For decades, business software has been designed around screens.
If you want to understand sales, you open a dashboard. If you want to investigate conversion, you find the right report. If you want a different answer, you change filters, create a segment or ask an analyst to write a query.
Generative AI is changing that interface.
But I don't think the interesting future is simply putting a chatbot next to every dashboard.
I think we're approaching a bigger shift: business users will increasingly interact with data through intent rather than software navigation.
We've spent years teaching people how to use analytics software
Consider a simple ecommerce question:
Why did sales drop last week?
The person asking doesn't really want a dashboard.
They want an answer.
Yet traditional analytics turns that question into a software workflow. Which dashboard should I open? Which period should I compare? Should I segment by customer type? Product? Traffic source? Should I investigate conversion or traffic first?
The software exposes its analytical capabilities, but the user still has to determine how to use them.
We have been working on reversing that relationship at Stormly.
Through an MCP-connected AI client, someone can ask the business question directly. The system can determine whether an existing AI-based analysis is appropriate. If there isn't one, it can generate SQL to investigate the underlying data.
It can also incorporate external market-trend information where relevant.
The user doesn't necessarily need to know which analytical mechanism produced the answer.
That's a significant change in how we design business software.
Natural language is only the surface
It would be easy to describe this as replacing dashboards with chat.
I think that's too simplistic.
The difficult part isn't creating a conversational interface. LLMs are already remarkably good at that.
The difficult part is what happens after someone types:
Why did sales drop?
A reliable system may need to determine whether traffic changed, whether conversion changed, which products contributed to the decline, whether customer behavior shifted, and whether the same category is declining in the wider market.
That's why I don't think enterprise analytics becomes “LLM + database.”
The natural-language interface needs analytical systems underneath it that can actually investigate the question.
The chat box is the simple part.
MCP changes where the software interface lives
MCP makes this even more interesting.
Historically, if we built a useful analytics capability, customers had to come into our application to use it.
That assumption is disappearing.
A user can increasingly interact with business systems from the AI environment where they're already working. The analytics product becomes a capability available to the agent rather than another application the employee needs to visit.
This may eventually change how we think about SaaS itself.
Software companies have spent enormous effort optimizing navigation, menus and dashboards because the graphical interface was where customers experienced the product.
In an agentic world, some of the most valuable software may increasingly sit behind the interface.
The user expresses an intention. The agent determines which capabilities it needs.
Proactive AI changes the workflow again
There's another step beyond asking questions.
Today, most analytics still starts with human curiosity. Someone notices something or remembers to check a metric.
But an AI system can initiate that process too.
For example, an analysis can run automatically and the findings can arrive by email every Monday. Instead of starting the week opening dashboards to discover what happened, the team can start with the changes that deserve attention.
That sounds like a small workflow improvement, but I think the underlying shift is significant.
The old model is:
Human → software → data → insight.
The emerging model is increasingly:
Data → AI → insight → human decision.
Humans don't disappear from that chain. Their position changes.
Business software will compete on what it can do, not how many screens it has
I don't expect dashboards to disappear. Visual exploration remains useful, particularly when someone wants to investigate data manually.
But dashboards will no longer need to be the front door to every analytical capability.
That's an important change for technology leaders designing the next generation of enterprise software.
We should spend less time asking how AI can help people operate our interfaces faster and more time asking whether they need to operate those interfaces at all.
The most consequential AI interface may ultimately be the one users barely notice, because instead of learning where the answer lives inside the software, they can simply ask the business question they actually care about.