AI Data Pipelines in Stockholm

Manufacturing / Industrial Engineering

A Swedish manufacturing group (SEK 180B annual revenue, 45,000 employees across 35 countries, producing industrial tools and equipment for mining, construction, and manufacturing) needed unified sustainability data infrastructure for CSRD compliance and climate strategy execution. (1) Carbon data fragmentation: scope 1 and 2 emissions data existed across 120 production sites in 35 countries — in different formats (energy invoices, fuel receipts, building management system exports, manual meter readings), different units (kWh, MJ, BTU, therms), different emission factor databases, and different reporting calendars. Consolidating this data for group-level carbon reporting required 3 months of manual effort by the sustainability team (18 people), with a 12% error rate due to unit conversion mistakes, missing data, and double-counting. (2) Scope 3 was largely unmeasured: the group's scope 3 emissions (supply chain, product use phase, end-of-life) were estimated at 85% of total emissions — but actual measurement was limited to tier-1 supplier surveys (32% response rate) and generic emission factors. Accurate scope 3 data was essential for credible SBTi target-setting and CSRD disclosure. (3) EU Taxonomy mapping was manual: determining taxonomy alignment required mapping each business activity against technical screening criteria — assessing substantial contribution, DNSH (Do No Significant Harm), and minimum safeguards. The group had 2,400+ product lines and 15 business divisions — manual taxonomy mapping had covered only 40% of revenue after 8 months of work. (4) Circular economy tracking was ad hoc: the group had ambitious circularity targets (50% recycled materials by 2030, product-as-a-service for 30% of revenue) — but material flow data, reverse logistics data, and refurbishment metrics were scattered across ERP systems, logistics systems, and production databases without unified tracking. (5) CSRD double materiality assessment lacked data infrastructure: the CSRD's double materiality requirement (assessing both the company's impact on sustainability matters AND sustainability matters' impact on the company) required integrating financial data with sustainability data — but these existed in completely separate systems with no pipeline connecting them.

We built a unified sustainability data pipeline over 14 weeks. (1) Carbon data unification: automated ingestion from 120 production sites — energy invoices (OCR extraction from PDF), building management systems (API integration), fuel purchase records (ERP integration), and refrigerant data (maintenance system integration). Unit normalisation: all energy data converted to standardised units (kWh primary, with conversion to tCO2e using site-specific emission factors from IEA, DEFRA, and local databases). Automated scope 1 and 2 calculation: real-time carbon footprint tracking by site, country, and group — updated daily (from annually). Data quality monitoring: automated validation for completeness (are all 120 sites reporting?), consistency (are values within expected ranges for each site's activity level?), and accuracy (cross-validation against financial data — energy costs vs. energy consumption). (2) Scope 3 data pipeline: supplier data collection (automated sustainability questionnaire distribution via supplier portal — structured data collection replacing ad-hoc survey emails), supply chain emission calculation (combining supplier-specific data with spend-based estimation using EEIO models for suppliers without specific data — creating a hybrid scope 3 calculation that improved accuracy as more suppliers provided specific data), product use-phase modelling (integrating product specifications, customer usage data from connected products, and energy consumption models — calculating use-phase emissions for the group's product portfolio), and end-of-life estimation (modelling disposal, recycling, and recovery based on product material composition and regional waste management data). Scope 3 data quality score: improving from 22% (survey-based estimates only) to 68% (hybrid approach) — with a roadmap to 85% as supplier-specific data collection matured. (3) EU Taxonomy automation: activity mapping engine — automatically classifying the group's revenue, CapEx, and OpEx against EU Taxonomy economic activities using product catalogue data, financial data, and production data. Substantial contribution assessment: automated evaluation against technical screening criteria (climate change mitigation targets, energy efficiency thresholds, material circularity criteria) — using production data and engineering specifications. DNSH assessment: automated screening against DNSH criteria for each eligible activity. Taxonomy coverage: from 40% (8 months manual) to 92% (automated) within 6 weeks. (4) Circular economy tracking pipeline: material flow data (input materials by type, source, and recycled content), production waste data (waste by type, destination — recycling, recovery, disposal), reverse logistics data (products returned, refurbished, remanufactured), and product-as-a-service metrics (subscription revenue, utilisation rates, product lifecycle data). Circularity dashboard: tracking progress against 2030 targets — recycled material percentage, product-as-a-service revenue share, waste reduction. (5) Double materiality data integration: financial-sustainability linkage — connecting carbon data with financial data (carbon cost exposure, stranded asset risk), supply chain sustainability data with procurement financial data (supplier sustainability risk weighted by spend concentration), and climate scenario data with business planning data (physical risk and transition risk assessment under different warming scenarios).

Carbon reporting: from 3 months manual effort to 2 weeks automated — including validation and audit preparation. Error rate: from 12% to under 1%. Real-time tracking: daily carbon footprint updates (from annual) — enabling operational teams to see the emission impact of production decisions in near-real-time. Scope 3 measurement: data quality from 22% to 68% — providing a credible basis for SBTi target-setting. The group submitted near-term SBTi targets validated by the SBTi initiative — a milestone that had been delayed by insufficient scope 3 data. EU Taxonomy: alignment coverage from 40% to 92% of revenue — completed in 6 weeks (from 8 months of manual work). Taxonomy-aligned revenue: 34% (better than the 28% estimated from manual assessment — the automated system captured eligible activities the manual process had missed). Green financing: the improved taxonomy data enabled the group to issue its first sustainability-linked bond — SEK 5B at 25bps below conventional pricing (saving SEK 12.5M annually in financing costs). Circular economy: unified tracking enabled data-driven circularity initiatives — recycled material content increased from 28% to 35% in the first year (accelerated by visibility into material flows that identified quick wins). CSRD compliance: ESRS-compliant sustainability disclosure prepared with 60% less effort than projected — the pipeline automated 75% of data collection and calculation. Carbon reduction: real-time tracking enabled operational carbon reduction initiatives — 8% absolute emission reduction in the first year (partly from operational efficiency and partly from visibility-driven behaviour change). Annual value: SEK 450M (carbon reduction — avoided carbon costs at projected ETS prices), SEK 185M (EU Taxonomy uplift — higher taxonomy-aligned revenue percentage improving access to ESG investment capital), SEK 125M (green financing — lower cost of capital from sustainability-linked instruments), SEK 85M (circular economy — cost savings from increased recycled material use and reduced waste), SEK 45M (reporting efficiency — reduced sustainability team effort, redirected to strategic initiatives) = SEK 890M. Development cost: SEK 4.2M. Payback: under 2 days.