From "So What?" to "Now What?"

1. From "So What?" to "Now What?"

Once you have a firm grasp of the data, the next step is to turn those insights into recommendations. In this phase, we'll use AI to generate and test hypotheses and evaluate possible scenarios.

2. Example: Reducing Operational Costs

We'll stick with the same scenario of advising a mid-size European manufacturer facing operational costs that are too high.

3. Progress Check

At this point, the analysis is complete. You now have a structured dataset with two tables and a dashboard that highlights the key information for your team.

4. Progress Check

Normally, AI systems perform better when interpreting structured raw data rather than already built charts, so we'll go back to the raw data for testing hypotheses. If you prefer the AI to extract insights directly from the charts, you can use the same prompts shown in this video, but use it right after generating each chart. This allows the model to rely on both the structured data and the visual output, which can lead to improved responses.

5. Generating Hypotheses

Starting from the two data tables, which we upload and attach, we can translate analytical insights into hypotheses, which are testable statements grounded in facts. We can guide the model with a prompt asking for three hypotheses, each written as a clear statement. These should be followed by a short explanation and a reference to the supporting data. As we observe, the model provides the three hypotheses in a structured manner.

6. Strategic Options

Next, we ask the model to group those hypotheses into potential strategic levers. These are the major choices that follow from the hypotheses. In this case, we get Workforce Optimization, Asset & Service Rationalization, and Operational Efficiency in IT.

7. Challenging Hypotheses

Since AI systems can make mistakes, we need to assess how robust and relevant their insights are - we don't want to accept the first list of hypotheses as fact.

8. Challenging Hypotheses

For example, we can ask the model to show, for each strategic lever, which rows in the dataset support it.

9. Challenging Hypotheses

We can also ask the model to challenge its own hypotheses. One simple approach is to prompt it to propose alternative explanations for the observed cost patterns and flag any data quality issues. This helps us assess robustness in a conversational way. We can observe that the model effectively provides alternatives, although mentioning that they are more unlikely.

10. "What If?" Scenarios

Finally, we can use AI systems to run what-if scenarios. The goal is to compare alternative recommendations quickly. These scenarios show how quickly AI can help us explore alternative recommendations. The model adjusts one lever at a time providing us a simple view of how different choices affect total operational costs. At this point, we can go back to the strategic options and refine them.

11. Consultant & AI Collaboration

Throughout this process, notice the role of the consultant. AI helps with structure, calculations and first drafts. Our job is to set the context, choose which levers to explore, judge what is realistic based on the data, and deliver better recommendations faster.

12. Let's practice!

Now it's your turn to begin generating and challenging hypotheses!

Create Your Free Account

By continuing, you accept our Terms of Use, our Privacy Policy and that your data is stored in the USA.