Roman Unterstöger: Making answers hold up in practice

Generating an answer is quick. Whether a team can work with it depends on other questions: Which definition applies? Where does the figure come from? Who is allowed to see it? And who resolves the exception?

At beyond chaotic analytics, I work on enterprise AI and putting decisions into everyday practice. I focus on the point where data, business rules and daily work meet. That is where a technically possible answer needs to become a traceable, usable one.

Why I write this blog

When the same question keeps landing with the same expert, a better report often is not enough. The applicable rule, its source or a known exception still lives in someone’s head. Adding a chatbot does not automatically close that gap.

Here, I write about how organisations can make that context explicit and usable for people and AI alike. Three themes run through my articles:

• Evidence before wording: An AI answer needs an agreed business foundation. Definition, calculation and permissions must be correct before a language model turns them into a sentence.
• Context with ownership: Terms, relationships and rules need somewhere they can be maintained and checked. An ever longer prompt does not replace that work.
• Data for specific users: A data product must serve a purpose and make dependable commitments. A dataset with a name and an owner is not enough.

My perspective on execution

My work connects SAP and analytics environments with data quality, governance and enterprise AI. What matters to me is whether the solution holds up in the actual workflow: Who checks the answer? When does a rule change? What happens when the available data does not support a clear conclusion?

Putting a solution into daily practice therefore requires regular routines, clear responsibilities and feedback from use. A team should know what it can decide independently and when it needs an expert to clarify something. Time saved can then be measured by comparing handling, checking and follow-up questions before and after the change.

This perspective complements Adrian’s work on decision architecture. He describes the path from decision to impact. I focus on how data context, governance and committed routines support that path in daily operations. Our shared aim: the capability stays with your team.

Where to read my work

• Evidence-first enterprise AI: Why an answer needs evidence
• Enterprise context: Why it does not belong only in a prompt
• Data products: Why a dataset with an owner is not yet a product
• Our services: beyond chaotic analytics
• LinkedIn: Roman Unterstöger on LinkedIn

Start with a recurring question

If your experts keep explaining the same figure or rule, that is a concrete starting point. In the Clarity Audit with Adrian, you work through one case: the rule, source and known exceptions. You also establish owners, next steps and a template that helps your team apply the method to further questions.

For questions about my articles, you can reach me at roman@beyond-chaotic-analytics.ch.