The hardest part of working with geospatial data isn't the analysis. It's the query. Finding every park with a water feature in San Francisco, or every ice cream shop within 500 meters of a streetcar line in Berlin, requires stitching together disparate public datasets, understanding their coordinate systems, and writing complex spatial joins. DataMonkey, a Berlin-based startup, is betting that a large language model can be taught to do that work for you. Its core product is a GeoAI agent that accepts natural language prompts and returns structured, location-based datasets, aiming to turn a niche technical skill into a simple conversation.
Founded in 2022 by CEO Irene Ternes and CTO Paul von Berg, DataMonkey operates with a small team of four under the legal entity Urban Monkeys GmbH [RocketReach, retrieved 2026]. The company describes its mission as making AI-based geographic data "as easily accessible as possible for business experts, data analysts, or IT teams" [DataMonkey, retrieved 2024]. While the broader market is crowded with mapping APIs and business intelligence tools, DataMonkey's wedge is specificity: it focuses on blending proprietary location data with open-source geospatial databases like OpenStreetMap, then letting users interrogate that combined dataset in plain English.
The technical wedge: From SQL to sentences
The company's public claims center on a concept it calls GeoAI, which it defines as "combining large language models with geospatial intelligence to make map-based data more intuitive and insightful" [ffplus-project.eu, 2026]. In practice, this means training LLMs to understand relationships in the physical world,proximity, containment, and adjacency,and to translate those relationships into executable geospatial queries. A user asking for "water in parks in San Francisco" would trigger a process where the agent identifies relevant OpenStreetMap tags for parks and water features, performs a spatial intersection, and returns the results as a clean table or visual overlay. This moves the interface from writing code to describing a need, a significant reduction in friction for non-specialist users.
A market of public data and private insight
DataMonkey's potential customers are organizations that need to make decisions based on location but lack dedicated geospatial teams. Urban planners, retail site selection analysts, logistics operators, and environmental researchers all work with questions that are inherently geographic. The startup's bet is that these users are currently underserved by generic BI tools, which treat location as just another column, and overwhelmed by the complexity of professional GIS software. By acting as a translator between human intent and geospatial databases, DataMonkey aims to own a new slot: the conversational interface to the mapped world.
The company's early positioning and technical focus suggest a product-led growth strategy, targeting technical individual contributors who can adopt the tool without a lengthy procurement cycle. The lack of disclosed funding or named investors points to a bootstrapped or very early-stage venture, which aligns with its current headcount and focused development on core AI capabilities rather than broad sales and marketing.
What could go wrong at scale
While the concept of a GeoAI agent is compelling, scaling it presents several technical and commercial hurdles. The accuracy of the system is entirely dependent on the quality and freshness of the underlying geospatial data it accesses; stale or incorrectly tagged OpenStreetMap data leads to incorrect answers. Furthermore, natural language is famously ambiguous. A query for "near" needs a defined radius, and "parks" could mean anything from a national forest to a pocket playground. The model's ability to disambiguate intent and ask clarifying questions will be critical.
From an infrastructure perspective, the cost of running inference on large language models for every user query is non-trivial. As usage scales, DataMonkey will need to optimize its model serving and potentially develop smaller, specialized models to keep latency low and costs manageable. The competitive landscape is also indirect but vast. Established players like Google (BigQuery GIS), ESRI, and Mapbox offer powerful programmatic tools, while a host of no-code analytics platforms are adding more sophisticated mapping features.
- Data fidelity risk. The product's utility is gated by the accuracy of open-source geospatial databases, which can be incomplete or inconsistently maintained [DataMonkey, retrieved 2024].
- Query ambiguity. Natural language prompts for location are often vague ("near," "walkable"), requiring robust disambiguation logic that must be built, not inherited from a general-purpose LLM.
- Compute economics. Serving real-time, complex geospatial queries powered by LLM inference could become cost-prohibitive before achieving sufficient pricing power.
The next twelve months for DataMonkey will likely focus on proving two things: that its AI agent can reliably handle a wide array of real-world queries without constant human correction, and that organizations are willing to pay for this abstraction. The team's small size is an asset for focused iteration but will soon face the classic startup tension between refining the core technology and building the sales motion required to move beyond early adopters. If they can navigate that, they may well define a new category for how businesses ask questions about where things are.
Sources
- [DataMonkey, retrieved 2024] DataMonkey - Ready for map-tastic insights? | https://www.datamonkey.tech/
- [ffplus-project.eu, 2026] Interview with Irene Ternes, Co-Founder and CEO at DataMonkey | https://www.ffplus-project.eu/en/multimedia/blog/interview-with-irene-ternes-co-founder-and-ceo-at-datamonkey/
- [RocketReach, retrieved 2026] Urban Monkeys GmbH employee data | https://rocketreach.co/urban-monkeys-gmbh-profile_b5c4c4f9f42e4a6d
- [LinkedIn, retrieved 2026] Paul von Berg profile | https://www.linkedin.com
- [Prospeo, undated profile] Prospeo profile for DataMonkey | https://prospeo.io/c/datamonkey