If you have been on the internet recently, you might have noticed that dashboards are dead because now we have AI (Artificial Intelligence). This message is iterated by clueless AInfluencers, who have no idea how data works, but are hyped by AI a lot. We, people in data, still consider AI as a solution looking for a problem (as of 2026 Q1), it helps us to vibe-code some python scripts, but we still do not ✰trust✰ when AI chatbots tell us numbers. The situation is changing quite rapidly, so will AI cause the advent of dashboards one day?

This piece is an opinion article based on my experience in various roles in the business – both on passive user and active creator sides, working with people, systems and, of course, AI.

The “threat” of AI

First of all, let’s define what AI can do in theory, how can it make dashboards dead and how it will steal our jobs.

Will it happen at some point is the same as asking will we ever have McDonald’s on Pluto. Well, in far future that is not impossible.

But let’s go back to AI. It can do two things.

The first of them is automating the manual work in creating data products so well, that we will need just one person to orchestrate a horde of AI agents, just like a Project Manager orchestrates a horde of data engineers right now. We are talking about gathering data from systems, cleaning it, ensuring data quality, providing data visualisations in forms of reports, presentations, dashboards and of course extracting feedback from end-users about how exactly those dashboards should work and what information should they provide.

As you may see, the farthest thing from happening is not the data cleaning, rather this extracting feedback from end-users. What do they want? We humans still struggle with that, but hey, one day we will have McDonald’s in Pluto, so let’s say it’s also feasible.

In the end, all the known infrastructure and the way we interact with data is still there, just instead of people all this work is done by AI.

The other thing that could happen is a more radical change. Imagine, that instead of building your data pipelines and dashboards AI connects directly to all the systems where data lies, it can understand that information and spew out insights based on any of your requests.

It will be like an all-knowing chatbot from sci-fi movies, but still a chatbot. “Give me sales by region, and indicate where we have a decline.” Here you go, explore this interactive exploration interface I created for you. “At the beginning of every week, provide me an overview of store performance and show which need my attention and why.” Sure, an email with a summary will be sent to you every Monday. (I believe emails will still be a thing in XXIII century)

There will be a skill in prompt engineering required, but at some point we will be able to handle it like we handle google queries now.

That second change is more radical because instead of replacing humans, it will replace the whole infrastructure (including traditional dashboards) and the whole paradigm of how data in business works.

Now we know what to be afraid of because in both cases our skills as data analysts and engineers will be harder to sell.

Will AI take over the Business Intelligence?

Business Intelligence is a set of ways data can reach business people, how can they interact with the data, and how these interactions influence their decisions. This is not an official definition, but I believe it summarizes the core processes quite well.

A small doubt that most radical scenarios will ever happen is still within me because of two reasons.

The first one is that this shiny Big Data, Data Science, Business Intelligence, Dashboarding, Data Warehousing stuff did not manage to kill Excel. There are reasons I’m leaving for maybe another article, but Excel is still there and is not going to leave soon. There are fundamental needs that this tiny piece of software is covering, which you cannot get from your dashboard by design. Just click Alt+MP and you will immediately see which cells are used to calculate your result. You are assured. Now try this with Power BI or Purview or whatever we have. You have to ✰trust✰ the system blindly.

The same way low-code-no-code revolution did not kill the code. There is something magical in understanding how things actually work.

Second one is related to the fact, that new technologies are still not changing how human brain works. I bought a Samsung tablet to make digital notes. Handwritten notes. All this technology progress just to come back to the point, where I can write again – typing is simply less natural than handwriting. Now, every millennial hipster owns a reMarkable – a tabled specifically tailored to handwriting.

Which means new technologies might kill old technologies, but they do not kill the ways we work as people. There will still be fast food outlet in Pluto.

How do we interact with data?

This leads us to the question of how do we interact with data in business, and in my experience there were three main ways.

The first one is a question. A CEO or a manager simply calls the analyst they ✰trust✰, or sends them a message and asks a question. The analyst replies to the question with a quick answer or presents their findings in a presentation. From the CEO standpoint, they just ask a question.

The second one is ad-hoc analysis. Analyst gets a question and then goes to do the analysis. They might pull the data from a Data Warehouse and do some crazy coding, run a model, or simply load the data in a Power BI or Tableau and do visual analysis. Or they might even export data from somewhere, XLOOKUP it in Excel, add some crazy calculations, play with Goal Seek. This is a messy interaction, involving a lot of digging in and understanding the data very well.

The third one is the looking at a ready-made data visualisation. After getting the same question from a CEO again and again, analysts decides to build a data extraction pipeline, load the data into BI software and create a dashboard which alerts the CEO every Monday. Then the CEO opens the same dashboard again and again to see the same number every week.

Here we have it:

• Asking a question
• Analysing the data
• Seeing the data

So, how can AI be involved with these interactions?

Asking a question

We like to ask. No matter how many times everyone tells us to just google things before asking we still like to ask because if you ✰trust✰ someone, you know their answer will be much better and more relevant to you than some message by some random dude on StackOverflow on an issue which is just similar to yours but not exact.

And here we have AI. We do rely on AI on many questions and usage of StackOverflow is decreasing, but at some point we want someone to solve the exact problem we are having instead of trying to vibe-solve it with AI which just keeps hallucinating non-existent functions. (Happens to me all the time with questions on Tableau!) So, we ask people.

The same happens with asking questions about data. When the CEO will ask AI instead of their favourite analyst? When they start ✰trusting✰ AI. At some point, the AI itself and all the data governance infrastructure behind it might be so good that their favourite analyst will assure them: “Now you can ask our chatbot.”

Analysing the data

We like to dig into numbers ourselves. An analyst is someone who has curiosity to uncover insights, but also the one who wants to verify that those insights are really backed by data, so they want to check, and then they press Alt+MP on Excel.

And honestly, most BI tools are not very good at allowing quick ad-hoc analysis. It is a pain to write another DAX measure in Power BI just to provide another angle to quickly view the data.

This is where AI might shine – providing more flexible ways to look at the data, not limited by the constraints of the tool you are using. Most hardcore analysts would probably still refrain from doing that because they ✰trust✰ the code they wrote and not so much results from AI. But at some point, the AI itself and all the data governance infrastructure behind it might be so good that they will.

Seeing the data

Opening a dashboard you know very well beats asking the same question every time. You can ask simple question like “Siri, what’s the weather tomorrow?” every day, but I don’t think you would like to repeat something like this often: “Give me last week sales by department sorted from the highest margin to lowest excluding products which were on stock clearance sale, and home deliveries which were ordered a week ago”. You just want this thing to be done in background and ready to consume – this is called automation.

Here, AI might still be involved in creating this kind of view, but I don’t see voice and text-based requests replacing the need for visual in-advance prepared and tailored dashboards any time in the foreseeable future and maybe never.

The ways data visualisation is prepared might change, but it will be there until the vision loses its purpose, which, based on all sci-fi I have consumed, will never be.

Our place in future processes

In the future, there will still be those three ways of human-data interaction – asking a question, doing analysis, and seeing a data visualisation.

Unless AI is able to connect directly to business systems on its own, and on demand, the human will still be in the loop to set everything up – the infrastructure.

A great example is Data Agents in Microsoft Fabric. They allow surprisingly accurate answers about your data, but to make this all happen you need to invest plenty of time to prepare your semantic model for this functionality. As of now (2026 Q1) this is something AI is not capable of doing. You must feed the data from the right sources to the semantic model. While AI can aid in that, the decisions which data has to be there is still made by people.

Our first role in the age of AI will be to set up the infrastructure for it to work properly. While more and more processes will be AI-ized, the demand for people knowing how those processes are set up will be there. Even the more basic knowledge about data technology we have now will still be relevant, there will be humans who just need to plug in the wires in the right places to fire it up.

The next thing is similar to what happens when you play electronic music on stage.

There are multiple plugins playing different parts of your song, and you’re standing there with all the buttons turning them on and off, changing the way they play or what they play – you don’t need to even be able to play any instrument yourself. Doing analysis will be the same, you will manage plenty of agents doing different parts of analysis and research, writing and running the code for you. They might even arrive at conclusions autonomously, but human will still take the role of Project Manager. Someone has to have a question for AI to answer.

Our second role in the age of AI will be to orchestrate those AI agents. In the future, there will be plenty of agents doing various kinds of works, and some of them might be even orchestrating other agents, still, most of them will come from different vendors, so the integration of all different protocols and result formats might require a great conductor.

Finally, I believe humans will still consume data, this will not be machine-to-machine only interaction, while we just sit there enjoying life. We like to ask a colleague to evaluate the work we have done. We like to visualise results provided by neural network before claiming the result is meaningful. The humans will still be in the loop to check whether data visualisation is readable for humans if it was made for humans. Or simply to check if AI output makes sense.

Conclusion

We will ask questions about data, we will analyse it, and we will look at prepared data visualisations.

But for this, we will set up the AI infrastructure, orchestrate AI agents and check their work.

This is how the future will (probably) look like. Your job will be still there, just different. It will be the same feeling as when after years of writing SQL code to see a nicely formatted table, you launch BI tool for the first time, and now analysis works fundamentally different because you can drag things and view results immediately. But you had to set up the tool so you could ✰trust✰ it.