Designing Graphy's new hybrid AI interface

How we integrated AI into Graphy for seamless data prep and meaningful insights

Over the past few years, the user experience landscape has undergone a seismic shift, one driven by the rapid advancement of artificial intelligence. At Graphy, we see AI not as a buzzword but as a practical way to help users uncover and communicate insights from their data, without sacrificing clarity or ease of use.

This is the story of how we introduced AI into Graphy to tackle the toughest challenge in data workflows: seamless data preparation and meaningful insights, while still preserving the art of storytelling and the slick UX our users want.

The challenge: data cleaning and UI overload

As our user base (150k!!) and feature set grew, so did the complexity of the user interface. Power users appreciated the granular control, but new or occasional users often felt overwhelmed by the sheer volume of settings to toggle. Meanwhile, another pressing issue loomed: data cleaning. Before users could even get to those carefully designed editor panels, they had to manually wrangle their data into just the right shape—renaming columns, restructuring rows, or eliminating missing values. It was tedious, and for many, a barrier to fully utilizing our tool.

Why AI became our next step

In October of last year, we reframed our core question: What if Graphy could automatically parse, clean, and interpret data, no matter how unstructured it was? That is when AI became central to our vision. We saw an opportunity to move from a beautiful chart-customization tool to a complete data companion.

This approach is not just about convenience; it also unlocks insights (and the stories behind them) for users who might be deterred by the tedious nature of data preparation. By adopting AI, we lowered barriers to entry, enabled faster experimentation, and helped teams quickly turn raw data into clarity.

A familiar but innovative interface

When designing our new input screen, we took cues from other AI-driven tools and purposefully chose a layout that feels approachable. Users can drop in any dataset or type a prompt, and the system does the rest. Under the hood, a powerful pipeline handles everything from CSVs with typos to half-structured spreadsheets.

Introducing our hybrid interface

Once a user selects or refines a chart, we introduce what we believe is a best-of-breed hybrid model: a conversational co-pilot positioned next to traditional editing panels. Even though I personally love the potential of conversational AI, I also wanted to respect the speed and familiarity of direct UI controls for final tweaks. Sometimes it’s just more efficient to click a button than type out “Switch to a heatmap” so we want to give our users the option to do both. This balance between AI-driven suggestions and hands-on customization helps users feel both guided and in control at every step of the visualization process.

Building trust through the recipe

One feature I’m especially proud of is what we’ve coined ‘the recipe’ – a clear, step-by-step account of how the system transformed raw data into the final visualization.

Each entry in the recipe includes the same icons and terminology found in the editing panels, effectively bridging the gap between AI-initiated suggestions and user-driven customization. This transparency:

• Instills confidence by showing why certain recommendations were made,

• Demystifies tasks like data joins or pivoting columns, and

• Empowers users to replicate or fine-tune those steps themselves.

User testing consistently shows that the recipe is one of the most appreciated features; bolstering trust and providing a valuable learning tool for those who might be newer to data visualization.

Standing apart in a growing field

Many tools are going all in on conversational interfaces. While that approach can be powerful, I believe a balanced blend of AI-driven suggestions and direct user interaction is more sustainable. Manually clicking a few toggles or sliders is sometimes the quickest way to refine a chart, and we want to keep that fast functionality front and center.

By integrating AI for complex data prep and pairing it with a clear, robust editor, Graphy stands out in a space where others might be losing themselves in endless chat prompts. Our users appreciate being able to do the “heavy lifting” through AI while retaining the speed and familiarity of traditional UI elements.

The road ahead

Design, like data, is always evolving. We plan to continuously refine our AI models, improve UI interactions, and stay closely attuned to user feedback. Every insight we gain helps us iterate, ensuring Graphy remains both powerful and approachable.

For now, we’re excited to see how an AI co-pilot approach transforms workflows. By letting users skip tedious data cleaning steps and then letting them decide whether to rely on AI or manual controls for visualization tweaks, we hope to make data storytelling more accessible, intuitive, and ultimately more valuable for everyone.

We're already mocking up ideas to make the UI cleaner and make a stronger connection between the editing controls and chat interface.