Administration and documents
Classify records, extract fields, monitor deadlines and prepare summaries.
Final validation and administrative use remain human responsibilities.
How to read this page
These sections describe possible reuse of workflow patterns.
They are not a catalogue of pre-packaged services. Any real use requires checking sources, access rights, requirements, responsibilities and metrics for the target context.
Problemwhat must be observed
Datasources, cadence, licences
Automationtriggers, rules, models
Outputcards, reports, dashboards
Controlgates and supervision
Evaluationmetrics and limits
Explore by problem
Search by what you want to achieve, not by workflow name.
AI and data for industrial processes
Verifiable examples of anomaly detection, soft sensors, energy optimisation and process models using public or explicitly simulated data.
Explore →RPA automation and data workflows
RPA automation examples for collecting data, applying checks, producing reports and publishing outputs with traceability and error handling.
Explore →Anomaly detection and predictive monitoring
How to structure anomaly detection with baselines, drift controls, residuals, thresholds and validation while separating statistical alerts from real faults.
Explore →Soft sensors and Virtual Analyzers
A verifiable soft-sensor approach for estimating quality between lab assays with delay, OOD detection, abstention, baselines and point-in-time validation.
Explore →AI, data and energy
Workflows on energy stress, hydrogen, fertilisers, renewables, steam networks and the AI supply chain using prices, scenarios, forecasts and explicit limitations.
Explore →Forecasting with public data
Forecasting examples across markets, macroeconomics, energy, weather and Etna using benchmarks, uncertainty intervals, vintages and mature-error tracking.
Explore →Etna: AI, satellites and public data
Experimental Etna workflows for VONA, tremor, seismicity, thermal data, satellites, nowcasts, forecasts and atmospheric trajectories with explicit limitations.
Explore →Use cases
Ten use cases, without generic promises.
Each example separates the activity that may be automated, the possible technique and the decision that must remain human. Labels do not mean the case is already implemented.
Industry and maintenance
Collect reports, detect anomalies and estimate consumption or failure probability.
RPA · Statistics · ML · Human oversight
Experimental output does not replace operational procedures and systems.
Energy
Update indicators, compare scenarios and detect regime changes.
RPA · Time series · ML
Decisions require operational data and defined responsibility.
Logistics and supply chains
Compare routes, availability and delays and flag exceptions.
Rules · RPA · Forecasting
Quality depends on source freshness and coverage.
Environment and territory
Collect sensors and bulletins, update maps and recognise events.
RPA · Image AI · Geospatial
It does not replace official sources or warning systems.
Research and universities
Update bibliographies, classify papers and extract concepts.
RPA · ML · Text AI
Citations, interpretation and scientific correctness require review.
Culture and heritage
Catalogue images, transcribe texts and build narrative paths.
RPA · Text AI · Image AI
Attribution, historical context and rights remain central.
Communication and information
Compare sources and prepare controlled summaries or publications.
RPA · NLP · Generative AI
Editorial review prevents a draft from automatically becoming a fact.
Accessibility
Produce chart descriptions, simplified versions and alternative formats.
Text AI · Image AI · Human oversight
Quality should be tested with users and assistive technologies.
Economic and financial analysis
Calculate indicators, classify regimes and compare scenarios.
Statistics · ML · Time series
It is not a recommendation or guarantee of results.
Connections to implemented workflows are provided in the sections and matrix below.
Transferability map
From implemented workflows to possible new contexts.
| Workflow | Technical pattern | Possible industrial or organisational applications | Possible research and communication applications | Required condition |
|---|---|---|---|---|
| WF-1 · Weather automation | Rule-based multi-variable index | comfort, maintenance, operational risk, multi-site priority | rule transparency and risk-communication studies | domain-validated thresholds, without clinical inference |
| WF-2 · The New Margin | Derived KPI + forecast + track record | energy, yield, efficiency, cost or margin indicators | model-baseline comparison and temporal stability | economic definition of the KPI and matured outcomes |
| WF-3 · Market Overview | Cross-source radar, regime and anomalies | operational monitoring, KPI portfolios, early warning | drift, anomaly review and regime interpretability | classification must not become an automatic decision |
| WF-4 · AI Supply Chain | Thematic baskets, normalisation and z-scores | value chains, suppliers, technologies or industrial assets | proxy construction, basket robustness and network analysis | documented proxies and survivorship-bias controls |
| WF-5 · Satellite | Event trigger + geospatial fusion + safety gate | environmental events, distributed infrastructure, territorial monitoring | sensor fusion, uncertainty propagation and educational scenarios | separate observations, simulations and official sources |
| WF-6 · Energy Crisis Thermometer | Composite index + confidence-aware fallback + state machine | energy stress, supply chains, country risk and operating continuity | research on proxies, regimes and decisions under uncertainty | public weights, quality, confidence and limitations |
| WF-7 · Etna Forecast | Multimodal nowcasting + modality gate + scorecard | rare-event monitoring and systems with incomplete sensors | calibration, Hawkes baseline and modality comparison | strict separation from alerts and authoritative sources |
| WF-8 · Bluesky multi-source | Multi-source NLP + source policy + facts-only templating | bilingual technical dissemination and thematic observatories | source independence, communicability and anti-reuse gates | copyright-aware publication, disclosure and source links |
| WF-9 · World Economy Engine | Macro nowcasting + model ensemble + structural graph + point-in-time governance | economic intelligence, planning, stress testing, KPI portfolios and supply-chain intelligence | forecast comparison, regimes, graph ablation, revision/vintage bias and uncertainty | graph assumptions disclosed; benchmarks and time-t data only; not financial advice |
| WF-10 · Etna Sentinel | Multi-source ingestion + deterministic model + ML anomaly context + fail-closed event relay | event monitoring, infrastructure screening, digital operations and governed automated communication | trajectory comparison, anomaly context, policy gates and quality degradation | not a concentration/deposition model, alert, safety or navigation aid |
| WF-11 · Hydrogen Route Observatory | Physical balance + prices + carbon policy + sensitivity | route economics, make-or-buy scenarios and decarbonisation | emissions boundaries, sensitivity and break-even robustness | PPA is a scenario; no site-specific cost representation |
| WF-12 · Ammonia & Fertilizer Chain | Physical cost chain + econometric transmission + maturation | fertilizers, feedstocks, cost pressure and procurement intelligence | lags, elasticity and ML-vs-physical-model comparison | global benchmark kept distinct from modelled European cost |
| WF-13 · Italy Variable Renewable Forecast | Capacity-normalised point-in-time forecasting | load shifting, energy planning, storage and flexibility | vintage bias and skill versus physical/operator baselines | V1 covers solar+wind only and does not certify RFNBO |
| WF-14 · Process Sentinel | Multivariate monitoring + frozen synthetic data (simulated data) evaluation era | condition monitoring, early warning and assisted diagnosis | detection-delay/false-alarm trade-off and model gating | performance applies only to the stated benchmark/simulator |
| WF-15 · Virtual Analyzer | Delayed-label estimation + OOD abstention | soft sensing, inferential measurement and quality monitoring | conformal coverage, OOD as domain distance and model error versus laboratory uncertainty | truth and laboratory are simulated; no real performance claim |
| WF-16 · Energy & Steam Optimizer | Utility balances + LP/MILP + frozen reference policy | energy management, steam balancing and capacity debottlenecking | LP duality, unit commitment and sensitivity | utility demand based on synthetic data (simulated data); modelled opportunity, not measured plant savings; electrical balance not modelled |
| WF-17 · Etna Fusion Lab | Bounded acquisition + data health + quality-aware fusion + prospective ledger | sensor fusion, condition monitoring, anomaly evidence and governed communication | prospective calibration, modality ablation, drift and comparison with strong baselines | Activity Index, escalation and VONA ash-positive are distinct targets; no output replaces official sources |
| WF-18 · Climate Comfort Hours | Multi-city hourly forecast + transparent heuristic score + semantic priority + graceful degradation | outdoor planning, territorial communication, tourism and non-operational weather dashboards | penalty sensitivity, class robustness and comparison with appropriate bioclimatic indices | not a medical index, PMV/PPD, UTCI, weather alert or normative indoor-comfort metric |
| WF-19 · Lab Intelligence Community | Deterministic editorial templates + registry-driven rotation + fail-closed publishing | technical communication, knowledge dissemination, portfolio discovery and governed campaigns | controlled comparison across editorial families and public engagement measures | no language model in the writing path; claims are limited to documented workflows and publishing stops when gates cannot be verified |
| WF-20 · The Public Service Run | Traffic-aware alternatives + calibrated uncertainty + risk-aware departure optimisation | commute and logistics planning, route/departure selection and non-operational decision-support demonstrations | interval calibration, decision regret, distribution shift and provider stress scenarios | educational use only: not navigation, dispatch or safety advice; the near-departure ETA proxy is not GPS ground truth |
| WF-21 · Port Operability · NYC Ferry | Multi-horizon forecasting + operational-risk classification + prospective outcome verification + event-driven triggering | availability of piers, terminals, berths and logistics assets exposed to wind/wave conditions, after site-specific adaptation and validation | lead time, event recall, alert burden, calibration, drift, champion/challenger comparison and source robustness | research shadow: not an operational decision, not a substitute for NYC Ferry or competent authorities, and never an automatic close/open command |
Four application families
Periodic intelligence
Collection, normalisation, control, synthesis and publication of updated sources.
Conditions and anomalies
Triggers, thresholds, regimes and cases for human review.
Documentable indicators
Readable outputs that expose sources, uncertainty and reasons.
Reproducible protocols
Versioned datasets, baselines, temporal splits, ablations and outcome reports.
Before reuse
A pattern is not transferred by copying the model.
- define who will use the output and for which decision;
- verify access, licence, quality and cadence of sources;
- build a relevant baseline;
- define block gates and missing-data handling;
- separate prototype, pilot and production;
- document responsibility, maintenance and failure modes.
Operating boundary
Applications described here are transfer hypotheses based on public workflows. They do not imply that a system is already suitable for corporate processes, regulated decisions or production environments.
Applications described here are transfer hypotheses based on public workflows. They do not imply that a system is already suitable for corporate processes, regulated decisions or production environments.
Discuss a possible application
A preliminary discussion only needs a non-confidential description of the problem and sources.