AI can’t see what you’re missing: the art of intelligent data visualization.
Graham WattsPublished 2025-08-13Summary: Modern data tools include many visualization types. However, without guidance, choosing the right visualization for the job can be a daunting task. This is where well-designed BI apps come into play. Informed by the deeper meaning and context provided by metrics, AI and a semantic layer, these analytics tools make choosing from this sea of options easy for everyone.
We are inundated with data visualization and framework tool choices, making it possible for anyone to create visualizations from any data set. This abundance of BI tool options has led many industry experts to declare data visualization a solved problem. In fact, many of the most innovative data tools and platforms to recently hit the market don’t even have a visualization layer, and are instead driven by AI-based conversational systems. The current theory is that if the data is well-modelled, with semantics, it should be easy for anyone to use any visualization library to consume it.
However, this simplistic approach ignores a key factor. Data visualization is more than just the ability to generate a visualization, it’s also about using the right visualization at the right time, and consistently. Deciding which chart type to use for a given data analysis can be difficult for many users. That’s where modern BI tools come in. Recent technological advances in semantic layers, metrics and AI have created the opportunity for these tools to be smarter at determining how data is consumed and visualized.
Your brain craves pictures, not numbers
Good visualizations are powerful because they instantly make data easier to understand. Compared to the effort and expertise required to sift through raw values in tables, with the right visualization you can recognize patterns and gain insights at a glance. Data encoded in a visualization leverages the most powerful pattern-matching parts of the human brain, the retina and the visual cortex. For example, it’s easy to see a point on a line chart that’s out of place, but it can be difficult to pick out that same point from a table of raw values. It’s no wonder analysts and business users rely heavily on data visualizations for interpreting and sharing information.
Surprisingly, data visualization hasn't always been so appreciated. One of the earliest pioneers in statistical graphical methods was William Playfair (1759 - 1823). Often ridiculed by his peers, Playfair experimented with data visualization as a way to make numerical analyses easier for the general public to understand. The widespread use today of the charts he pioneered (including line, bar and pie charts) are a testament to his ingenuity and an indication of how effective visualizations are at helping people comprehend data.
More than 200 years ago, Playfair introduced visualizations as a novel analytics concept. Fast forward to today where data visualizations have become the standard interface between consumers and their data. Looks like he was onto something!
Chart choice is everything
Today’s consumer has a vast array of visualization types to choose from, each with its own unique purpose.The right visualization will help consumers get answers from their data. The wrong visualization may mislead them into invalid conclusions and, in the worst-case scenario, faulty business decisions.
Unfortunately, most traditional BI tools don’t understand your data and what you’re trying to learn from it and, as a result, can’t really help you make good choices. They’re rich in capability but poor in guidance. The challenge faced by BI tools isn’t a lack of knowledge about the purpose of each visualization type – these concepts are well-known. Where these tools often fall short is in understanding what you want or need from your data.
Instead of interpreting your needs and offering you the ideal chart type, most BI tools provide a large set of visual choices, allowing you to play with the options and see what you can uncover. If you understand the benefits of each visualization type and know your data really well, you have everything you need to quickly find insights. However, if the data or tooling is unfamiliar to you, finding answers can be an exhausting exercise of trial and error.
Thankfully, modern BI tools have the potential to better understand your analytics goals and the meaning in your data, leading to more helpful responses and more effective data visualizations.
Innovative tools for smarter chart choices
These days there aren’t many new visualization types being added to BI tools. Instead, these tools are innovating to help users quickly get to the most effective visualizations for their data. Two advancements in particular have made this possible:
Metrics and semantics: The recent addition of metrics and semantic layers into the modern data stack brings a richer, formalized description of the data. This deeper understanding enables visualization tools to make good guesses about meaningful patterns and insights that can be surfaced. The result? Modern BI tools, supported by metrics and semantic layers, can pick visualization types that best communicate insights to the consumer, something not possible when the data model has only a basic description of the data.
Natural language interfaces and AI: Natural language interfaces, when implemented effectively and backed by a system that understands that language, provide a more intuitive way to interact with data. This interaction usually involves a user entering a description on what they want and why. Understanding what a user is looking for from the data helps the consumption tool pick an appropriate visualization, or at least make a very good guess.
Why chat interfaces don’t always get it right
AI applications are quickly becoming the tool of choice for business users looking for answers from their data. Charts, based on specific requests by users, can be generated via an AI-powered chat interface. Unfortunately, the resulting chart is often not what the user needs to successfully interpret their data.
LLM-based AI models are optimized to return answers in natural language, or, in some cases, images or other artifacts. These answers are generated by closely imitating learned responses to similar questions in their training data, and are not based on an understanding of the user’s question.
It’s important to note that charts may look similar and yet tell different stories about the data. Generating a chart that’s similar to other charts often results in something that’s not needed or doesn’t even make sense. To create the right chart for effective analysis, the system must understand the user’s question and apply nuanced reasoning, and do that consistently, which is beyond the current capabilities of LLMs.
Rule-based systems and AI for informative, consistent visualizations
Even with advancements in AI, expert-authored charts are key in creating a consistent, intuitive way to consume business data. Curated dashboards are enduring artifacts that decision makers can continually rely on to track the health of their business. Presented clearly and consistently, dashboards and their charts help business users quickly spot and act on important changes in the data.
When combined with AI, these systems impose a set of logical guardrails that ensure AI-generated charts meet the user’s intention. Instead of generating a visualization directly from the data, the AI selects the data and invokes the visualization system with options that describe the user's question. A well-designed system can then apply standard visualization rules to create a chart that fits the user’s question. Since the generation of the visualization is rule-based, it will be consistent and unambiguous every time.
How metrics power intelligent visualizations
Metrics are especially well-suited at helping data tools effectively visualize data. A metric contains not only semantically-described data and a well-known structure, it also has an attached business meaning. Additionally, each metric has a set of expected analyses it would typically be used for. These inherent qualities help define which charts are best for visualizing the data, and make it easier to predict the kinds of questions users might ask, particularly when compared to a more generic data model.
Metrics should, and often do, provide insight as soon as they are loaded up in a metric consumption tool. These applications can associate specific metrics with expected visualizations, immediately presenting something useful. This automatic visualization selection may be exactly what the user is looking for or it might be a good starting point for data exploration. Consumers should still have the option to manually choose from a rich set of visualization types but the first default metric visualization they see should be helpful and easily-understood. It should also reinforce the business meaning that’s defined in the metric.
Metrics also lend themselves to AI and natural language interfaces to create and interact with visualizations. The same richness of metadata and definition that helps data tools consume metrics effectively provides the context needed for AI systems to interpret the data and match it to the user's intention and goals. Based on the rich description of data in the metric, a typical AI system can recognize which visualization type and apply a metric that will answer each user’s question.
You can have all this today
Klipfolio PowerMetrics makes it easy to consume and visualize metrics. Its AI-backed natural language interface further enhances the metric visualization experience. With PowerMetrics AI, users can ask questions of their metrics and get meaningful answers in the form of a visualization. Responses are based on the rich metadata built into each metric and the product's knowledge of visualization types and their uses.
If you haven’t already, try the AI-enabled natural language interface in PowerMetrics. It’s an effortless, fast way to get answers from and discover insights in your data.