--- title: "Stop Automating Reports. Automate the Decision to Investigate" url: "https://cooinsider.com/insight/stop-automating-reports-automate-the-decision-to-investigate/" author: "Aleks Sztemberg" published: "2026-10-01" updated: "2026-10-01" --- # Stop Automating Reports. Automate the Decision to Investigate For years, one of the first things companies did when their data became difficult to manage was build another report. Then we automated the reports. Daily sales reports. Weekly conversion reports. Monthly retention reports. Alerts when a metric crosses a threshold. AI has made it possible to summarize all of those reports faster. But I think that still leaves the most important part of analytics unchanged: someone has to know what deserves investigating in the first place. For [ecommerce teams](https://www.stormly.com/features), that is becoming increasingly difficult. A growing store may have hundreds or thousands of products, multiple markets and acquisition channels, different customer segments and millions of behavioral events. Somewhere inside all that data, conversion may be declining for a valuable customer group while a particular product is unexpectedly taking off. The information exists. The problem is knowing where to look. This is why I think the next step for [analytics for ecommerce](https://www.stormly.com/features) isn't another generation of automated reports. It's an AI agent that can recognize when something meaningful has happened and start investigating before a person asks. ### A report still requires someone to interpret it Imagine an ecommerce company's revenue falls 12% this week. A dashboard can show the decline immediately. An automated report can put it in someone's inbox on Monday morning. An AI summary can explain that revenue is down 12%. But none of those necessarily answers the question the business actually cares about: Why? Did traffic decline? Did customers continue visiting but stop adding products to their carts? Did checkout conversion change? Was the decline concentrated in one market? Did one previously successful product suddenly stop selling? Did returning customers behave differently? Or did something outside the store change? These questions require investigation, not reporting. That distinction has become central to how we think about Stormly. ### From AI chatbot to AI analyst A lot of AI analytics currently begins with a chat box. Ask, "Why did sales drop?" That's already a major improvement over requiring someone to understand which dashboard, filter and report contains the answer. But there's still an important limitation. **Someone has to know that sales dropped before they ask the question.** At Stormly, we're increasingly focused on what happens before that moment. An AI agent can monitor ecommerce analytics, detect an unusual change and start investigating the underlying data. It can look at customer segments, products, funnels and behavioral patterns, choose an appropriate analysis and generate SQL when an existing report isn't enough. The goal isn't simply to make querying data easier. It's to reduce the amount of analytical work that depends on somebody remembering to initiate it. ### The opportunity may be more important than the anomaly This isn't only about detecting problems. One of the most useful things analytics can uncover is something going unexpectedly well. Suppose a product suddenly starts converting significantly better among a particular customer segment. A conventional alert might not fire because nothing is "wrong." But commercially, that could be more important than a small decline elsewhere. Maybe there is an emerging customer need, an opportunity to increase inventory, or a marketing campaign worth expanding. This is where I think an AI agent can become genuinely useful operationally. Instead of producing a fixed list of metrics every week, it can ask: **What changed enough that somebody should care?** That's a very different job. ### Your store doesn't exist in isolation Another limitation of traditional ecommerce analytics is that it tends to explain the business using only data generated inside the business. But customers don't live inside your analytics platform. Their behavior is influenced by trends, competitors, seasonality and what is happening elsewhere in the market. This is why we've also been working with external Trends data at Stormly. Imagine searches and interest around a product category suddenly increase at the same time that product views and purchases rise in your store. Without the external context, you might credit a marketing campaign or website change. With it, you may discover that you're benefiting from a broader shift in demand. For analytics for ecommerce, understanding the difference matters because the business response is different. One tells you your marketing worked. The other may tell you there's a market opportunity you need to move quickly to capture. ### The real productivity gain is removing initiation This has changed how I think about AI in business operations more broadly. Most AI projects start with: "How can AI do this task faster?" I think there's a better question: **"Why does a person need to initiate this task at all?"** If someone spends three hours every Monday reviewing ecommerce reports, using AI to summarize those reports might reduce that work substantially. But the bigger opportunity is asking whether the Monday review needs to exist. What if an AI agent continuously looks for meaningful changes, investigates them and brings the important findings to the team when they actually happen? Then the workflow changes from: "Check everything and find what matters." to: "Investigate what matters and tell me when you find it." For a COO, that difference is important. The objective isn't to create more analytics. It's to reduce the operational effort required to turn data into decisions. ### Not every decision should be automated There is an important boundary here. I don't think the goal is to let AI make every business decision autonomously. There is a meaningful difference between an AI agent saying: "Conversion among returning customers in Germany has dropped significantly, and most of the decline appears at this checkout step." and saying: "I changed your checkout because I decided that was the solution." The first removes investigative work while giving a person better information. The second removes judgment. For many business operations, I think the most valuable near-term role for AI sits between those two points: **autonomous investigation, human decision-making.** Let the AI monitor, query, compare, investigate and surface opportunities. Let people decide what those findings mean for customers and the business. ### Analytics should compete on what gets done This also changes what I think analytics products will compete on. For years, the competition was about dashboards. Who has the cleanest interface? Who makes charts easiest to build? Who has the best filters? Those things still matter. But once an AI agent can work with the analytics system on your behalf, the interface becomes less central. Through technologies such as [MCP](https://www.stormly.com/mcp-for-ecommerce-analytics), for example, Stormly analytics can be connected to compatible AI clients. Instead of leaving a conversation, opening an analytics platform and manually recreating the question, the AI can work with the relevant project data directly. The analytics platform increasingly becomes infrastructure the agent can use. The competitive question then shifts from: "Who has the best dashboard?" to: **"Which system can actually help me understand what deserves my attention?"** ### The operational question I'd ask If I were evaluating an AI opportunity inside a business today, I wouldn't begin by listing the reports people spend the most time producing. I'd look for recurring workflows where someone is responsible for checking whether something happened. Checking whether conversion changed. Checking whether a product is underperforming. Checking whether customer behavior shifted. Checking whether an opportunity is emerging. Those are interesting candidates for AI because the work starts before the report. The biggest improvement may not come from generating the answer faster. It may come from recognizing that there is a question worth asking. --- Aleks Sztemberg is Co-Founder of [Stormly](https://www.stormly.com/), an AI-powered analytics platform for ecommerce businesses. Stormly combines an AI agent with ecommerce analytics, behavioral data, external Trends and MCP connectivity to help teams identify, investigate and understand changes in business performance.