Reflections from GeoWeek 2026: AI, trust, and the gap between technology and use case
.png)
Last week I had the opportunity to attend GeoWeek in Denver which was three full days of geospatial conversations, hardware demos, AI debates, and some very big ideas about where this industry is heading.
Conferences like this are always a mix of inspiration and validation. You see what’s emerging, what’s maturing, and what’s still trying to find its footing. You also get a clearer sense of where your own thinking sits within the broader landscape.
Here are a few personal reflections that stood out.
GeoAI is moving quickly and trustworthy outputs matter
There was significant focus on Geo AI throughout the conference, not just as a future concept, but as something actively being implemented across the sector.
What stood out to me wasn’t hype. It was momentum.
AI models are becoming more capable, segmentation tools are more refined, and the speed at which spatial datasets can be analyzed is accelerating rapidly. The underlying models are evolving quickly, and the advancements between iterations are substantial.
One particular technical observation stood out to me: many of the open models currently being used have been trained predominantly on ground-based imagery. As more aerial data from drones, manned aircraft, and satellite sources is incorporated into these models, we can expect a significant improvement in spatial performance and reliability. We’re already seeing the impact of this shift in-house at Birdi, where aerial-specific training data materially improves the quality of results.
Alongside that progress, there’s an increasing emphasis on the importance of trustworthy outputs.
Traditional survey-grade data is trusted because it is grounded in known methodologies and standards. As AI becomes more embedded in spatial workflows, the conversation naturally turns toward transparency and defensibility:
- How is that insight formed?
- What data underpins it?
- How repeatable is the outcome?
- How clearly can the result be explained to a decision-maker?
AI in geospatial will only create real value if the outputs are explainable, defensible, and embedded within existing operational workflows. Trust isn’t a barrier, it’s a requirement for adoption.
Gaussian splats vs point clouds vs photogrammetry
Another consistent technical theme was the comparison between visualization technologies — particularly Gaussian splats, point clouds, and photogrammetry.
Gaussian splats are visually impressive. They produce highly realistic renderings and scale efficiently, even across large environments. From a visualization standpoint, they are compelling.
But from a measurement perspective, the accuracy conversation is still evolving.
For volumetric calculations or any workflow requiring high relative and absolute accuracy, point clouds and traditional photogrammetry remain essential. Gaussian splats may be well suited to visual inspection and assessment use cases, but precision measurement still demands rigor.
This isn’t about one technology replacing another.
It’s about aligning the right tool to the right use case.
In an industry filled with innovation, discipline matters. Not every new capability needs to be adopted immediately. The question is always: does it improve the outcome for the end user?
The real gap: you capture data… then what?
One of the strongest themes for me was the gap between:
- Capturing data
- Processing data
- Running AI on data
- And embedding that insight into operational workflows
Many demonstrations showcased impressive outputs. But identifying an object or anomaly in imagery is only the beginning.
What happens next? Who acts on it? How is it validated? How does it integrate into existing systems?
That “then what?” question still feels under-addressed.
A particularly interesting example was a nationwide LiDAR initiative in Scotland, where comprehensive datasets are being captured and made publicly available ( and excitingly potentially on a recurring multi year basis). It’s a significant move toward trusted, accessible spatial data.
But once the data exists, the challenge becomes interoperability and workflow integration. Large volumes of data sitting in storage do not create value. Data that is accessible, contextualized, and embedded in day-to-day operations does.
That transitional space, between capture and practical use, is where much of the industry’s opportunity lies.
Use cases still anchor everything
While much of the conference leaned technical, there were strong examples of geospatial embedded directly into operational practice.
One US general contractor shared how they operate around 70 drones, conducting approximately 15 flights per day and processing close to 100,000 images per month. Their applications span site scanning, earthwork monitoring, utility mapping, quality control inspections, and pre-task planning.
That’s not experimentation. That’s operational infrastructure.
Another perspective that resonated deeply came from the Jane Goodall Institute, emphasizing that more data does not inevitably lead to better outcomes. For data to influence decisions, it must be trusted, contextualized, and filtered through local knowledge and real-world constraints.
More technology does not automatically mean better decisions. Trust and relevance determine whether insight becomes action.
A personal takeaway
GeoWeek reinforced something important for me.
The geospatial sector is not short on innovation. The hardware is advancing. The software is evolving rapidly. AI models are becoming more capable. Compute power continues to scale.
But the real differentiator will not be who has the most technically impressive demo.
It will be who can bridge:
- AI and trust
- Capture and workflow
- Insight and action
Technology in geospatial is accelerating rapidly. That is clear. The opportunity lies in making it usable, defensible, and embedded in the way organizations actually operate.
That’s the lens I’m bringing back from Denver.