Will AI Kill BI?
Yes it will. But BI will rise again!
AI is going to kill BI, no question about it. Ask my friends Joe, Jean-George and Ole or countless others who see this one clear fact: Artificial intelligence will soon be better - or at least much faster and cheaper - at creating visualizations and dashboards than human beings. I get tagged in these conversations a lot as our little data clique’s resident BI guy. So I’m going to weigh in definitively here.
Yes AI will kill BI. As we know it.
But something new will rise in its place and that thing will still be BI.
Confused? Welcome to Summer 2025.
BI has already died at least once
When people say AI will kill BI, what do they mean? There are two common themes; that AI will kill both the development style and form factor of BI. Consider:
AI is or will soon be so much better and doing BI development tasks that we will no longer need nearly so many BI developers, thus killing BI. This is true.
AI driven experiences like chatbots, AI driven alerts and agents will replace the BI front end completely, thus killing BI. This is also true.
What many people don’t know is that we’ve already gone through at least one total and complete ‘death of BI.’ It was called Tableau.
In the late 2000s, BI had a development style defined by a centralized team of technical professionals with almost no business knowledge operating in a waterfall development lifecycle. And BI had a form factor, the standardized web report paired with the mass distributed PDF.
This development style and form factor was the basis for tens of thousands of careers and multiple 1B+ vendor exits. It was BI.
And that’s completely gone now1. It was replaced with ‘Visual Data Discovery,’ originally a niche subset of BI where data analysts embedded in the business used their technical and business acumen to rapidly answer questions and build visually stunning interactive dashboards, with minimal IT oversight.
If that sounds like BI at your company, that’s because it likely is. ‘Visual Data Discovery’ completely replaced traditional BI and the job of data analyst was born. And yet, we still call it BI.
And that is the paradigm that AI is going to kill very soon.
How will AI kill BI
BI takes an enormous amount of work with middling payoff. I am the BI guy, I can say it. There are three main reasons for this, in order of importance:
It is too hard to get the right data in the right place at the right time to have a consistent positive impact on business outcomes.
It is too time consuming to develop front end visualizations, dashboards and reports to keep up with demands from data customers.
It is too challenging for data customers to learn BI interfaces to answer questions themselves.
AI is rapidly closing these gaps.
BI chatbots are vastly easier to use than drag-and-drop viz tools for 95% of people, even when they aren’t very good, which they mostly aren’t. But keep in mind that every BI chatbot you use is the worst it’s ever going to be. The need to ping a data person for every modestly complex query is rapidly fading.
Likewise for building visualizations and dashboards. AI can already do this, and do it quite well. It struggles with building what the user actually wants vs what they asked for, which is the art of good BI. But even now, the cost-benefit is tilting in favor of allowing AI to build the BI front end for routine requests because it’s simply so much faster and cheaper.
Finally, getting the right data at the right time. This is the toughest part, because it requires coordination across domains, business knowledge and deep technical skill. But even here, AI is making data engineers vastly more productive on its path to eventually replacing them. BI’s role is often to put the finishing touches on data and make it relevant to a specific data audience
When it’s easy to get the right data fast, simple to build the data front end, and trivial to query and iterate as a data consumer without ever calling your BI team, BI as we know it is dead.
BI is dead. Long live BI!
Once AI can do all the core technical tasks of the BI team, what is left? Ironically, it’s the same tasks that separate the good from great BI practices today - being the human face of data while managing the meta tasks relating data to consumer.
Being the human face of data is simple and devilishly complex at the same time. Because the BI team traditionally sits and the end of the data pipeline, they are the ones in contact with business reality. The best thing a BI team can do is not to clear adhoc queries requests, it’s to represent the business consumers and be their advocates in the data gristmill.
This role is not going away until data consumers get radically better at articulating their needs, goals, desires and fears, or until AI’s are able to intuit them from a combination of institutional knowledge and reading non-verbal and visual cues. The automation and ease coming to BI will only amplify the need for someone to combine technical mastery, business and domain knowledge and human understanding and empathy. This is the ‘soft stuff’ that will be the key to BI in the future.
Then there’s managing the ‘meta tasks’ of BI. This is the hard, technical stuff that will remain when AI is answering queries and building visualizations. Examples of meta tasks in BI include:
Creating, customizing and maintaining the front end data assistants, chatbots and agents that end users interact with. These are complex systems that require a ton of care and must be infused with highly relevant data context.
Curating, securing and enriching the metrics library available to AI. Even if AI is defining all the metrics, someone has to maintain the metadata and business context for those metrics.
Contributing to domain and enterprise knowledge graphs, metadata catalogs, context libraries, etc. Especially translating business speak into technical instructions for machines.
Defining and maintaining libraries of ‘approved’ front end assets. Why let the AI generate the same chart 1000x a day when it can just take it off the shelf?
Building mission critical data assets alongside AI peers - the dashboard the CEO relies on cannot be 100% an automated system.
Doing deep data research and answering the most critical, business altering questions. Yes, with an AI assist.
Other technical tasks we can’t imagine yet.
These BI meta tasks are vastly more valuable than building dashboards, but building dashboards has soaked our BI bandwidth. In fact the top complaint of BI teams is something like this:
‘We spend so much time building dashboards and answering ad-hoc queries on slack that we can’t focus on solving the business problems that really matter!’
The good news for BI practitioners is that AI will relieve you of this problem quite quickly. The bad news for many is that having technical skills in a BI suite is rapidly losing it’s economic edge.
What’s next for practitioners and platforms
This is the transition to manage then. As a BI practitioner, skill up rapidly in AI assisted data development. Learn about knowledge graphs, ontologies, context windows, semantic layers. Don’t assume Tableau and Power BI skills will keep the family fed. And most of all, get laser focused on understanding the people who consume your data and what drives them.
As far as the BI platforms themselves, they are rapidly evolving AI capabilities. If history is any guide, some of the current leaders will not adapt fast enough and will fade from prominence, while startups or very innovative mid-level firms will push the industry rapidly forward.
But what does the BI platform of the future actually look like? That’s a question for a future post. Subscribe so you don’t miss it!’
Catch up on the best of Super Data Brothers
Season 3 of Super Data Brothers has wrapped for the summer, and we are knee deep in planning Season 4 kicking off in September. Have a topic or guest you’re dying to see? Hit me up here.
Season 3 is available on demand on YouTube:
Revolutionize your data mindset with Malcolm Hawker - How to ditch loser mentality and become the beset data leader
How to get a data job in 2025 with Aaron Wilkerson - the lowdown on getting paid in 2025.
Gartner BI MQ 2025 revealed - who moved up, who moved down and why!
Will AI doom or save the data industry with Joe Reis - The legend himself tells us where it’s all heading.
AI will steal your soul, not just your job with Ramona Truta - How to use AI systems in an ethical, uplifting way
Data is the new bullshit with Scott Taylor - Get boardroom buy in and do work that matters!
Databricks AI/BI tool review - Full breakdown of AI/BI and Genie’s strengths and gaps
Making data more human with Tiankai Feng - Design and empathy in data work
Fight health insurance denials using AI with Kolden Karau - Healthcare AI to save your life
Surviving the data engineer job crunch with Eevamaija Virtanen - Career strategy and good vibes
A BIG thank you to the Season 3 sponsor, GoodData!
Yes there are actually still tens of thousands of people doing this as a career and the tooling still exists with tens of thousands of customers. ‘Completely gone’ is hyperbole. But it is completely gone from our collective data imagination.
Love this quote, generalized so it applies wider:
“Don’t assume <insert tech platform here> skills will keep the family fed. And most of all, get laser focused on understanding the people who consume your data and what drives them.”