What I learned asking people at a data viz conference how they’re really using AI
Ask someone at a data conference how they’re using AI in their day-to-day work, and you’ll get a strange kind of answer: confident vocabulary, vague substance. “We’re exploring agentic workflows.” “We’re looking at MCP.” “It depends on the use case.” That was my experience at a major data visualization conference this year — three days of hallway conversations, keynotes, and sessions, with the same handful of terms surfacing again and again: MCP, headless analytics, conversational analytics. Confident-sounding words, and almost nobody could tell me what they actually meant for their own work.
The honest answer is that the tooling landscape is fragmented and nobody has quite settled on the vocabulary for what they’re actually doing yet. A lot of the product promises center on tools that require several layers of access just to get started, because organizations are increasingly cautious about how much data employees feed into agentic systems and what those systems are then allowed to do. The tools getting the most stage time are often the ones most practitioners will never touch in their day-to-day work.
So I decided to write the article I wish someone had handed me at that conference: a practical guide to how analysts are actually using AI in data visualization today, with the tools most of us already have.
Start with your budget reality
There are two very different conversations here depending on where your organization sits.
If your organization has access to premium AI-native analytics platforms, conversational analytics can build visualizations for you end to end. You describe what you want, and it builds it. Most people in data viz, whether practitioners or enthusiasts, won’t have access to these tools, and the ROI of buying them at an individual level isn’t clear right now.
So here are three ways I use AI in data visualization as someone with a Gemini subscription and Tableau Desktop, which is probably closer to most people’s reality than a Tableau Cloud enterprise license.
Three workflows I have tested and use regularly with AI and LLMs
Workflow 1: Google Sheets + Gemini
If Google Sheets is your primary environment, Gemini is a practical starting point. It is integrated directly, requires no coding knowledge, and can take you from raw data to a working visualization through conversation. The barrier to entry is low, and the setup is minimal.
Say you have a dataset with sales data where each deal has been booked at a certain amount on a certain date and you want to see how the total booking value has progressed over time. Here is the conversation with Gemini that gave me a perfectly usable visualization. You may need to follow up on annotations, legends, and formatting, but it removes the grunt work of manually pulling fields into boxes and configuring a chart from scratch.
If you have spent any amount of time manually building charts in spreadsheet tools, you will know how much cognitive overhead goes into what should be a simple task. Gemini takes that friction away and lets you stay focused on what the data is actually telling you.
Workflow 2: Python Notebooks + Gemini
This is the one I have stress-tested the most, approximately six weeks of real use, and it is the most capable of the three.
The Gemini add-on in Colab notebooks lets you build and refine visualizations through conversation. I used it to build a line chart for forecast values and iterated entirely through conversation — adding annotations, adjusting colors, changing layout — without touching the code manually. The output is updated with each exchange.
I have working knowledge of Python, so I could go into the code directly when I wanted to. The conversational layer holds up on its own as well. You need to know what you want to show. The coding is optional.
Workflow 3: Tableau Desktop + Any AI tool
Tableau Desktop has no built-in AI capabilities without an additional license. Most practitioners do not have that license. Here is the workaround that has genuinely changed how I build.
Open ChatGPT, describe what you want to build, ask it to generate dummy data that matches your schema, explain your fields and the story you are trying to tell, and ask it to recommend the right visualization type. Then ask it to walk you through building it in Tableau step by step.
Go one more step if you are not sure how that should look.
Tableau is a deep tool. A lot of its most powerful features are hidden, and it can take years to discover the most efficient way to do things. Prompt engineering through ChatGPT has helped me build complex visualizations faster, surface features I would not have found through trial and error, and explore the tool in a much more direct way. It functions as an on-demand Tableau expert who never runs out of patience for follow-up questions.
The underlying principle
Across all three workflows the pattern is the same. The value of AI in data visualization right now is iteration speed. The ability to describe what you want, get something back, react to it, and refine it in conversation is what compresses the time between idea and output. That is the actual unlock, regardless of which tool you are using.
You do not need a premium license to access that. You need to describe your intent precisely and treat the conversation as a working session.
What this means for you
Here is the question worth sitting with: as AI tooling in data visualization matures, what kind of analyst does it produce?
I think it creates two distinct profiles, shaped almost entirely by how someone relates to the technology.
The first is the conversational analyst. This is someone working with agentic, high-capability AI tools. She has become fluent at directing AI, knowing what to ask for and when to let the tool run. She offloads the build. Her leverage sits upstream — translating insights, communicating the value of the visualization to stakeholders, connecting the output to the business question. The AI builds. She interprets and communicates.
The second is the architect analyst. This is someone who uses AI as a working partner. She has done the harder thinking before the AI touches anything: articulated the requirements, pushed back when the output was off, worked through the problem iteratively. The resulting visualization is something the AI helped construct, but she is the architect behind every decision. She understands why each design choice was made because she made it, working alongside the AI.
Is one better than the other? I think there are people in both camps. The tools you have access to, the work you do, and the environment you are in will pull you toward one or the other naturally. Nothing is better than the other. What matters is that you are intentional about how you are showing up in that relationship with the technology. It is still the human decision and intervention, whether at the architecture level or the communication level, that will shape how AI gets adopted in data viz.
That’s also, maybe, the missing vocabulary from those hallway conversations. Not a tool name or a feature list, but a clearer sense of which kind of analyst the tool in front of you is asking you to become.
Rituparna Das
Rituparna is a Senior Business Analyst leading sales analytics and dashboard development at a data management company, with six years of experience across data visualization, business intelligence, and stakeholder communication. She mentors analysts transitioning into the field through the University of Washington and the Data Visualization Society. Originally from India and now based in Seattle, she writes on Medium and can be found on LinkedIn.
