its unique challenges and integrations, it's possible to take similar approaches and build specialized tools. In addition to tackling rote tasks, agentic data professionals can help answer ad hoc requests submitted by business users. Historically, only those skilled in SQL could query databases to make decisions, making it difficult for business users to access the data they needed to make decisions. Text-to-SQL solutions have cropped up to fill this gap. Powered by large language models (LLMs), text-to-SQL systems allow non-technical users to query databases using everyday language. The issue is that most text-to-SQL solutions assume that users will ask very specific, well-crafted, reasonable questions. In reality, users' queries are seldom straightforward. Sometimes, questions are incorrect. Sometimes, they can't be answered due to missing data. Sometimes, it's both. As a result, data analysts spend a lot of time understanding and correcting incoming questions. They must work with business users to refine questions before responding to questions and building the resulting reports. Human data analysts develop a deep understanding of the subject matter to navigate business users' queries. When dealing with requests, they engage in conversations with stakeholders to understand what they're looking for and what can be done based on available or potentially available data (pending changes to data pipeline, collection or transformation). For AI agents to be effective, they need to navigate these requests to arrive at a set of questions that can be answered correctly. To do this, they need to perform deep research on the existing data model and connections between different data assets. If the current model can't answer the question, then they need to run multiple prediction scenarios to see what changes can be made to the existing model to generate the correct answer. Suppose AI agents for data and analytics have the proper context and knowledge of the data. In that case, they can clarify the question and understand if the right underlying data assets are available to answer it. Building effective AI agents for data and analytics is nearly impossible without that framework and context. Research, understanding of the existing model and predictive modeling can't be done without a knowledge framework. The knowledge framework AI agents need is a universal semantic layer. The semantic layer maps business-level definitions, key performance indicators (KPIs) and organizational corporate jargon to data fields. Instead of requiring users (or AI) to know table and column names, the semantic layer ensures that phrases like "lifetime customer value" or "annual recurring revenue" are automatically translated into precise database references. These connections allow AI agents to navigate and understand business requests as they come in. In previous posts, I mentioned the need for training and change management in implementing a universal semantic layer—and that remains true. Often, the technology is quite simple. The bigger job is managing change throughout the organization. The first step in moving toward an agentic AI workforce is to be aware that no organization can successfully make the transition without implementing a universal semantic layer, and the effort is well worth it. New AI agents will allow knowledge workers—including data engineers—to concentrate on high-value tasks. Ultimately, organizations that successfully implement AI agents will save costs, optimize workflows and productivity, drive innovation and accelerate insights that benefit the business. Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?
