Original LinkedIn Post
A Discussion with a Former Tableau Expert and My Next Steps for Agentic BI
On April 7, thanks to an introduction by Matthew Miller, I had an in-depth video meeting with David Spezia, a former Tableau expert who recently joined Anthropic. This conversation was not just a review of my recent automation tools, but a discussion on the future evolution of enterprise internal data products.
We discussed the dilemmas facing the current BI ecosystem. Using Tableau as an example, the underlying XML structure is complex and lacks standard documentation. Major tech companies also tend to focus only on data loops within their own ecosystems. Meanwhile, traditional Data Catalogs only solve the "where is the data" problem but cannot directly answer business questions, leaving users stuck with the "blank canvas" problem.
David shared his vision for OSI (Open Semantics Interchange). This is an excellent architecture concept: establishing a universal semantic standard that allows different AI Agents to read and understand data with business context in a standardized way, rather than just executing SQL. Notably, his idea of using worksheets directly as data sources to establish business standards at the base level is something he proposed 10 years ago. Now, with large language models, this architecture finally has the potential to be realized.
During the meeting, I demonstrated my cwprep and cwtwb projects, which were developed entirely with AI assistance. By having the LLM reverse-engineer the undocumented XML and combining it with the MCP (Model Context Protocol), I have built the backend pipeline for generating Tableau dashboards using natural language. However, as we agreed during our talk: MCP is a great backend communication standard for Agents, but the current CLI (Command Line Interface) is not friendly for regular users. As a data product developer, I will not stop at the code level. My next step is to build a complete Web App (UI frontend) for these tools. This will encapsulate the complex Agent interactions and protocols, lower the barrier to entry, and allow more people to use agile data workflows. Thank you to Matthew Miller for making this conversation happen. I wish David all the best in his new role at Anthropic and look forward to his future breakthroughs in enterprise white-collar process automation. I also expect to have more communications with him and other industry peers in the future to jointly drive the implementation of Agentic BI.
Lately, I have been busy with work and interviews. I am also designing a monetization model for my projects to support continuous iteration. In short, I am on my way.
Southard Jones Elif Tutuk Matthew Miller Adam Mico Allan Folting Roy Raviv Would love your thoughts!
#Tableau #DataVisualization #ArtificialIntelligence #BusinessIntelligence #OpenSource #DataAnalytics #MCP #AIAgents #DataEngineering #datafam #gemini #google #ai #bi #agent #cwtwb #codex #chatgpt #salesforce #datacooper #openclaw #qwen