Data Mesh, Data Meh? Why Many Companies Need to Reassess Their BI and Data Organization

The Problem: In Many Places, Principles Turned Into a Blueprint

And between a target architecture and everyday reality in BI, reporting, and business departments lies an uncomfortable truth. Not every business unit is a data product team. Not every area has the analytical maturity, staffing stability, and methodological depth to sustainably carry semantics, KPI logic, quality, and governance on its own.

That is exactly why the debate is now being reopened. And this time for more realistic reasons. Many companies are at the end of a tool cycle. Platforms, reporting landscapes, and governance models need to be reassessed. At the same time, AI is changing expectations around data access. Near-time reporting, on-demand analytics, and natural-language interfaces are currently reshaping what self-service actually means.

So the question is no longer which architectural buzzword sounded most convincing. The question is: Which BI and data operating model works today under the conditions that now matter? For many, the more honest answer is not more decentralization, but better self-service BI with strong guardrails, clean semantics, and an operating model that matches the actual resilience of the business units.
‍
‍

Data Mesh Was Often More Attractive Than Organizational Reality

Data Mesh introduced important impulses. Away from the “central team” bottleneck and toward more responsibility in the domains. There is nothing wrong with that. It became problematic where the resilience of the business units was overestimated.

The reality often looked like this:

• the same KPI existed in multiple variants
• local reporting logic existed without shared semantics
• power users became unofficial carriers of critical logic
• governance existed on slides, but not in day-to-day work
• more freedom came with less comparability

In a BI context, this becomes visible quickly. Because this is not just about data access, it is about reliable steering. If revenue, margin, inventory, or forecast are interpreted differently in every area, that is not a sign of successful decentralization. It is a steering problem.

The uncomfortable truth: Not every business unit becomes better through more autonomy. Some mainly becomemore inconsistent.

‍

The Real Shift Is Happening Now: Self-Service Is Changing Its Character

For a long time, self-service BI mostly meant that business units built things themselves. Dashboards, reports, analyses. The guiding principle was freedom to build and as little dependency on the central team as possible.

That understanding fits the current reality less and less. Today, many users do not primarily want to build. They want answers. They want to put KPIs into context, understand deviations, segment quickly, pull a report on short notice, or ask a question in natural language. Directly and in context.

The expectation is shifting from “I want to build reports myself” to “I want to get reliable answers at any time.” 

That is not a detail. It is a different operating model. Self-service is moving from a build logic to a response logic. Near-time reporting, semantic layers, and AI assistants are accelerating this shift.

‍

More Access Does Not Make Governance Smaller — It Makes It More Important

The easier access to data becomes, the greater the risk that seemingly plausible answers are built on unclean logic. A good interface does not solve a semantic problem. An AI assistant does not replace a robust KPI definition. An on-demand report is not automatically trustworthy just because it is delivered quickly.

The lower the barrier to use, the more stable the foundation must be. Self-service does not scale through maximum freedom, but through clear guardrails.

These include:

• binding KPI definitions and curated semantic models
• certified reports and trustworthy data products
• clear roles and permission concepts
• traceable data origin and freshness
• defined escalation paths in case of conflicts in KPI logic

Guardrails are not a brake on self-service. They are the prerequisite for self-service to remain reliable under AI conditions. Without guardrails, you do not democratize insight, you democratize interpretive ambiguity.

‍

Not Every Business Unit Needs the Same Degree of Freedom

A common misconception: self-service is treated as a uniform target vision. In practice, self-service should be designed in stages according to maturity level. What matters is how much analytical and semantic responsibility a business unit can realistically carry.

‍

A simple maturity model:‍

Level 1 – Consumers

Use finished, certified dashboards, filter them, and interpret them safely. No own KPI logic or datamodels. Suitable where data maturity is low or standardization is high.
‍

Level 2 – Explorers

Guided self-service. Own reports and analyses, but based on a centrally provided, certified semantic layer and clearly defined KPI logic. Freedom increases, but the foundation remains centrally managed.
‍

Level 3 – Owners

High data maturity and analytical independence. Broad autonomy is possible, but not as decoupling. It requires central platform guardrails, shared standards, role models, and binding semantic guidelines.

The central insight: Not every business unit should be a producer or owner. For many, a well-guided explorer mode is the better target model.

‍

It Is Not the Industry That Decides, but the Substance Within the Business Unit

Whether more decentralization makes sense is rarely decided by the org chart. It is decided in day-to-day reality. The stronger the analytical substance in the business unit, the more likely decentralized structures will hold up. Where that substance is missing, decentralization is quickly romanticized. Then “more ownership” in practice often means nothing more than more local logic, more KPI deviation, and more dependency on individual people.

What matters is the combination of:

• the company’s technical maturity
• the analytical strength within the business unit
• process stability
• the criticality of the KPIs
• realistically sustainable semantic responsibility

Anyone who ignores this doesnot build a modern operating model. They merely redistribute complexity.

‍

What Many Companies Actually Need Right Now

The task is not to continue old ideological battles between central and decentralized models. The task is tore align the BI operating model under new conditions.

Most companies primarily need three things:

1. An honest assessment of business-unit maturity. 
   Not every area needs maximum autonomy.
2. A modern understanding of self-service. 
   Not building everything yourself, but getting fast, understandable, and reliable answers, increasingly on demand and through natural language.
3. Strong guardrails as enablers. 
   Clean semantics, reliable KPI logic, and clearly guided governance are the foundation.

For BI and data teams, this changes their role. They are no longer just ticket receivers or tool administrators. They become architects of a reliable usage model. Not building every request themselves, but creating the conditions under which many requests can be answered quickly, consistently, and trustworthily.

The goal is not maximum centralization. The goal is scalable usability.

‍

Conclusion

For many, Data Mesh was an important mental model. But it was also often a buzzword of the last tool and transformation phase. Today, companies face a different reality: platforms are being reassessed, governance must be robust, and AI is changing how business units want to work with data.

What is needed now is a more mature BI operating model. Less ideology, more testing of actual resilience. Modern self-service on demand, with strong guardrails in the background. Because in the end, what matters is not how modern a model looks on a slide. What matters is whether business units can work with data reliably, quickly, understandably, comparably, and in a way that remains controllable even under AI-shaped usage patterns.

Not every business unit needs maximum data autonomy.
But almost every business unit needs better self-service.