al Transformation Summit 2022 Session 1

### Overview Chaired by Mark Steven, this opening session of the 8th Annual Digital Transformation Summit brings together three distinct perspectives on digital transformation, technology adoption, and organizational impact. The presentations and subsequent panel Q&A focus on keeping human needs at the center of change, overcoming operational hurdles in data science, and leveraging open satellite earth observation data for environmental and business intelligence.

### Key Topics

#### Human-Centric Transformation and Tech for Good Graham Johnson (CEO of Retail and Distribution at The Donaldson Group) argues that digital transformation should fundamentally serve people rather than technology for its own sake. Drawing on his experience, he outlines core transformation lessons: treating programs as continuous evolution rather than rigid, one-off changes; avoiding isolated "VIP clubs" of top talent that pull domain experts away from day-to-day realities; winning "hearts and minds" instead of imposing unrealistic timelines; and focusing on what new tools enable for colleagues and customers rather than simply what systems they replace. Highlighting digital inequality—where only 44% of a representative 100-person global village has internet access—he calls for an "Industry 6.0" centered on humans and social equity. He details the evolution of the Scottish Tech Army into the Tech for Good Alliance, an initiative addressing social justice, digital inclusion, skills development, and climate sustainability by mobilizing skilled tech volunteers for the third sector.

#### Practical Realities of Data Science and MLOps Leanne Fitzpatrick (Director of Data Science at the Financial Times) discusses the organizational realities of delivering impact through data science, machine learning, and AI. She warns against unrealistic expectations where organizations assume models immediately create business value without solid data infrastructure. Impact requires balancing technical execution (dependency management, scaling, Version Control, observability, and MLOps pipelines) with an open team culture (trust, curiosity, peer review) and organizational embedding. Fitzpatrick highlights the challenge of calculating return on investment (ROI) in ML, advocating that teams map technical metrics (such as NLP accuracy or maximum likelihood estimation) directly to core business KPIs like engagement and bottom-line revenue. She also emphasizes deprecating obsolete projects and fostering cross-functional skills such as change management and data governance.

#### Satellite Data, Earth Observation, and Natural Capital Professor Edward Mitchard (Co-founder and Chief Scientific Officer at Space Intelligence) examines the vast, underutilized repository of free satellite data provided by agencies such as the European Space Agency (ESA). He details three core remote sensing modalities: Optical satellite imagery (e.g., Sentinel-2), which captures spectral bands but is impeded by night and persistent cloud cover; Synthetic Aperture Radar (SAR, e.g., Sentinel-1), which penetrates clouds and night to measure surface roughness, moisture, and forest biomass; and spaceborne LiDAR (e.g., NASA's GEDI on the International Space Station), which provides precise tree height measurements along orbital tracks. Combining these datasets with machine learning enables continuous mapping of forest biomass, carbon sequestration, and land-cover change. Mitchard explains how this intelligence underpins nature-based carbon credits, Scottish Government natural capital accounting (such as the Natural Capital Asset Index), regulatory compliance (e.g., zero deforestation, Biodiversity Net Gain), and verifiable ESG metrics.

### Technical Details - **Remote Sensing and AI Pipelines**: Space Intelligence processes terabytes of multi-sensor data in AWS and GCP. Optical (10 m resolution Sentinel-2), SAR microwave backscatter (Sentinel-1), and LiDAR photon arrival timings (GEDI 25 m footprints) are cleaned, aligned, and merged using machine learning models to interpolate biomass maps and produce land-cover classifications (including 22 distinct Scottish land categories). - **Data Architecture Foundations**: Successful ML requires an underlying hierarchy: structured data platforms and pipelines (data engineers), business intelligence/reporting, analytics, and finally predictive/prescriptive data science and MLOps at the apex.

### Takeaways Successful digital transformation relies on robust technical foundations, cultural alignment, and user-focused outcomes. Whether mobilizing skilled volunteers for charities, connecting ML models to operational decisions, or turning satellite feeds into environmental dashboards, technology acts as an enabler rather than an end in itself.