Written by Maurice Sikkink
For most of the history of business analytics, we’ve assumed that a human starts the process.
Someone opens a dashboard. They notice sales are down. They compare this week with last week, segment the data, look at individual products, and eventually ask: Why did this happen?
AI is already making that investigation much faster. But I think focusing on speed misses the bigger change.
The real shift happens when the human no longer has to notice the problem first.
Dashboards depend on someone looking
I’ve spent years building analytics products, and one limitation keeps coming back: even a perfectly designed dashboard only works when someone looks at it.
Imagine an ecommerce business with thousands of products. Overall revenue might look relatively stable while one previously successful product category is quietly deteriorating. Perhaps customers still view those products, but fewer add them to their carts. Or conversion has fallen only for returning customers.
All of that information may already exist in the company’s data.
The problem is that somebody has to think to investigate it.
This creates an overlooked bottleneck in analytics. Companies don’t just miss insights because analysis takes too long. They miss insights because nobody asked the question.
AI can reverse the process
This is something we’ve been working on at Stormly, and it has changed how I think about AI analytics.
Instead of starting with a dashboard, you can start with a business question: Why did sales drop last week?
An AI system can determine which analysis is appropriate, examine behavioral and product data, and, when an existing analysis doesn’t answer the question, generate SQL to investigate further. It can also bring external market trends into the analysis.
That last part matters.
Suppose demand for a particular product falls 20%. Looking only at internal data, you might conclude that something is wrong with your pricing, website or marketing. But if search interest for the entire category dropped at the same time, you suddenly have a very different explanation — and probably a very different decision to make.
The interesting next step is removing the initial question as well.
Rather than waiting for someone to ask why something changed, an AI agent can continuously analyze the data, decide that a change deserves investigation, perform the analysis, and deliver the findings to the team.
The Monday morning workflow then changes from:
“Let’s open the dashboards and see what happened last week.”
to:
“Here are the three things that happened last week that deserve our attention.”
That’s a much bigger productivity change than simply generating a report faster.
Automation should reduce what people have to remember
We experienced this ourselves when we started using automated AI reporting internally.
Initially, I thought about the benefit in fairly conventional productivity terms: if someone spends several hours preparing a weekly analysis and AI can do the first pass, we’ve saved several hours.
But the more important benefit turned out to be different.
We removed the requirement that somebody remember to look in the first place.
That’s important because businesses accumulate enormous numbers of recurring tasks that depend on human initiation: check this metric, review that funnel, compare this period, investigate that category.
Each individual task may only take a few minutes. The problem is that people are busy, priorities compete, and the thing nobody remembered to check might be exactly where an important problem was developing.
AI should decide what deserves human attention
I don’t think the goal should be autonomous companies where AI makes every decision.
Quite the opposite.
The valuable division of labor is to let machines do more of the exhaustive looking while humans make the consequential decisions.
AI can inspect thousands of products, customer journeys and behavioral changes far more consistently than a person can. A human can then bring something AI still struggles with: context, judgment, priorities and an understanding of what the business is actually trying to accomplish.
For leaders adopting AI, I’d therefore suggest asking three questions:
- What do people currently have to remember to check? Those workflows are good candidates for proactive analysis.
- Where does analysis still depend on someone knowing the right question? AI may be able to surface the question before the team asks it.
- Can AI bring evidence to a person rather than make the final decision itself? Automation is often most useful when it improves human attention rather than replaces human judgment.
We’ve spent a lot of time talking about AI making knowledge workers faster.
I suspect its bigger impact will come from something less obvious: helping us notice the things we otherwise would have missed.
Author Bio:
Maurice Sikkink
CTO & Co-Founder, Stormly
Maurice builds AI-powered analytics technology for ecommerce businesses, with a focus on turning behavioral, product and market data into actionable business insights.