Did we choose the wrong book title?
"Generative data apps" might be the new lingo, but the principles remain as important as ever.
Hi all,
Andy here. This time:
Are they even called dashboards anymore?
Details about Chart Chat 71 (“AI: Sabotage or Supercharge”) on June 18th. Register here.
Are they even called dashboards?
“Dashboards are dead,” rang through our heads as we wrote Dashboards That Deliver.
Thoughtspot had banged that drum for years; business users were fatigued by too many disappointing dashboard projects; and during our writing process, Generative AI swept in, promising to sweep away the traditional dashboard entirely.
So, did we write a book for a previous era?
No. But we might have given it the wrong title. Whatever its title, the principles in the book endure. I’d even like to make a case that the principles are more important than before.
Talkin’ bout the Generation
Tableau and Power BI are fading from their peak, and the industry is aggressively hyping the next big thing: “generative apps.” Have you seen what’s been announced and released recently?
Microsoft just launched Fabric Apps for pixel-perfect data analytical displays.
Hex has “generative data apps” (fancy a fairytale-style storybook data display? Here you go).
Golden has launched with conversationally generated stories and dashboards.
Veezoo has a new “canvas” feature.
Lightdash is going all-in on “data apps”.
The narrative is clear, and it sounds fantastic: “The traditional dashboard is history. The AI-generated data app is here: if you can imagine it, you can build it.”
Alas, there are hard truths to consider.
Slop will tear us apart
If anyone can build anything with a single prompt, Excel and Dashboard hell are nothing compared to the AI Slop Mountain we are about to create.
When you lower the barrier to creation to zero, the requirement for discipline skyrockets.
AI agents are, at best, intern-level analysts. They’ll tirelessly answer all your questions with enthusiasm and misplaced confidence. But they don’t know your business context, they don’t understand human nuance, and certainly don’t have a critical eye for design.
This is exactly where the principles and frameworks we created in Dashboards That Deliver are needed.
What’s skill go to do with it?
There is good news: you can build excellent generative data apps. To do so, users need a framework built on a strong foundation. A good dashboard isn’t built from a single sentence. A good generative app can’t be built from a single prompt.
If you don’t work with users, you won’t build them an app that works for them. Discovery can be assisted very well by AI tools, but you can’t replace the hard work of working with them to understand their needs.
A skilled “design eye” during Prototyping is vital for a successful user experience. LLMs have generic design knowledge but it’s not enough to fit your organisation’s context.
Corporate style guides exist for a reason: if anyone can make anything, what kind of cognitive overwhelm will we end up with when every single app has its own unique (and probably very poor) interface and design style?
Adoption. “Hey, you! I built this. Use it.” That approach didn’t work with dashboards. It won’t work with AI apps either. Adoption is a human process that requires time and planning. And speaking to humans a lot.
That’s why the framework in Dashboards That Deliver works for generative apps too.
I don’t give a monkey what you call the thing you’re building (heck, I wrote one of the chapters on that topic). What matters is that a successful data communication asset must be built on a successful foundation of analytical and design knowledge, aligned with the context of your own organisation's needs.
That foundation is not yet strong enough in any AI tools, despite what the marketing teams say. Only with a well-informed human at the helm can these tools deliver success beyond marketing. Do check out the book if you are interested.
My friend Mark Bradbourne said to me the other day, “If you want to be a great AI Data Analyst, be a great Data Analyst.” Amen!
Chart Chat and other goings on:
Chart Chat #71 is on June 18: we will be discussing our latest thoughts AI’s place in analytics. (You could also call this our mea culpa episode, following the AI-created issues in our last episode🙂). Click here to register.
Elsewhere, plenty is happening:
I’ll be debating AI and Analytics with Francois Lopitaux, SVP of Product Management at Thoughtspot. Click here to register.
Amanda and I will be hosting and speaking at the Outlier conference, June 24-26.
Looking ahead to September: I’ll be doing a data storytelling workshop at Data Decoded in Manchester, and there is a special event cooking for Big Data London (I’m very excited about it!)
That’s all for now. Take care, and keep building amazing dashboards generative data apps.
Andy