Agentic AI is reshaping how organizations collect, process and act on geospatial data. But while the hype is everywhere, operational adoption remains elusive for many. LIDAR Magazine spoke with Chancee Vincent, principal AI architect at NV5, about where organizations really are in their AI journey, why agentic workflows matter and models alone are not enough, and how GeoAI will transform lidar and remote sensing over the next decade.
Q: Many organizations are experimenting with AI, but few have operationalized it. From your vantage point, where do geospatial teams stand with regard to their AI maturity today?
Vincent: Most organizations are still in the pilot phase. They’ve tested models, run proofs of concept, maybe automated a task or two. But moving from experimentation to operational intelligence is where things stall.
The challenge isn’t enthusiasm, it’s workflow integration. Geospatial teams often have strong tools, but they live in silos. Analysts still spend hours stitching together datasets, running quality control (QC), harmonizing schemas and pushing results into downstream systems. AI pilots may solve an individual task, but often do not address the broader workflow around it.
Operational GeoAI often requires reasoning, orchestration and cross-platform interoperability. To deliver the most value, geospatial agentic AI must understand intent, select the right tools and execute multi-step workflows across imagery, lidar, GIS and operational systems. That’s the maturity gap most organizations are trying to cross.
Q: What makes that leap from pilot to production so difficult for organizations?
Vincent: There are two key things: complexity and expectations.
Geospatial workflows are inherently multi-step: Find the right data and validate it, run the right analysis and interpret results, then push them into operational systems. Traditional AI focuses on models. But organizations don’t invest in AI because they want a new model, they invest because they want better decisions.
That’s why NV5 emphasizes workflow intelligence. The value isn’t just in a classifier or a model; it’s in connecting AI to enterprise systems, business processes and analyst workflows to deliver actionable insights quickly and easily.
This is why we developed the GeoAI Acceleration Program, which provides a structured path to help organizations assess readiness, prepare data, deploy workflows and maintain AI operations. Without that type of foundation, AI adoption becomes a series of disconnected experiments.
Q: Geospatial professionals often ask whether AI will replace human expertise. How do you address that concern?
Vincent: AI isn’t replacing geospatial experts, it’s augmenting their work and freeing them up to do more high-value work.
Agentic AI can handle repetitive, time-consuming steps, such as QC, feature extraction, catalog search, flight planning, SLAM processing and more. But experts will still need to guide interpretation, validate outputs and refine custom tools.
We call this human-in-the-loop GeoAI. Analysts can author specialized workflows, then let the system operationalize them at scale. Non-experts gain conversational access to complex geospatial data, but experts remain essential for context, nuance and quality.
Q: For lidar-heavy organizations, where does GeoAI deliver the fastest ROI?
Vincent: Anywhere you have massive volumes of data and repetitive analysis.
The biggest early wins we see include:
- Quality assurance (QA) and QC on incoming lidar and imagery
- Asset inventory and feature extraction
- Sorting and triaging large imagery collections
- Coordinating multi-sensor workflows
ROI shows up in hours reclaimed, which frees analysts to focus on high-value work. When you factor in analyst wages, project load and throughput, compressing a 60-minute workflow into a few minutes isn’t just convenient; it’s transformative.
Q: You’ve emphasized the difference between “models” and “workflows.” Why is that distinction so important?
Vincent: Models are powerful, but a model by itself is still only one component of an operational workflow. They classify, detect or segment. Workflows are where enterprise value lives. A workflow understands which data to trust, which tools to run and in what order, and where results need to go.
That’s the power of agentic AI. It connects models to business processes.
Q: NV5 often describes its generative AI solution GeoAgent™ as “technology agnostic.” Why is that important for lidar and sensing teams?
Vincent: Because no organization wants to abandon the ecosystems they already trust. Our approach is to ensure that generative AI can integrate with ArcGIS, Trimble Unity, Planetary Computer, our ENVI image processing and analysis solution, and more. We don’t want AI to replace your existing stack, we want it to orchestrate across it. That orchestration carries the data, parameters and geospatial context needed to coordinate each step across the different systems.
This matters for lidar because workflows rarely live in one place. You may process point clouds in one place, manage assets and visualize results in other systems, and store data in cloud catalogs. A technology-agnostic generative AI agent ensures results flow seamlessly from analysis to action.
Q: How is agentic AI changing lidar and remote sensing today?
Vincent: Agentic AI is starting to change the parts of lidar and remote sensing workflows that sit around the model or tool itself. A lot of AI discussion has focused on classification or feature extraction, but the more operationally challenging issues typically center around:
- Which dataset is the right one?
- Does it cover the area of interest?
- Is it current enough?
- What processing path should be used?
- What parameters are safe?
- Where should the output go once it is created?
A GeoAI agent can help turn that into a repeatable workflow instead of a series of manual handoffs. For lidar, that could mean finding the right point cloud or derived surface, checking coverage and metadata, running the appropriate processing step, comparing against prior conditions, and producing a reviewable layer or summary.
Q: How will agentic AI change how geospatial analysts work in the future?
Vincent: I think the analyst role will move closer to workflow direction and review. The analyst still owns the judgment, especially in lidar where quality, classification, density, collection conditions and local context matter. The difference is that the analyst should not have to manually rebuild the same path every time.
A good agentic workflow can propose the data, run the first-pass analysis, expose the steps it took and stop at review points before anything becomes authoritative. Human-in-the-loop should mean practical review gates, including:
- Approving the input data and the tool path
- Inspecting flagged areas
- Adjusting thresholds
- Deciding whether the output is good enough to publish or send downstream
Q: What workflows are possible today that weren’t realistic even two years ago?
Vincent: The realistic change is chaining remote sensing, GIS and delivery into one workflow from a plain-language request. For example: “Show me where this corridor changed since the last collection and which assets need review.”
Today that request can be broken into executable steps: identify the area of interest, find the relevant lidar or imagery, clip the data, run change or classification logic, intersect the result with asset layers, and publish a map layer or report for review.
Two years ago, each of those pieces existed somewhere, but making them work together usually required a custom script. The newer opportunity is making that pattern repeatable across projects without pretending the agent is the expert. The agent handles the orchestration. The analyst handles the judgment.
Q: Looking ahead, what excites you most about what lies ahead for agentic AI in geospatial in the next two, five, even 10 years?
Vincent: In the near term, I’m most excited about practical, repeatable workflows that save real time without overpromising autonomy. In five years, I think the bigger shift is cross-system geospatial work where remote sensing data, GIS layers, asset systems and reporting tools can be connected through the same workflow.
Looking ahead 10 years, I believe the most interesting outcome will be persistent geospatial understanding: systems that track what was collected, what changed, what was reviewed and what still needs attention. Lidar and remote sensing data should become part of a living operational picture that can be questioned, updated and acted on as conditions change. 
Chancee Vincent is Principal AI Architect at NV5, where he leads the development of enterprise-grade geospatial AI and agentic systems that help organizations transform complex geospatial data into actionable intelligence. Drawing on more than a decade of experience in geospatial intelligence, GIS and advanced imagery analysis (including support for U.S. Special Operations Forces and the National Geospatial-Intelligence Agency) Chancee specializes in integrating AI into operational workflows. His work focuses on building scalable, secure GeoAI solutions that accelerate decision making across government and commercial organizations.