BI and generative AI still don’t mix well
Generative AI is starting to creep into BI but not in the way we need it to.
A LinkedIn post from Roman at GoodData caught my attention this week—they're launching functionality to import sketched dashboard designs. Most dashboards suck, but sketching them first is a great step to making them suck less!
BI tools are evolving, but most still cater to data experts, not decision-makers.
Generative AI can fundamentally change how people interact with data—removing technical barriers to analysis.
Sketch-based dashboarding (like GoodData's new feature) helps, but only incrementally and only for power users.
The real BI revolution will come from tools that prioritize decision-making over data exploration.
We should build tools like Loveable.dev—intuitive, empowering, and designed for the non-technical majority—not just better tools for the 1% of expert users.
What we need is not an “AI data team,” but an AI-powered co-designer of decisions.
My own analytical work has been completely transformed by AI. I no longer write SQL queries; I simply chat with a custom GPT tuned to our database structures. This shift hasn't merely saved time—it's fundamentally expanded what I can analyze. Creating moving averages with different weights—essential for decision-making but technically challenging—is now trivial. I'm generating better analyses, decisions and products because the technical barriers have disappeared.
This experience illustrates a fundamental truth: the BI industry has been building increasingly elaborate tools that still fail to deliver genuine insight to decision-makers. The problem isn't technical sophistication—it's accessibility and relevance. Maybe it is just me, but I don’t feel like that shift is happening fast enough. If I have to build my own tools that deliver 10x on existing experiences, something is off in the business of business intelligence.
The Hidden Problem with Business Intelligence
Our dashboards have become like elaborate paper maps—technically impressive but fundamentally limited in how they serve their purpose. While we've perfected the ability to display data, we've largely failed at making that data truly useful for decisions.
As they say, when you're holding a hammer, everything looks like a nail. The BI industry has been wielding SQL and visualization libraries for so long that we've forgotten our actual purpose: enabling better decisions through data.
I’m not sure that industry titans realize that LLMs and generative AI in general are the key to getting back, but for me they already are.
Early Signs of Transformation
The BI industry is beginning to acknowledge this reality. GoodData's move to support importing sketched dashboard designs creates leverage for better practices—but it's just the first tentative step toward a more profound transformation.
My last company, Arch.dev has pivoted into a very similar direction, providing your “AI data team” to help me “Focus on decisions.” Thing is, that’s not gonna happen when I “Ask,. Answer. Act.”
Build the Lovable, Not the Cursor
Loveable.dev and Cursor.ai are both two different classes of super star AI products. While Cursor AI went from 1 million ARR to 100 million within 12 months (fastest ever), Loveable went to 10 million ARR within 2 months.
But as products, they couldn’t be more different. While Cursor enables the existing coders to become 10x or 100x coders, giving them all the freedom they could have (by providing a custom IDE), loveable takes all that freedom away to enable EVERYONE (but coders!) to deliver great software products.
I’m saying the big challenge of BI is to build the loveable.dev, not the cursor. And it sure looks like all the “this is for everyone” kind of features are only marketing crap, while all the actual working features are only aimed at people already data-capable.
Why is that the case? Because data teams are support teams, supporting other value driving parts inside the company. Whereas software teams sit directly inside the value stream, and thus profit directly from cursor, and every improvement there directly translates into more value for the company.
GoodData's implementation resembles a cursor for data engineers and analysts. It helps the 1% of people who already create dashboards work marginally faster. But what about the other 99% who should be working with data but can't due to technical barriers? It imports sketched dashboards. That is great but that only helps people who already sketch dashboards (a super tiny margin), it doesn’t actually help everyone. How would that look like? If I were to empower everyone to use the best practice of sketching, that would mean I’d build a good interface that starts with my business problem, then plans out things (including the sketching of dashboard!) and helps me to make that great.
The BI industry is still focused on building better cursors when what we need are lovable products that democratize data analysis entirely.
Arch promises to be "the AI data team for every business" with their "Ask, Answer, Act" approach. But as a decision-maker, I don't want an AI data team. I want an AI version of a product designer married to a sales person and data analyst. I need a system that helps me think through the right decisions, identify meaningful metrics, and design experiments worth running.
The analogy that comes to mind is photography. Professional photographers didn't clamor for smartphone cameras. The transformation came when the industry stopped focusing exclusively on professional needs and instead created tools that made photography accessible to everyone. The result wasn't just more photographers—it was entirely new forms of visual communication.
Similarly, the BI revolution won't come from making analysts more efficient. It will come from making analysis accessible to everyone who makes decisions.
The companies that understand this shift won't build faster horses for the existing 1%—they'll build cars for the other 99%. They'll focus less on dashboards and more on decisions, less on data processing and more on insight generation.