Glossary

An approach in which multiple AI agents work together in a coordinated manner to handle complicated, multi-step business processes autonomously. Each agent takes on a specific sub-task — such as research, analysis, execution, or quality control. In low-surprise environments, this approach can unlock significant automation potential.

An AI system that does not respond to a task in a single step but instead plans autonomously, executes intermediate steps, uses tools, and evaluates outcomes — until the goal is achieved. AI agents are the building blocks of agentic AI systems. Example: an agent that independently researches, compares, and completes a travel booking, rather than simply providing a recommendation.

The capacity of an organization to strategically prioritize, organizationally anchor, and technically operate solutions based on artificial intelligence. AI readiness encompasses technical, organizational, and strategic prerequisites in equal measure.

See data strategy.

The systematic analysis of enterprise data with the aim of identifying patterns, understanding relationships, and supporting decisions — from descriptive reporting to predictive models. Analytics is a core component of data-inspired organizations.

Unity of the distinction planning/exploration.

Software that produces useful outputs from data — such as text, forecasts, classifications, pattern recognition, or recommendations — without requiring explicitly programmed rules. These solutions are based on statistical models and suggest "intelligence" in that they can perform tasks previously reserved for human cognition. Large language models are one of the most prominent examples.

See data strategy.

The discipline of transforming enterprise data into structured reports and interactive dashboards that provide decision-makers with a consistent view of the company's past and present. Business intelligence is a core component of an organization's analytics landscape and depends on sound data management. Common tools include Qlik Sense, Power BI, and SAP Analytics Cloud.

Command and control transfers existing knowledge to where action takes place — as directives, process descriptions, performance metrics, or rules. It presupposes a knowledge advantage on the part of those issuing instructions: only those who know better can direct meaningfully. In complicated environments, command and control is highly efficient. In complex environments it fails — because the critical competence develops where the problems arise, not where the instructions come from. There, it must be complemented by commitment by choice. Together they form the distinction management.

Commitment by choice is a social phenomenon grounded in mutual resistance and voluntary followership. It is based on earned trust and perceived competence, not formal authority. It cannot be mandated — it can only be observed after the fact. Within teams, it shifts situationally between the skilled practitioners. In management, it complements command and control: it surfaces problems, brings talent together, and creates the right conditions.

Unity of the distinction knowledge/skill.

Complex problems are high in surprises and cannot therefore be solved with available knowledge alone. Ideas are required to address them.

Complicated problems are low in surprises and can therefore be mastered with existing knowledge.

A data and AI operating model describes how an organization delivers its complicated value creation in data and AI — documenting the interplay of processes, roles, capabilities, technology, and governance. Design approaches imply a definable target state that is achievable through direct intervention. The operating model is one component of organizational design — it addresses the directly manageable part, not the structural conditions required to handle complex surprises.

The shared habits, values, and unspoken rules that determine how data is used in an organization to create value. Data culture is not an object of design — it is an emergent property of a complex social system and therefore cannot be directly controlled. It is effect, not cause: a symptom of the conditions under which data value creation takes place. Its practical value lies in diagnosis — it makes invisible barriers visible that block the use of data as an asset. Attempting to directly engineer data culture will fail. Changing the conditions, however, allows a fitting data culture to emerge as a side effect. By contrast, data governance is directly designable.

The totality of activities that ensure data is available in high quality and kept secure — through processes, roles, and technologies. Data governance addresses the complicated part of data value creation: it is directly designable and forms the reliable foundation on which data culture can emerge. Unlike data culture, governance can be consciously designed. It is an essential component of comprehensive data management.

The totality of activities that consolidate data from multiple source systems and make it available in a unified form. Data integration is a core task of data management and forms the technical foundation for analytics and AI applications.

A data platform that stores large volumes of structured and unstructured data in its raw format — as a flexible foundation for analytics, machine learning, and AI applications. Unlike a data warehouse, a data lake imposes no predefined structure on the data.

The totality of activities that ensure enterprise data is available, accurate, consistent, and secure. Data management encompasses disciplines such as data integration, data governance, and the operation of data platforms such as data warehouses or data lakes.

The term is used in practice with two distinct meanings. In the strict data mesh sense, a data product is a self-contained, domain-owned unit of analytical data together with everything needed to serve it reliably: transformation code, metadata, access interfaces, and quality commitments. In the broader, widely used sense, a data product refers to any useful output that generates value from data — including a BI dashboard or an AI solution. Anyone using the term should clarify which meaning they intend.

A data strategy is the strategy of the functional area responsible for data, analytics, and/or AI. Depending on its focus, it may also be called a BI or AI strategy. It defines how this organizational unit supports and amplifies the organization's overarching business strategy by delivering data-based services to internal (and potentially external) users. A data strategy does not define how the organization achieves competitive advantage through data and AI — that is and remains the task of the business strategy. Data and AI can strengthen an existing competitive advantage, reinforce existing capabilities, or open up new playing fields. A data strategy is neither a plan for building data and AI capabilities nor a supplement to the business strategy — both are widespread misconceptions.

Unity of the distinction data governance/data culture.

A data platform that integrates, historizes, and makes available structured enterprise data from multiple source systems for business intelligence and reporting. Unlike a data lake, a data warehouse follows a predefined data model and is optimized for structured analysis.

Data-inspired organizations use data, analytics, and AI to continuously experiment and learn, to ask new business-relevant questions, and to realize competitive advantages by combining data, creativity, and intuition.

A capability of modern AI systems that autonomously searches a large number of sources, consolidates information, and produces structured reports. Deep research builds on large language models and extends their capabilities with autonomous research and synthesis. Available in tools including ChatGPT, Claude, and Gemini.

Exploration is the activity of discovering an unknown solution path through ideas and experiments. Together with planning, exploration forms the distinction approach — the way an organization works on problems. Which approach is appropriate is determined by the character of the problem: planning for complicated components, exploration for complex ones.

A function is a bundle of specific services that solves a permanently recurring problem. Important: function here does not refer to the organizational unit that produces those services.

Generative AI is a class of AI systems that produce new content from training data — text, images, audio, video, or code. Unlike conventional AI systems that analyze or classify existing data, generative systems independently produce new outputs. Large language models such as ChatGPT or Claude are the most prominent example — they generate language. Generative AI is the technical foundation for applications such as RAG, AI agents, and Vibe coding.

Knowledge consists of uncontested statements about how to solve a problem. It arises through learning, is transferable, and is the decisive resource for complicated problems. Together with skill, knowledge forms the distinction competence.

A specific class of AI software trained on very large volumes of text data, enabling it to process and generate language — as the foundation for applications such as chatbots, text summarization, code generation, and document analysis. Modern LLMs increasingly process images and other data formats as well. LLMs are the technical foundation for AI agents and RAG systems. Well-known examples include ChatGPT (OpenAI) and Claude (Anthropic).

A class of AI software that does not receive explicitly programmed rules but derives them autonomously from data. Example: rather than being given rules for when a machine is likely to fail, a machine learning model learns this pattern from thousands of historical sensor readings. Machine learning is the technical foundation for predictive analytics and forms — at a much larger scale — the basis of modern large language models.

Unity of the distinction command and control/commitment by choice.

A misjudgement arises where knowledge is still lacking and action is nonetheless required. One decides on the basis of an assumption that turns out to be false. This is not a deficiency — it is the only way to learn and move forward in uncertain situations. Unlike mistakes, only misjudgements generate new knowledge.

Whoever makes a mistake knew better. The solution was known. Mistakes do not arise from ignorance but from non-compliance. Nothing can be learned from them: the knowledge was already there. The only appropriate response is prevention. Unlike misjudgements, mistakes generate no new knowledge.

An open standard that connects AI agents and other AI systems with existing enterprise applications — such as CRM, ERP, or ticketing systems — standardizing the exchange of data and actions between AI and enterprise software.

The deliberate design of structures, roles, and conditions with the aim of creating an environment in which the right decisions become more likely. Organizational design works indirectly — not through direct control of processes, but by shaping the structural conditions that guide the organization in the desired direction. A data and AI operating model is one component of this: it describes the complicated, directly manageable part of value creation. Organizational design goes further by also creating the structural conditions that enable the organization to respond to complex surprises. For organizations seeking to become data-inspired, it is a critical lever — because technology alone does not change decision logic.

Planning is the activity of breaking a known solution path into steps in advance and making decisions ahead of time. It presupposes that sufficient knowledge about the problem and the path to the solution exists. Together with exploration, planning forms the distinction approach — the way an organization works on problems. Which approach is appropriate is determined by the character of the problem: planning for complicated components, exploration for complex ones.

A technique in which an LLM does not rely solely on its training but retrieves relevant content from internal enterprise data sources in real time and incorporates it into its responses — such as from documents, manuals, or databases. RAG significantly improves the factual accuracy and currency of AI-generated answers.

Skill is the person-bound and non-transferable capacity to generate problem-solving feelings in concrete, often poorly understood situations. It is grounded in talent and practice and is the decisive resource for handling surprises — since by definition no knowledge yet exists for what is surprising.

A strategy is an integrated set of choices under uncertainty that articulates a plausible theory of how an organization intends to win in competition. It is the entrepreneurial bet on success. It encompasses choices about ambition, the selection of playing fields, how to win there, which capabilities are critical, and which systems reinforce them. Good strategic choices — particularly the theory of winning — have the character of principles: they rule out options without prescribing specific actions. From the strategic clarity gained, the required functions and capabilities emerge — whose implementation requires either planning or exploration depending on the nature of the problem. The strategy design process itself addresses a complex problem and therefore requires an exploratory approach. Every indivisible business unit, every level of aggregation, and every functional area of an organization requires its own strategy.

The identification and evaluation of use cases in which data, analytics, and AI generate concrete business value. Since discovering relevant use cases is an inherently complex endeavor, use case innovation requires an exploratory approach — serving as the bridge between business goals and technical implementation.

A practice in which software developers — and increasingly non-developers — generate functional source code by giving natural language instructions to an LLM, without writing every line manually. The term describes the shift from manual programming to AI-assisted code generation.

Unity of the distinction complex/complicated.