Sergey Davidovich is Co-Founder and President of SparkBeyond, pioneer of the AI-powered Always-OptimizedTM platform that drives constant improvement in business operations. The Always-Optimized platform extends Generative AI’s reasoning capabilities to KPI optimization, discovering complex patterns across disparate enterprise data sources like CRM, ERP and system logs. SparkBeyond’s technology solves the hardest challenges in customer and manufacturing operations.
In this interview, Sagie shares his insights on the emerging trends in AI taking shape in 2025 that will impact how enterprises leverage this technology long-term. Sagie also highlights why Agentic AI is central to this discussion.
M.R. Rangaswami: Why is Agentic AI gaining momentum as an industry classification?
Sergey Davidovich: Agentic AI has rapidly emerged as a focal point in artificial intelligence. Unlike traditional AI tools that primarily respond to user prompts, Agentic AI is designed to autonomously analyze data, predict outcomes, and execute decisions with minimal human oversight.
According to Deloitte, 25 percent of companies using generative AI are expected to launch Agentic AI pilots by year end. Companies like Google, Salesforce, Microsoft, and HubSpot have already begun integrating Agentic AI.
These developments signal a shift from static automation to dynamic systems capable of continuous learning and adaptation, just like human knowledge workers would. At its core, this revolution revolves around two key concepts:
- Defining KPIs as actionable objectives for agents: By aligning agents with measurable business goals such as customer retention or revenue growth, organizations can ensure these systems focus on what truly matters.
- Enabling continuous improvement through real-time learning: This allows agents to refine their strategies dynamically based on evolving data and conditions.
M.R. Rangaswami: What is the missing piece in Agentic AI and how do we address it?
Sergey Davidovich: While execution capabilities are essential, future-proof autonomy in Agentic AI requires more than just task completion—it demands introspection and self-improvement.
At the heart of strategic intelligence lies hypothesis generation & testing—the cornerstone of the scientific method. This iterative process allows agents not only to learn what is true, but also to adjust their strategies and workflows when new data challenges existing assumptions. Hypothesis testing enables agents to explore possibilities beyond predefined guidelines, fostering innovation and adaptability.
As one industry example, we support clients with the ability to generate billions of hypotheses that ensures agents operate with a deep understanding of their environment and agent builders have the information they need to improve the agent and its ability to achieve its goals.
M.R. Rangaswami: Can you give a real-world example of Agentic AI at work?
Sergey Davidovich: In a marketing ecosystem powered by Agentic AI, optimization examples include:
- A Campaign Management Agent optimizes ad spend across channels
- A Customer Segmentation Agent refines audience targeting based on real-time behavioral data
- A Content Optimization Agent tests messaging strategies tailored for each segment
- A Marketing Mix Modeling Agent reallocates budgets dynamically based on performance metrics
These agents work collaboratively, adjusting strategies in concert to maximize overall marketing ROI while balancing KPIs such as customer acquisition cost and lifetime value.
M.R. Rangaswami is the Co-Founder of Sandhill.com