Source-linked AI summary
Reviving our data foundations is the most disruptive step to data maturity
Valentina Carapella, Ernesto Jimenez-Ruiz
TL;DR
Small- and medium-sized enterprises need evidence-backed, low-impact ways to strengthen data foundations before pursuing AI advances. The paper proposes dynamic, modular, process-focused knowledge graphs with cross-functional semantic mapping, culminating in an interconnected ecosystem that supports interoperability across business units.
Problem
Small- and medium-sized companies struggle to secure support for foundational data work and to adapt data strategies to continuously changing processes and flows.
Method
The paper proposes a low-impact strategy based on process and workflow ontologies, modular domain-focused knowledge graphs, and cross-functional ontology alignment.
Results
The paper's supported conclusion is that interconnected, modular knowledge graphs can provide semantic interoperability on demand across diverse business units.
Takeaways & Limitations
Companies should step back from technology hype and consolidate data foundations through lightweight, dynamic, and cross-functional knowledge-graph practices.
Abstract
from arXiv · showhide
The most disruptive step that enterprises of small-medium size and maturity can take to make the most of the latest technological advances in AI is to step back from the hype and focus on establishing or reviving a good knowledge foundation layer. It is a hard message to present to the executive team; therefore, it needs to be backed by evidence, and its implementation needs to be of minimal impact on the existing processes. In this vision statement, we discuss how we need to rethink what evidence speaks to the decision-makers and propose a low-impact data strategy that adapts to the existing and ever-changing data flows and processes across the company. We firmly believe that knowledge graph techniques will increasingly become non-negotiable in the data strategy of an AI-powered enterprise, provided that we approach their design in a modular, dynamic and cross-functional way.
1 Introduction
The paper argues that small- and medium-sized companies should resist technology hype and first strengthen their data foundations through incremental, minimally invasive steps. Foundational work must be connected to business KPIs to secure executive and product-team support.
- Small- and medium-sized companies often rush toward new technologies but experience high implementation failure rates when adoption strategy is neglected.The paper attributes the problem to adoption strategy rather than technology itself.
- The paper proposes a counter-intuitive step back: build solid data foundations across business functions through minimally invasive, incremental improvements.This approach is presented as the most disruptive step toward AI-driven transformation for companies with limited size and maturity.
- Executive buy-in is difficult because foundational work is commonly framed as technical debt rather than as a source of business value.
- Showing that AI outcomes depend on input data is insufficient to change priorities without connecting foundational improvements to KPIs such as ROI and TTM.
- Evaluating new technology with business-relevant parameters can help researchers identify promising directions and create opportunities for funding and academia-industry collaboration.
2 Business perspective
The paper frames Data Mesh and knowledge graphs as a basis for low-investment data strategies that accommodate changing business processes. It proposes process-focused ontologies, modular cross-departmental models, and cross-functional semantic mapping.
- 2 Business perspective: Data Mesh is presented as a framework for unlocking business value from data, while knowledge graphs naturally support domain ownership and distributed governance.Small- and medium-sized companies still face awareness, resource, and financial pressures that make adoption difficult.
- 2 Business perspective: Business processes generate changing data, while workarounds and ad hoc sub-processes make short- and medium-term standardisation and streamlining often unfeasible.
- 2.1 Ontologies of processes and workflows rather than just about data and relationships: Process and workflow ontologies should model sources of change and how change is applied, rather than only representing static data and relationships.Structured metadata produced through these ontologies can provide a more complete and reliable passport for data across its life cycle.
- 2.2 Modular and minimal ontologies for cross-departmental interoperability: Modular, minimal ontologies focused on priority domains and business processes can formalise knowledge with minimal overhead while supporting rapid prototyping and routine updates.The approach follows domain ownership across the company and focuses ontology use on a few key applications.
- 2.3 Ontology alignment to the next level, mapping cross-functionally: Mapping across departmental glossaries and processes, instead of imposing one standard vocabulary, is proposed as a low-resistance route to collaboration and improved time-to-product KPI.
3 Technical enablers
The technical vision is a modular ecosystem of knowledge graphs that preserves stakeholder-specific perspectives while enabling semantic interoperability on demand. It requires process semantic lifting, graph modularity, and richer alignment across functional areas.
- Semantic lifting of existing business processes and workflows remains limited by infrastructure, while dynamic-environment approaches may not scale well.GenAI is identified as an opportunity to facilitate process semantic lifting and knowledge-graph creation at scale.
- Scalability and adaptability require moving away from monolithic ontology and knowledge-graph solutions toward modular networks tailored to different organisational stakeholders.Different business units may retain inconsistent needs and viewpoints within the network.
- Advanced alignment systems should discover complex semantic correspondences and transformations across functional areas while accommodating logical disagreements.In agentic architectures, agents representing functional units may negotiate correspondences to support cooperative tasks.
- The proposed endpoint is a modular ecosystem of interconnected knowledge graphs that provides semantic interoperability on demand across diverse business units.
4 Conclusions
Successful AI adoption for growing companies depends on data maturity, but the recommended path emphasizes minimal investment and dynamic, modular foundations. The paper therefore recommends stepping back from technology hype and consolidating data foundations and practices.
- AI adoption for start-up to scale-up companies is presented as being paved by data maturity.
- The proposed strategy seeks maximal gain from minimal investment through lightweight, dynamic representations of processes and workflows.
- The recommended approach emphasizes high modularity and cross-functional semantic mapping to accommodate changing data flows and processes.
- The paper recommends stepping back from technology hype and consolidating data foundations and good practices.