It is the classifying idea behind the Atlas. Every product is placed on one of two panels by asking a single structural question: is the AI the product, or is it embedded inside a product that already existed? Native products are AI-first and organized as a technology stack; Embedded products are incumbents that added AI and are organized by the host software category they live in.

Native AI means the product's core value depends on the models. Remove the models and there is no product, an AI coding agent or a foundation model, for example. Embedded AI means an established platform that predates the AI and runs without it has AI capability layered on top, such as a CRM, analytics suite, or SIEM that added a copilot or agent. The test is dependency: does the product still have a reason to exist with the AI switched off?

Each product is evaluated against five tests: dependency, architecture, origin, commercial model, and interface. Dependency is decisive; the other four resolve ambiguous cases. Every product also gets an agentic level from 0 to 3. Placements are checked for consistency after each update, and the one deliberate edge case is documented in the open.

Agentic level is a 0-to-3 scale recording how autonomously a product acts: 0 is none, 1 is assistive (it helps a human), 2 is agentic (it completes multi-step tasks), and 3 is autonomous (it operates with minimal human involvement). It is an attribute recorded on every product, separate from the Native or Embedded classification.

As of August 2026, the Atlas covers 319 products from 275 vendors across 44 category cells, split into 193 Native and 126 Embedded products, with 607 cited sources. The counts grow as new products are classified.

Most AI landscapes are logo collages: static snapshots with no method for telling an AI-native product from an incumbent that added a copilot. The Atlas applies one consistent, published test to every product, records the rationale and at least two sources for each placement, and is revised continuously rather than reprinted. It is built to be audited, not just admired.

The Atlas is a research project of Arion Research, a global AI and cloud advisory firm. It is built and maintained by Michael Fauscette, Arion's founder and chief analyst, who led IDC's worldwide enterprise software research for nearly a decade and served as Chief Research Officer at G2.

Continuously. New products, agents, and acquisitions are added and re-classified as they ship, rather than frozen in a quarterly report. New classifications and market moves are also covered each week in The Digital Workforce newsletter.

You can browse the interactive map on the site, filtering by panel (Native or Embedded), category, agentic level, and vendor, and open any product to see its classification, the five test results, and its sources. Browsing is open to everyone.

Every product carries at least two cited sources, 607 in total, drawn from vendor documentation, primary announcements, product pages, and independent analyst coverage. The sources are the evidence used at the time of classification, so each placement can be checked.

Placement is editorial and method-driven, not purchased. Vendors and readers who believe a product is missing or misclassified can reach out through the contact page; corrections are evaluated against the same five tests.