The added value of artificial intelligence (AI) lies not in its ability to answer questions, but in its potential to speed up decision-making. While platforms like Airwallex, an intelligent financial platform for international businesses, use AI to answer questions, the real challenge is whether AI can help users make decisions more quickly. According to product launch pages and discussions on LinkedIn, the future of data analysis seems to involve a large interface where users can enter a question in natural language and receive a chart labeled "AI-powered analysis." This idea is appealing because, for a long time, extracting insights from data required someone who understood the data warehouse, its structure, and enough SQL (a programming language used for managing data) to avoid disrupting operations. A chat-like interface offers a more accessible way for people to "talk to their data." Despite this, it's important to note that while chat interfaces make data more accessible, access alone is not enough to move from identifying a problem to solving it. Dashboards, which are visual tools that display key performance indicators and other important metrics, remain central to how teams understand their progress. With modern technology and some help from AI, it's possible to create a dashboard for growth experiments in about fifteen minutes. These dashboards are especially effective when teams already know what they're looking for, serving as shared communication tools where everyone is aligned on what to track and what success looks like. When numbers fall out of the expected range, it becomes immediately visible. Natural language interfaces have real utility, allowing users to ask questions in everyday language and receive sensible answers. This reduces the time, skills, and context-switching needed to extract insights from company data. This is particularly useful in teams where the person asking the question isn't the one who knows SQL. However, these chat interfaces do not help users know what to ask. It's still up to the user to decide which questions are relevant, which groups of people or time periods to focus on, and which comparisons are worth studying. These reflections lead to a key question: How can AI best be used for analytics if the answer isn't simply in making dashboards easier to create or using chat interfaces for convenience? The concept of agent-based analysis emerges as a potential solution, shifting from tools that answer questions to systems that assist users in decision-making. This approach reduces the delay between a market change and a team's response. Instead of waiting for a user to ask, "What was the performance of the last campaign," an agent-based system would continuously monitor data, understand what matters to the company, detect what is normal and what has changed, suggest specific actions to try, and track the results over time. The key metric to focus on is decision-making time. How long does it take an organization to recognize an important issue, decide on a course of action, and implement it? Dashboards make important numbers visible to everyone, and conversational interfaces help more people get personalized views and detect anomalies without the delays of traditional business intelligence systems. However, the limitation of these tools is determined by how quickly humans can process data, analyze it, agree on a response, and act on it. Improving AI systems to understand the business context behind changes, explain what caused them, test responses, and measure their effectiveness could significantly reduce decision-making time. However, faster decisions also carry risks. If an AI system simply optimizes the speed of suggesting and executing changes, it might make errors more quickly. Good AI systems must understand not only the data but also the context and governance levels, adapting their actions to each situation and respecting policies and best practices. The distinction between knowing and acting is crucial. It might be tempting to view AI-assisted analysis as just a user interface problem, adding a conversational layer to existing business intelligence systems. However, the risk is investing in AI to make it easier to ask questions without considering how long it will take to implement the answers. A better approach would be to start with decision-making time. Ask: At what point in the current process is learning blocked? Which decisions are too slow, too complex, or too dependent on a single person? What would it take for an AI system to not only answer "What happened?" but also help determine "What should we try now?" and "Did it work?" Starting from this point, the chat becomes a powerful tool within the overall strategy. Dashboards will continue to serve their purpose: providing everyone with a shared view of the facts. Conversational interfaces will continue to make data less intimidating and easier to use. However, the real advantage will come from the systems built behind the scenes—agents that work in the background to reduce the gap between the appearance of a new element in the data and the reaction of the teams. AI-assisted analysis doesn't serve to replace dashboards or create the best chat interfaces. Instead, it aims to reduce the gap between knowledge and action, ensuring decisions are based on data and context. This is how organizations can be equipped to learn reliably at the speed of their data rather than the speed of their meetings.