Glossary Terms

The lawful collection and analysis of information about competitors and markets to inform decisions. Validate sources, respect confidentiality and distinguish observed evidence from speculative conclusions.

AI methods for interpreting information in images or video, such as identifying objects or classifying scenes. Evaluate performance under actual operating conditions and consider privacy, error consequences and human review.

A table comparing predicted classes with actual classes in an evaluated dataset. It makes different types of classification error visible, helping teams choose measures that reflect the consequences of mistakes.

A neural networkA model made of connected computational units with adjustabl architecture using convolution operations to learn local patterns, commonly in images. Business evaluation should test the relevant conditions and failure cases rather than rely only on a benchmark score.

Capabilities that help an organisation create distinctive value across its activities. AI decisions should consider how technology supports those capabilities and where human expertise remains essential.

Decisions about the organisation’s overall direction, portfolio and allocation of resources. AI opportunities should be assessed against those priorities rather than pursued as disconnected tool experiments.

A factor that influences the cost of an activity or process. Understanding cost drivers helps teams evaluate whether an AI-enabled change reduces total effort or merely shifts cost elsewhere.

Conditions or capabilities that are important for achieving an objective. Make them explicit and measurable where possible, so implementation decisions focus on what must work rather than a long list of activities.

A method for evaluating a model across different training and validation splits of available data. It helps assess generalisation, provided the splits avoid information leakage and reflect the intended use.

An approach that considers customer needs and outcomes in business decisions and operations. AI-supported improvements should be tested for usefulness to customers, rather than judged only by internal efficiency.