Patterns are defined by behavioral boundaries — what qualifies as each pattern versus an adjacent one — established by CoDomain’s technical fingerprint layer. The line between an agent and a chatbot, or between multi-agent orchestration and sequential chaining, is not semantic. It is structural, and it determines which frames a deployment belongs in.
Customer-Facing Workflow Agents
Customer-facing workflow agents are embedded directly in the products and workflows your end users interact with. They act on behalf of the customer — resolving issues, guiding decisions, completing transactions, or personalizing experiences — within a defined scope. This pattern generates some of the strongest commercial signal in the corpus. It is also where the gap between vendor claims and practitioner experience tends to be widest. CoDomain tracks where operator conviction is forming versus where capital is running ahead of validated deployment. What qualifies: An agent embedded in a customer-facing product surface that takes goal-directed action on behalf of the user, with memory or context persistence across at least part of the interaction. What does not qualify: A chatbot that answers questions from a static knowledge base without taking action or persisting context.Developer Agents
Developer agents augment or automate the workflows of software engineers, data scientists, and technical practitioners. They assist with code generation, debugging, testing, documentation, code review, and increasingly with end-to-end task completion across the development lifecycle. This is one of the most active patterns in the corpus by practitioner discourse volume. Field sentiment here is detailed and technical — developers write about what works and what fails with unusual specificity, which makes this pattern one of CoDomain’s higher-confidence signal areas. What qualifies: An agent operating within a development workflow that takes multi-step action — generating, reviewing, running, or modifying code — with some degree of autonomous execution. What does not qualify: A code completion tool that produces suggestions inline without executing, reasoning across context, or taking action beyond the autocomplete surface.Horizontal Do-Anything Agents
Horizontal do-anything agents are general-purpose agents without vertical specialization. They are positioned to handle a broad range of tasks across domains — marketed as capable of operating wherever a user points them. This pattern carries significant analytical weight in CoDomain because it is where capital and field sentiment most frequently diverge. Broad positioning is easy to fund and hard to validate. CoDomain tracks where horizontal agents are building genuine field conviction versus where spend is running ahead of demonstrated reliability. What qualifies: An agent explicitly positioned for general-purpose task execution across domains, without a defined vertical or workflow specialization. What does not qualify: A vertically specialized agent marketed with broad language. Positioning language alone does not determine pattern classification — the behavioral boundary does.Internal Ops Agents
Internal ops agents are deployed for internal business operations — finance, HR, IT, procurement, legal operations, and back-office workflows. They do not touch the end customer directly; they automate, assist, or orchestrate internal processes. This pattern often shows stronger practitioner conviction than customer-facing patterns, because internal deployments face lower regulatory scrutiny and faster feedback loops. CoDomain tracks where internal ops adoption is outpacing infrastructure readiness and where it is consolidating into durable workflow integrations. What qualifies: An agent deployed inside an organization to automate or assist internal workflows, with no direct customer-facing surface. What does not qualify: An internal tool that routes requests or surfaces information without taking autonomous action across a workflow.Multi-Agent Orchestration
Multi-agent orchestration covers coordinated systems of multiple specialized agents working together toward a shared goal. Rather than a single agent handling an end-to-end task, orchestration patterns involve routing, delegation, specialization, and coordination across agent boundaries. This is the most technically complex pattern in the map, and the one where infrastructure signal is most consequential. The architecture required to run reliable multi-agent systems reliably is still maturing — CoDomain tracks where deployment ambition is outrunning the orchestration primitives available to support it. What qualifies: A system in which two or more agents with distinct roles or capabilities coordinate, with a defined orchestration layer managing routing, context handoff, or task delegation. What does not qualify: A single agent with multiple tools or a sequential pipeline without coordination logic between agent boundaries.How Patterns Relate to Industries and Infrastructure
Deployment patterns are one axis of the CoDomain map. Each pattern is measured across industries — so you can see where, for example, internal ops agents are gaining ground in financial services versus healthcare — and against infrastructure signal that reveals whether the underlying architecture supports the deployment conviction being funded.Industries
See how each deployment pattern maps across the 10 industry verticals CoDomain currently covers.
Technology Infrastructure
Understand the infrastructure signal layer — the leading indicator for where deployment patterns are heading next.