Module 2 · The Offering Landscape: Who Plays Where
Lesson 4 of 5 · 4 min · English · Read-only
Four specialized SaaS categories overlap with Symphony: customer support agents, developer-facing agent tooling, voice and conversational agents, and enterprise agent governance. Each shares Symphony's orchestration substrate — tracker-driven dispatch, isolated per-task execution, and observable runs — but optimizes for one surface. This lesson places each category on the comparison matrix and names one shared and one unique capability per category.
Symphony is a long-running automation service that continuously reads work from a configured issue tracker, creates an isolated workspace for each issue, and runs a coding agent session for that issue inside the workspace. Its specification describes it as a scheduler/runner and tracker reader, not a model or a chatbot product. That gives us a baseline of orchestration-level capabilities: tracker-driven intake, per-issue isolated execution, a runner that launches an agent session, workflow policy versioned in the repository, and operator-visible observability.
The four specialized categories in this lesson — customer support agents, developer-facing agent tooling, voice and conversational agents, and enterprise agent governance — each sit on top of or alongside that substrate. They share the orchestration pattern (dispatch work, isolate execution, observe runs) while differentiating on the surface they optimize. The matrix makes this visible: shared capabilities fill the same row across all four columns; unique features appear in only one column.
Source support
Source support
Inputs: The supplied grounding describes GitHub Copilot cloud agent and its Linear integration. Copilot cloud agent can research a repository, create an implementation plan, and make code changes on a branch; you can create multiple custom agents that specialize in different types of tasks; Copilot automates branch creation, commit message writing, and pushing. When you assign a Linear issue to Copilot, it analyzes the issue contents, opens a draft pull request, works independently in its own ephemeral development environment powered by GitHub Actions, streams progress updates back to the Linear activity timeline, and requests a pull request review when work is completed. Task: place this offering on the comparison matrix by naming one shared capability and one unique feature.
Step 1 — Identify the orchestration substrate. The Linear integration reads work from an issue tracker (Linear issues), dispatches it to an agent session, and runs that session in an isolated environment (its own ephemeral development environment powered by GitHub Actions). That matches Symphony's pattern: tracker-driven dispatch plus per-issue isolated execution. So the shared-capability cell is filled: tracker-driven dispatch and isolated execution.
Step 2 — Identify the surface the offering optimizes. The agent's output is code: it researches a repository, creates an implementation plan, makes code changes on a branch, and opens a draft pull request. The surface is the code and API workflow, not ticket resolution, not speech, not policy audit. So the unique-feature cell is filled: code and API workflow surface.
Step 3 — Check the matrix row. The shared row (tracker-driven dispatch) is now filled for the developer-tooling column, matching the other columns. The unique row (code and API workflow surface) differs from the other columns. The matrix now shows one shared and one unique capability for this category.
Source support
The shared substrate exists because every specialized category faces the same operational problems that Symphony's specification names: turning work into a repeatable daemon workflow, isolating agent execution per task, versioning the workflow policy with the code, and providing enough observability to operate multiple concurrent runs. Those problems are surface-independent — a support agent, a code agent, a voice agent, and a governance agent all need dispatch, isolation, and observability.
The surface differs because each category optimizes for a different output. Developer tooling optimizes for code artifacts (branches, pull requests, test results). Support agents optimize for ticket outcomes (resolved, escalated, or handed off). Voice SaaS optimizes for real-time speech turns (latency, turn-taking, audio quality). Governance tooling optimizes for policy and audit (which actions were allowed, what was recorded). The matrix captures this by keeping the shared row constant and letting the unique row vary per column.
This is why specialized agent SaaS usually sits on top of or alongside orchestration platforms rather than replacing them. The orchestration layer handles dispatch, isolation, and observability; the specialized layer handles the surface-specific output. A learner who assumes replacement will misread the matrix: the shared row would look empty, when in fact it is the row that connects all four categories to Symphony.
Source support
Think it over
Pick one of the four specialized categories and describe a feature that would look like a shared capability but actually belongs to the unique surface. What in the feature description tells you which row of the matrix it fills?
Relationship summary
Each specialized category depends on the same orchestration substrate (tracker-driven dispatch, isolated per-task execution, observable runs) while differentiating on the surface it optimizes. The substrate is the shared row on the matrix; the surface is the unique column.
Customer support agents
Optimized for ticket resolution
Developer agent tooling
Optimized for code and API workflows
Voice agent SaaS
Optimized for real-time speech interaction
Enterprise agent governance
Optimized for policy, audit, and observability
Source support