Geo for the Agentic Era: Highlights from #SDSC25 London
On May 15th, over 400 data analysts, GIS professionals, and geospatial developers gathered at the Royal Geographical Society for this year’s Spatial Data Science Conference in London (#SDSC25). It was clear from the start: we're entering a new era of GIS.
From AI agents and open data formats to climate monitoring and advertising optimizations, leaders from across the industry gathered to share their takes on the challenges and opportunities as geospatial faces one of its biggest reinventions to date.
Missed out? Keep reading for our summary of the day, or watch the sessions on-demand here.
Kicking off the day, Javier de la Torre, Founder & CSO at CARTO, delivered a call to action: it’s time to rebuild GIS for the agentic era.
Agentic AI is transforming geospatial analysis from static dashboards to dynamic, conversational systems that interpret context, refine questions, and surface insights proactively. This shift makes spatial intelligence accessible to anyone - not just specialists - by enabling natural language interaction with data. CARTO AI Agents, now in public preview, are leading this change, letting users “talk to their maps” and unlock real-time insights with ease.
Beyond AI, Javier also spoke about how open table formats are making spatial data a “first class citizen” in the wider data space. Some of the key developments in this space are:
- GeoParquet & Apache Iceberg: Together they bring standardized storage, time travel, and multi-engine access to spatial data - eliminating duplication, boosting performance, and avoiding vendor lock-in. With vector data now fully supported in these formats - and raster data well on its way - there remain two key types of spatial data to make their cloud-native journey:
- Trajectory data: In the coming months, CARTO is enabling native support for movement and mobility datasets, allowing seamless ingestion, analysis, and visualization of location over time.
- Digital twins: While still early-stage, efforts to standardize large, automatically collected formats like LiDAR and point clouds are essential to integrating 3D spatial data into mainstream analytics platforms.
- Trajectory data: In the coming months, CARTO is enabling native support for movement and mobility datasets, allowing seamless ingestion, analysis, and visualization of location over time.
Wishing you’d caught the keynote? Watch it in full below!
One of the loudest themes across sessions? Spatial data is fragmented - and it’s costing us.
Fawad Qureshi (Global Field CTO, Snowflake) called out the "hidden tax" of siloed spatial formats and stressed the need for common frameworks like H3 and GTFS to enable seamless collaboration across industries. His session made a strong case for standardizing access to rasters, vectors, point clouds, and more. Meanwhile, Stuart Lynn (Senior Specialist Solutions Architect, Databricks) emphasized how the lakehouse architecture is helping organizations like Landmark and Natural England run geospatial ML workflows, from peatland restoration to coastal proximity modeling.
Across the board, the message was clear: to unlock real spatial intelligence, we need open formats and integrated tools that speak the same language.
Every year at SDSC, scale is a key theme. Organizations are continuing to push the boundaries of using more data to make decision-making an even more scalable process.
This year, the scale was bigger than ever.
Gennadii Donchyts (Cloud Geographer, Google Cloud) shared how they're building geospatial pipelines at a planetary scale using BigQuery, Earth Engine, and open formats like GeoParquet - enabling organizations to process and query massive datasets with ease. Dean McCormick (Account Manager, AWS) showcased how real-time geospatial intelligence at planetary scale is now a reality, powered by S3 pipelines, LLMs, and AWS Ground Station.
But scalability isn’t just about big data - it’s about scaling skills, access, and decision-making. Gerardo Martín Carreño (Full Stack Spatial Data Scientist at TJX Companies) showed how moving from desktop GIS to a modern spatial stack helped democratize insights across their business, reducing analysis time from weeks to hours and empowering more teams to make spatial decisions.
For all the buzz around new technology and data, SDSC25 had a grounded undercurrent: the fundamentals of geography will always be crucial. Tobler’s First Law (“everything is related to everything else, but near things are more related than distant things”) was referenced more than once, although this SDSC may be the first time that John Snow was not cited!
Some examples of this in action were Alfie Long (Data Scientist, BT Group) used spatial autocorrelation to reveal how footfall behavior varies across regions, and Dr. Patrick Ballantyne (Postdoctoral Researcher, University of Liverpool) who shared his work on using decision frameworks to improve pedestrian safety planning.
The rise of agentic AI marks a major shift for geospatial, but one thing we kept hearing from speakers was that they don’t expect AI to replace geospatial expertise - but to amplify it.
As AI takes on more of the routine analysis, the role of the spatial expert is evolving - from building dashboards to building geospatial agents, and from running models to designing how AI asks and answers questions.
The message from SDSC25? AI may be the future - but it’s the geospatial community that shapes what it delivers.
SDSC25 reflected a moment of transition - not just in tools and technology, but in how we approach spatial thinking itself. As AI reshapes workflows and data infrastructures evolve, the core goals remain the same: making better decisions and connecting people with the insights that matter.
Couldn’t make it to SDSC? Want to rewatch a talk or share it with your team? Watch the full conference on demand here! You can also follow SDSC on LinkedIn here to stay up to date with news, ticket drops and calls for papers!
.png)
