D39 What's Coming in Tableau: A Leader's Guide to the Best of TC26 with Candi and Kirk Munroe

Summary

Fresh from Tableau Conference 2026, Fi and Sarah sit down with Kirk and Candi Munroe - co-founders of Paint with Data, a Tableau consultancy and Salesforce partner based in Canada - to unpack what's actually coming for Tableau and what it means for the people making decisions about their data teams.

Kirk brings 25 years in business analytics, a newly minted Tableau Visionary title, as well as a Tableau Ambassador and a book on data modelling in Tableau. Candi is a Visual Analytics specialist, four-time Tableau User Group Ambassador, and heads up the Canada Tableau User Group. Between them, they've seen every era of Tableau - and they're genuinely excited about this one.

The conversation covers Tableau Solve and its write-back and forecasting potential, the Viz layers that will let you layer multiple data sources on a single chart axis, composable data sources and why they're the missing piece in Tableau's semantic layer story, agentic analytics and what it actually takes to make natural language querying work in practice, Tableau MCP - including a cautionary tale about a CPM field that cost 58 cents a query - and what the Devs on Stage roadmap means for the data leaders who need to decide what comes next for their teams.

Takeaways

Tableau Solve productionises record-level triage and forecasting - and write-back is finally coming, with Kirk calling it "way, way, way overdue"

Chart Layers will allow multiple data sources to sit on the same XY axis without any joining, opening up time series comparisons and complex formatting that previously required workarounds

Tableau Server is getting genuine love - including LLM features - signalling that Tableau sees it as a long-term platform, not a migration waiting to happen

Composable data sources let analysts extend a published data source on the fly without touching the original, and because everything is relationship-based, nothing already built gets broken

Well-governed, well-commented data sources are now a prerequisite for agentic analytics - without them, the AI will guess wrong and cost you money (Kirk's CPM example is a must-hear)

Tableau MCP is already live, easy to set up in Claude Desktop in under three minutes, and keeps row-level security intact - but you need your own personal access token, not a service account

AI-assisted field descriptions get you 90-95% of the way there - the blank box problem is real, and the AI solves it