Most GIS tools stop at the data, but Houseal Lavigne is looking to close the gap between that data and the finished deliverable. At this year’s Esri User Conference, the firm demonstrated its answer to this problem: two new AI-powered workflows, Euclid and PlaceEngine, that together form what the firm calls “Place Intelligence.”
Euclid is a jurisdictional knowledge platform that makes a community’s adopted plans, codes, and operational data conversationally available, with every inquiry logged back to the jurisdiction’s own ArcGIS Online. Launched commercially in January 2026 as one of the first AI-native solutions in the Esri ecosystem, Euclid acts as the knowledge layer of Place Intelligence.
PlaceEngine is an AI-native platform that turns GIS data into finished work like reports, maps, visuals, narratives, or presentations and works directly inside tools like ArcGIS Pro and CityEngine. Now in alpha testing and previewing at the conference, it’s aiming to be the production layer of GIS, closing the gap between data and decision-making.
From Renderings to Full Deliverables
The idea for PlaceEngine started small: the team was using AI to enhance CityEngine renderings, turning them into photorealistic scenes with different lighting, seasons, and weather. However, that led to a bigger realization. The 3D models behind those renderings were carrying rich data, and that data could feed an AI pipeline to generate entire deliverables, not just prettier images.
Today, PlaceEngine pulls together disparate sources like GIS layers, spreadsheets, APIs, meeting summaries, reference documents and turns them into finished PDFs, InDesign files, websites, ArcGIS Hub sites, and StoryMaps. To show off the range, the team fed the platform a 2022 plan for a 60-acre site in Colorado and had it produce 14 different deliverables from that single dataset: a pro forma, a marketing prospectus, and, just for fun, a children’s coloring book.
Though built for planning, the team found PlaceEngine’s usefulness extends well beyond it, which led to a planning-specific version called the “SMART Plan.” One example they showed: using Euclid’s zoning knowledge, the team generated a four-page wedding-hosting guide for Madera County, California, sourced entirely from the county’s own zoning code.
Human-in-the-Loop by Design
A theme that stood out throughout the demo was how careful the system is not to overstep. PlaceEngine interprets questions and drafts content, but it stops and asks a person when it hits a real decision point. Zoning answers are always pulled directly from a city’s adopted documents rather than generated from the AI’s general knowledge.
That distinction matters because of how most AI search tools work. Many rely on RAG, or retrieval-augmented generation, a method where a document gets broken into small chunks so an AI can search and pull from the most relevant pieces. It’s efficient, but it has a known weakness: breaking documents apart this way can lose the cross-references that tie sections of a code together, which is why Houseal Lavigne says standard RAG tools often land in the 50–60% accuracy range on zoning questions. Houseal Lavigne built its own architecture instead, which the firm says reaches roughly 98% accuracy — and, when it isn’t confident, the system will decline to answer rather than guess.
Five Specialized Agents
PlaceEngine is built on five AI agents, each handling a different task and running semi-independently to do it. Think of an agent as a focused digital assistant, one that’s been given a specific job, the tools to do it, and the ability to make small decisions along the way without a person walking it through every step. Here, that means: PhotoDesk enhances imagery and generates supporting infographics; Map Room critiques and improves cartography inside ArcGIS Pro; Auteur writes content through three editorial passes, checking for completeness, alignment, and accuracy; Profile handles demographic profiling and geo-enrichment; and Pencil runs deterministic calculations like pro formas, plugging in a user’s own financial models when needed.
The interface is split into two spaces. Studio holds source imagery pulled from ArcGIS Pro, CityEngine, Rhino, SketchUp, and AutoCAD, rendered with realistic context like vegetation, materials, and signage matched to the actual location. Canvas is where deliverables come together, laid out like a mural board, with “lenses” that let users prioritize which data sources the AI should lean on most as it writes.
Who It’s For
This tool is aimed at city managers and planning directors, and the goal is to keep pricing within reach of what they can approve on their own, without needing to go through city council. Euclid launched commercially in January 2026, while PlaceEngine is still in alpha and looking for early adopters ahead of a wider release.
