Monday, May 11, 2026 | 16 mins read05/11/2026 | 16 mins read
This edition of The Data Massagist explores the rise of agent ecosystems as the next evolution of enterprise AI. Instead of isolated copilots or chatbots, organizations are moving toward Multi-Agent Systems (MAS) where specialized AI agents collaborate, delegate tasks, and consume the outputs of other agents to execute end-to-end workflows. The article explains why this shift is happening now and how enterprises are adopting coordinated intelligence patterns such as triage (intake and prioritization), routing (dynamic task delegation), and orchestration (workflow execution and control). It highlights how industries like telecom, healthcare, manufacturing, and energy are already applying these models, and why the future of AI will depend on building scalable, governed ecosystems of interacting agents rather than standalone tools.
Thursday, May 7, 2026 | 7 mins read05/07/2026 | 7 mins read
Microsoft Fabric uses a unified, token-based AI billing model where all AI features — including Copilot for Power BI, Copilots in Fabric, Data Agents, and Operational Agents — consume Capacity Units (CUs) from the organization’s Fabric capacity. Instead of separate AI licenses or per-prompt fees, costs are calculated based on input and output tokens, with output tokens typically driving higher consumption. The article explains how AI workloads are monitored through the Fabric Capacity Metrics App, Admin Portal, and Activity Logs, giving organizations visibility into token usage, CU consumption, and workload spikes. It also clarifies licensing considerations for Power BI Copilot and highlights the difference between Data Agents (AI that answers) and Operational Agents (AI that acts autonomously). Ultimately, the model provides predictable, transparent, and centralized AI cost management within Microsoft Fabric.
Monday, April 13, 2026 | 9 mins read04/13/2026 | 9 mins read
Organizations can keep on-premises SQL Server systems while enabling modern AI by using Microsoft Fabric. Data is continuously mirrored into OneLake, avoiding complex ETL and enabling real-time analytics without disrupting operations. A key requirement is building a strong semantic layer that defines business meaning, ensuring accurate, governed AI insights. Fabric Data Agents then provide natural-language access to this curated data via tools like Copilot. This architecture separates operational and analytical workloads, improves scalability, and enforces governance. The result: faster insights, reduced maintenance, and trusted AI-driven analytics—while preserving existing systems and modernizing incrementally.
Tuesday, March 31, 2026 | 11 mins read03/31/2026 | 11 mins read
As organizations adopt AI-driven analytics, exposing trusted data to Copilot and Fabric Data Agents requires strong architecture—not just enablement. Microsoft Fabric Data Agents add conversational analytics over governed data, but report-embedded semantic models create fragile AI behavior, unclear cost ownership, and governance risk—especially as customers migrate from Power BI Premium to Fabric capacities. Through two Contoso case studies, the article shows why extracting reusable, standalone semantic models is essential for AI readiness. By combining governed semantic models with Fabric Mirroring for Oracle, organizations achieve predictable AI costs, stable and explainable AI responses, centralized security (RLS/OLS), and scalable foundations for Data Agents and future Copilot experiences. The key takeaway: AI succeeds when semantics are treated as first-class data products.
Tuesday, February 3, 2026 | 10 mins read02/03/2026 | 10 mins read
In the second edition of The Data Massagist, Pablo Junco Boquer explores what truly powers Agentic AI solutions such as Microsoft Copilot, Copilot for Power BI, and Fabric Data Agents. While AI adoption and ROI are accelerating, Pablo argues that real AI accuracy does not come from better prompts or newer models—it comes from better data foundations. AI failures, he explains, are data problems, not model problems. The “real magic” happens in preparation: trusted, well‑modeled, and governed data expressed through strong semantic models. Using Microsoft Fabric, organizations can turn raw data into AI‑ready knowledge by aligning business meaning, storage modes, and governance. Semantic models become the shared language between humans and AI, enabling agents to reason accurately, scale understanding, and deliver reliable business outcomes without confident mistakes.
Saturday, January 31, 2026 | 4 mins read01/31/2026 | 4 mins read
In Why Agentic AI Starts with a Calm, Governed Data Foundation, Pablo Junco Boquer introduces The Data Massagist newsletter and argues that agentic AI succeeds only when data is “calm.” Calm data is prepared, unified, observable, and governed, enabling AI agents to act with trusted context rather than hallucinate or over‑escalate. AI failures, he explains, are usually data operating system problems—not model issues. Using the “data massagist” metaphor, he frames his work as preparing, governing, and modernizing data so AI and analytics share a strong foundation. Grounded in global, hands‑on experience, Pablo emphasizes business outcomes over tools, focusing on revenue, cost, risk reduction, and faster time.
Subscribe onLinkedIn to get the latest Pablo Junco's perspectives and insights to move from messy data to measurable outcomes — governed platforms that power agentic AI.
If you’re a Chief Data Officer (CDO), a data leader, or simply someone who believes in the power of preparing data for AI—you’re already a Data Massagist.
Whether you have an idea, a challenge, or just want a fresh perspective, let’s connect. I’m always open to collaborating, learning, and helping others move forward.
