D40 Why AI Fails at Enterprise Data and What Tableau Is Doing About It

Episode Summary Dive into the evolving world of data, AI, and analytics with Will Pitzler, Director of Product Management at Tableau. This episode explores how Tableau is breaking down barriers to make insights accessible across platforms, the importance of governance, and what it actually takes to become a data-driven organisation.   In this episode, you'll learn about: ● Why leading LLMs scored just 6% on a Yale benchmark testing real enterprise databases and what that means for your AI strategy ● The role of composable data sources, the most requested feature in Tableau's history, and what they unlock for data teams ● How Tableau insights are now accessible directly inside Google Sheets, PowerPoint, Google Slides, and Word ● The uncomfortable questions around PII, data sovereignty, and governance when using MCP and third-party AI tools ● The Open Semantic Interchange initiative, what it is, why it matters, and how far away a real standard actually is ● What data leaders should actually do before their next AI project kicks off Timestamps: 00:00 - Introduction and welcome 01:22 - Will's background and role at Tableau 03:31 - Why TC26 felt different, the return to the practitioner 06:00 - The developer spectrum and Tableau's broad user base 06:30 - Tableau through the Salesforce acquisition, what's changed and what hasn't 09:42 - TC26 comes to Sydney, bringing the insights to local customers 10:32 - The Yale Spider 2.0 benchmark and the 6% problem 13:50 - Why context is everything for LLMs in enterprise environments 15:18 - PII, data sovereignty, and the governance gap in AI and MCP 18:08 - What data leaders are actually telling Will on the ground 19:39 - Open Semantic Interchange, the industry's attempt at a common standard 21:17 - Two schools of thought on how Tableau handles semantic layers 23:23 - Delegated semantics, Tableau's interim approach 26:06 - How the data market has gone in circles, monolithic to modern and back again 28:30 - Composable data sources, the most requested feature in Tableau's history 33:10 - Governance and ways of working as teams move faster 35:25 - The last mile problem, insights shouldn't live only inside Tableau 38:44 - Staying in the flow, self-service where people actually work 39:32 - The tension between AI text outputs and data visualisation 42:02 - Demo begins, third party integrations overview 44:38 - Demo: Tableau inside PowerPoint 46:09 - The timestamp feature and refreshing slides on demand 47:19 - Salesforce internal use case, automating operational reporting 50:40 - Why live dashboards weren't the answer 52:06 - Demo: Tableau inside Google Sheets 55:27 - Using published data sources in Google Sheets 56:58 - Licensing and permissions 57:22 - Pushing data back into Tableau from Google Sheets 59:48 - The Henry Ford problem, what customers say vs. what they need 1:02:28 - Closing question: What should a data leader actually do? 1:03:16 - Where to find Will 1:03:38 - Fi and Sarah's closing takeaways   Resources & Links: Connect with Will on LinkedIn https://www.linkedin.com/in/will-pitzler-603b915b/ Spider 2.0 Benchmark, Yale https://spider2-sql.github.io/ Tableau Add-on for Google Workspace https://www.tableau.com/blog/improve-collaboration-tableau-google-workspace Tableau App for Microsoft 365 https://www.tableau.com/blog/meet-tableau-app-for-microsoft-365-word-powerpoint-teams