AI for Business - AI Applications for Business Success

Explore global challenges in providing clean drinking water and examine how a last mile water delivery company uses artificial intelligence to create business value and improve lives.

Analyze a water-container business using swot analysis to balance intrinsic strengths and weaknesses with extrinsic opportunities and threats, emphasizing data fidelity, logistics, and supply chain risks.

Leads with a SWOT-backed plan to apply AI to improve data quality and reduce expansion costs, using smart goals to specify measurable, time-bound actions.

Explore the BI analyst workflow and three major limitations: complexity of operations, reliance on descriptive statistics, and correlation cannot illuminate causation, with AI offering causal insights to improve metrics.

Explore how descriptive statistics limit manual influencer analysis and how AI can automatically identify and test drivers to causally affect business metrics and objectives.

Explore how a business intelligence analyst assesses whether minutes between consecutive deliveries or deliveries scheduled drive total barrels delivered per shift, using correlation to inform CFO decisions.

Explore identifying the leading influencer driving water barrel deliveries per route by transforming time series into shadow manifolds and applying the k-nearest neighbors approach to match states.

Explore how multiple time series form a state space and how Takens theorem enables extracting dynamics from lagged representations, using the shadow manifold and driving forces.

Compare shadow and lagged manifolds using k-nearest neighbors to identify the most influential time series driving a business dynamic system, via three-nearest-neighbor projections of state-space points and plane-based scoring.

Quantify attainability of smart goals by building an AI model to simulate a delivery process with fine-grained data, assessing feature attribution and thresholds to reduce shift duration.

Explore gradient boosted machines with Manhattan distance, fitting models that sum to predict shift durations, using medians of residuals. Learn early stopping on validation to prevent overfitting and improve generalization.

Explore gradient boosted machines and how successive trees learn the median of residuals to refine predictions, using non-linear splits to capture feature interactions and produce non-linear outputs.

Learn how Shapley values fairly allocate a model's prediction cost across features using marginals, means, and the Shap library to derive the optimal allocation.

Explore Friedman's H statistic to test interaction effects in AI models, using partial dependence functions and Shapley values to assess attainability and feature interactions.

Apply Lime to provide local, model-agnostic explanations by perturbing inputs, fitting a surrogate generalized linear model, and revealing how input factors influence a specific prediction.

Explore ai-driven models that replace manual sensitivity analysis by learning from historical data. Identify top actions that forecast a reduction in shift duration using GBM regression and Shapley values.

Train a GBM AI model to forecast average shift duration, use a waterfall chart and Shapley values to simulate top actions and reduce duration.

Explore traditional causation methods using Welch's t-test to compare pre- and post-action shift durations, testing attribution and null hypotheses while guarding against spurious correlations in a water delivery use case.

Explore advanced causal methods on time series to assess how average shift duration responds to truck maintenance issues and performance write ups per day using Shapley values and convergent cross-mapping.

Explore Takens' theorem and convergent cross mapping to infer causation from time series, using shadow manifolds and diffeomorphism to link shift duration, truck maintenance, and performance write ups.

Forecast business metrics with ai models and identify key drivers through interpretability. Infer causation from historical data without experiments and prepare for final statistically validated experiments like an a/b test.

Explore the hybrid experiment approach to detect interaction effects in mixed design studies, using t-tests and the central limit theorem to attribute changes in key business metrics.

Explore quantile difference tests to assess statistically significant interaction effects at the 25th percentile using the asymptotic distribution of sample quantiles and the t-test.