Information Technology in Management: Computers, MIS and DSS, Artificial Intelligence and Big Data, Data Warehousing and Mining, Knowledge Management and Managing Technological Change

Artificial intelligence is the capability of a machine to perform tasks that normally require human intelligence — perceiving, reasoning, learning, deciding and using language. The term was coined at the Dartmouth workshop of 1956 (John McCarthy); Alan Turing's 1950 test asks whether a machine's conversation is indistinguishable from a human's. Narrow (weak) AI performs one task; general (strong) AI would match human breadth and does not exist. Machine learning lets systems improve from data without explicit programming: supervised learning trains on labelled examples (classification and regression — credit scoring, demand forecasting), unsupervised learning finds structure in unlabelled data (clustering for customer segmentation, association, dimensionality reduction), and reinforcement learning learns actions from rewards (dynamic pricing, robotics). Deep learning uses multi-layer neural networks and drove the advances in image recognition, speech and natural language processing after 2012; generative AI — large language models built on the transformer architecture (2017) such as those behind ChatGPT (released November 2022) — produces text, images and code and has pushed AI into every office function. Business applications include chatbots and virtual assistants, recommendation engines, fraud detection, predictive maintenance, robotic process automation and algorithmic decision-making in HR and credit. Ethical and governance concerns are bias in training data, opacity ('black box'), privacy, accountability, job displacement and deepfakes; the EU's AI Act (2024) is the first comprehensive regulation, and India's Digital Personal Data Protection Act 2023 governs personal data while the IndiaAI Mission (approved March 2024, outlay of about ₹10,372 crore over five years) builds compute capacity, datasets and skilling.

Big data technologies handle what a single machine cannot: distributed storage and processing (Hadoop's HDFS and MapReduce, 2006; Apache Spark for in-memory speed), NoSQL databases for unstructured and semi-structured data (document, key-value, column and graph stores), cloud platforms that rent elastic capacity (infrastructure, platform and software as a service), and the Internet of Things as the fastest-growing source. For managers big data underpins customer analytics and personalisation, dynamic pricing, supply-chain visibility, risk and fraud models, people analytics in HR and evidence-based strategy; its limits are data quality, privacy and consent, security, the skills gap and the danger of mistaking correlation for cause.

A data warehouse is a subject-oriented, integrated, time-variant and non-volatile collection of data in support of management's decision-making (Bill Inmon, 'the father of the data warehouse', 1992). Subject-oriented: organised around customers, products or sales rather than applications; integrated: cleaned to consistent names, units and codes from many sources; time-variant: holds history with time stamps; non-volatile: loaded periodically and not updated by transactions. Data reach it by ETL — extract from operational systems, transform (clean, standardise, aggregate), load — via a staging area; a data mart is a departmental subset; metadata describe the data. Inmon's top-down approach builds the enterprise warehouse first in normalised form and derives marts from it; Ralph Kimball's bottom-up approach builds dimensional marts and unites them by a bus of conformed dimensions. The dimensional (star) schema has a central fact table of measures (sales amount, quantity) keyed to surrounding dimension tables (time, product, store, customer); a snowflake schema normalises the dimensions into further tables. A data lake stores raw data of all types for later use; a lakehouse combines the two.

Data mining is the discovery of previously unknown, valid and actionable patterns in large data sets — the analysis step of knowledge discovery in databases (KDD: selection, pre-processing, transformation, mining, interpretation). Its tasks are classification (assign to known classes — decision trees, naive Bayes, k-nearest neighbours, neural networks; credit approval, churn), regression (predict a number), clustering (group without labels — k-means, hierarchical; segmentation), association-rule mining (items that occur together — market-basket analysis by the Apriori algorithm, Agrawal and Srikant 1994), sequence and time-series analysis, anomaly or outlier detection (fraud), and text and web mining. An association rule X → Y is judged by support (fraction of transactions containing both X and Y), confidence (fraction of transactions with X that also contain Y) and lift (confidence ÷ support of Y; lift above 1 means a real association). CRISP-DM (1999) is the standard process: business understanding, data understanding, data preparation, modelling, evaluation, deployment. Overfitting — a model that memorises the training data and fails on new data — is the classic hazard, guarded against by hold-out and cross-validation.

Knowledge management (KM) is the systematic process of creating, capturing, organising, sharing and applying an organisation's knowledge to achieve its objectives. Polanyi distinguished tacit knowledge — personal, experiential, hard to articulate ('we know more than we can tell') — from explicit knowledge, which can be codified in documents, databases and procedures. Nonaka and Takeuchi's SECI model (1995, 'The Knowledge-Creating Company') describes four modes of conversion in a spiral: socialisation (tacit to tacit, by shared experience and apprenticeship), externalisation (tacit to explicit, by dialogue, metaphor and writing), combination (explicit to explicit, by systemising documents and databases) and internalisation (explicit to tacit, by learning by doing); 'ba' is the shared context in which it happens. KM strategies are codification (people-to-documents; reuse; Hansen, Nohria and Tierney 1999) and personalisation (people-to-people; expertise networks). Tools are knowledge repositories and portals, expertise directories, communities of practice (Wenger), lessons-learned and after-action reviews, storytelling, mentoring and enterprise social platforms. The learning organisation (Peter Senge, 'The Fifth Discipline', 1990) rests on five disciplines — personal mastery, mental models, shared vision, team learning and systems thinking, the fifth that integrates the rest; Argyris distinguished single-loop learning (correct the error) from double-loop learning (question the governing assumptions). Intellectual capital (Edvinsson and Malone; Stewart) is human, structural and relational capital, and the Skandia Navigator was the first attempt to report it.

Managing technological change in practice follows the general change models — Lewin's unfreeze-change-refreeze, Kotter's eight steps — applied to technology: a clear business case, top-management sponsorship, user involvement in design, pilots before roll-out, training and support, redesign of jobs and processes rather than automating the old ones (Hammer's business process re-engineering), attention to resistance (fear of job loss, deskilling, loss of control, habit) and measurement of adoption. Industry 4.0 (the term from Germany's 2011 Hannover Fair) bundles the current wave — cyber-physical systems, the Internet of Things, cloud, big data, AI, additive manufacturing, robotics and digital twins; Industry 5.0 (European Commission, 2021) adds human-centricity, sustainability and resilience. Digital transformation is the reshaping of business models, not merely the digitisation of documents (converting analogue to digital) or the digitalisation of processes.