DATACON 2026: Building AI-Ready Microsoft Data Estates

Overview

• Speaker: Steve Weiss and David Hay
• Channel: Datavail
• Main topic: Modernizing Microsoft data estates to make them AI-ready
• Purpose: Provide organizations with a practical guide and lessons learned on how to audit, govern, modernize, and scale their data estates to prepare for AI implementation. Steve Weiss and David Hay from Datavail present lessons from the field on modernizing Microsoft data estates to make them AI-ready. Drawing on decades of consulting experience across thousands of databases and hundreds of managed service customers, they walk through a multi-step framework moving from discovery and inventory through governance, master data management, capacity planning, and fan experience optimization. They highlight common pitfalls such as lack of visibility into legacy infrastructure, unmanaged database sprawl, shadow IT, lack of definitions/governance, and compute capacity shortages in the era of AI workloads.

Topic Map

Introduction and Speaker Background

• Explanation: Steve Weiss and David Hay introduce themselves, outline their background in database consulting and AI solutions, and set up the session's football-themed presentation style.
• Key claims:
  • Datavail manages over 100,000 databases across 400+ managed service customers.
  • The ultimate goal of modernizing data conversations is making data AI-ready.
• Examples:
  • Combined 20+ years of consulting conversations around data and data modernization.
• Terminology:
  • Datavail
  • Solutions Architect
  • Managed Services
• Why it matters: Establishes credibility and frames the entire presentation around real-world consulting experience.

Phase 1: Discovery and Inventory

• Explanation: Organizations often don't know what data assets they actually have. Building a comprehensive inventory of source systems, owners, and last usage dates is the foundational first step.
• Key claims:
  • You cannot modernize what no one owns.
  • Uncovering legacy tools like Access databases, SQL Server 2016, SSIS, and old vendor platforms is critical.
• Examples:
  • A global container leasing company in transportation and logistics discovered countless shadow databases and legacy tools only after inventorying.
• Terminology:
  • Inventory Table
  • Source System
  • Midnight Gymnastics
  • Extended Support
• Why it matters: Without knowing what data exists and its dependencies, modernization timelines slip and unexpected legacy blockers emerge.

Phase 2: Migration and Infrastructure Planning

• Explanation: Moving data off legacy, end-of-life systems (such as SQL Server 2016) onto modern cloud platforms (Azure SQL Managed Instance, Azure SQL Database, Microsoft Fabric) with minimal friction.
• Key claims:
  • Moving from on-premise SQL Server to Managed Instance has the lowest friction and lowest rework.
  • Waiting too long until end-of-support deadlines causes panic-driven, poor architectural decisions.
• Examples:
  • Global equipment manufacturer using SQL Server 2016 facing end-of-extended-support deadlines.
• Terminology:
  • Managed Instance
  • Azure SQL Database
  • Microsoft Fabric
  • Extended Support
  • UAT
• Why it matters: Sticking with unsupported infrastructure creates severe security vulnerabilities and higher long-term maintenance costs.

Phase 3: Governance and Master Data Management (The Gold Layer)

• Explanation: Ensuring data quality, standardized definitions, and trusted master lists across the organization so different departments speak the same language.
• Key claims:
  • Trust in data is lost when different departments use conflicting definitions for the same metric.
  • Metadata and definitions must be codified in a gold layer or customer master list.
• Examples:
  • An energy infrastructure operator experienced alignment issues where finance and operations defined 'availability' differently.
  • A high-end women's apparel retailer suffered from duplicate records and unmastered customer lists across departments.
• Terminology:
  • Gold Layer
  • Data Quality Checks
  • Master Data Management
  • Data Stewards
  • Metadata
• Why it matters: AI tools will give rapid answers, but if the underlying data is unmastered and governed poorly, AI will rapidly scale incorrect business decisions.

Phase 4: Capacity Planning and Compute Scarcity

• Explanation: Addressing the surging demand for compute and AI capacity, avoiding sudden scarcity, and utilizing multi-region redundancy.
• Key claims:
  • Demand for AI compute has outstripped the availability of tools to service it.
  • Organizations must plan for multi-region discipline and redundancy rather than relying solely on single-region capacity.
• Examples:
  • A professional sports venue experienced fan experience failures when high traffic overwhelmed regional capacity during game day.
• Terminology:
  • Compute Capacity
  • H100
  • Multi-Region Discipline
  • Redundancy
  • Scarcity
• Why it matters: Without proper capacity planning and multi-region failover, scaling AI initiatives can lead to massive service outages and customer dissatisfaction.

Key Points

Inventory before modernizing

• Explanation: You cannot build an AI-ready data estate if you do not know what databases, applications, and legacy spreadsheets exist in your organization.
• Evidence: Consulting experience across hundreds of Datavail customers showing pervasive shadow IT and undocumented legacy tools.
• Practical implication: Build an inventory table tracking source systems, owners, and last usage dates before undertaking any migration.

Choose the path of least friction for migrations

• Explanation: Moving from on-premise SQL Server to Azure SQL Managed Instance reduces refactoring friction compared to jumping straight to alternative tech stacks.
• Evidence: Client migrations stalling due to complex application layer rewrites when shifting off end-of-life SQL Server 2016.
• Practical implication: Prioritize managed cloud database instances to minimize rework and meet deadlines.

Establish a Single Definition of Truth (The Gold Layer)

• Explanation: Self-service analytics fails when different departments have conflicting definitions for core business metrics like 'availability' or 'revenue'.
• Evidence: Energy infrastructure client where finance and operations had completely different formulas for equipment availability.
• Practical implication: Appoint data stewards and codify master definitions into a centralized gold layer before layering AI on top.

Anticipate Compute and Capacity Scarcity

• Explanation: The massive demand for AI compute means capacity in specific regions can run out, causing sudden outages if multi-region redundancy isn't planned.
• Evidence: Sports venue client whose app crashed during game days due to unexpected demand spikes and lack of multi-region failover.
• Practical implication: Incorporate multi-region discipline and workload redundancy early in your AI infrastructure planning.

Frameworks, Models & Processes

The Data Estate Modernization Journey (Left-to-Right Field)

• How it works: A chronological progression through discovering data, moving to modern cloud systems, establishing governance/master data, and scaling compute capacity.
• Components:
  • Discovery & Inventory
  • Migration & Modernization
  • Governance & Master Data (Gold Layer)
  • Capacity Planning & Multi-Region Compute
• When to use: When assessing an organization's readiness for AI and planning data architecture upgrades.

Examples & Case Studies

A global container leasing company maintained night jobs, joined Excel files by hand, and had undocumented shadow systems.

• Illustrates: Lack of visibility into existing data assets during discovery.
• Lesson: Always perform a thorough inventory first to avoid delayed timelines and uncover hidden legacy dependencies.

An energy infrastructure operator had correct data, but operations and finance defined 'availability' differently, leading to conflicting reports.

• Illustrates: The danger of poor data governance and lack of master definitions.
• Lesson: Data quality checks and codifying definitions into a gold layer are mandatory before building self-service or AI solutions.

A professional sports venue rolled out a new fan experience app right before football season without testing multi-region capacity, resulting in system crashes during peak game traffic.

• Illustrates: Compute scarcity and poor capacity planning during high-demand events.
• Lesson:

Actionable Takeaways

• Immediate:
  • Conduct an inventory of all existing databases and shadow IT systems in your organization.
  • Identify systems running on end-of-life software (e.g., SQL Server 2016).
  • Review cross-departmental metric definitions to check for discrepancies.
• Strategic:
  • Move legacy on-premise databases to modern cloud managed instances to minimize technical friction.
  • Establish a governed 'gold layer' for master data and data stewardship before deploying AI.
  • Design multi-region compute strategies to handle surging AI workload demands without hitting capacity walls.
• Questions to investigate:
  • Do we have a complete inventory of our data assets and their owners?
  • Are our core operational and financial metrics defined identically across all business units?
  • Is our current cloud infrastructure prepared for multi-region AI compute scaling?

Claims Worth Verifying

• Extended support for SQL Server 2016 ended recently (around 2026 timeframe reference). (product lifecycle timeline)
• Datavail manages over 100,000 databases across 400+ managed service customers. (company metric)

Notable Quotes

"You can't modernize what no one owns." "The best thing you can do out of the gate is to build a nice inventory table." "A gold layer is not a person. We all have them... The one person you go to that knows everything. That's our gold layer. And then Robin decides she wants to go build a farm, we'll make Robin, and then we have no gold layer left."

Compressed Summary

• Discovery and inventory are mandatory first steps in data modernization.
• Legacy systems and end-of-life databases create hidden project delays.
• Governed master data (the gold layer) prevents conflicting business metrics.
• AI readiness requires proactive multi-region compute capacity planning.
• Keywords: modernization, governance, inventory, migration, capacity
• Core insight: To successfully prepare a data estate for AI, organizations must progress methodically from discovery and cloud migration to rigorous master data governance and robust multi-region capacity planning.