Your critical context spans multiple application suites and data types.
Architecture comparison
What each enterprise AI category gives you, and what it leaves to your team.
See what model APIs, suite copilots, search, workflow automation, and an execution layer provide, including the work that remains with your team.
Category map
Different products begin at different layers.
This comparison describes typical architecture patterns, not every vendor implementation. Validate each product against your sources, identity model, actions, controls, and operating needs.
CapabilityModel APIsSuite copilotsAutomation toolsS/Runtime design
Enterprise contextBring your own retrievalApp-specific contextWorkflow inputsShared graph across systems, people, policy, and work
PermissionsApplication must enforceUsually source or suite scopedConnector and workflow scopedIdentity and source policy evaluated across the run
Agent executionDeveloper-builtSuite actionsStrong for deterministic workflowsContext-aware planning with bounded tools and approvals
Model choiceProvider-specificUsually suite-selectedVaries by automation toolTask-level routing designed across approved models
Governance evidenceApplication-ownedProduct activity logsWorkflow execution logsRequest-to-action trace with context and policy evidence
Reuse across teamsRequires platform engineeringBest inside one suiteReuse by workflow templateShared context, tools, policy, and agent patterns
Often evaluated alongsideGleanMicrosoft Copilot StudioGoogle Gemini EnterpriseServiceNowSalesforce AgentforceUiPathModel-provider platforms
Names identify representative evaluation categories only. The table above does not assert feature parity or deficiency for any named vendor; buyers should validate current product scope directly.
Decision framework
Choose S/Runtime when the problem crosses systems and teams.
Permissions and policy must follow the user through every retrieval and action.
The outcome requires reasoning, approvals, and work execution across tools.
You want one reusable intelligence foundation instead of an agent stack per team.
Fit boundary
Do not choose S/Runtime just to add another chatbot.
Stay with a suite-native agent
when the workflow, data, users, and actions live almost entirely inside one application suite.
Build on a model platform
when you have a strong internal platform team and want to own retrieval, identity, policy, evaluation, and operations.
Use deterministic automation
when the process is stable, rules are complete, and reasoning over changing context adds little value.
Choose S/Runtime
when consequential work spans systems, permissions, policies, teams, and reusable enterprise context.
Architecture session
Bring one consequential workflow. Leave with a governed agent blueprint.
We will map the context, systems, decisions, controls, actions, and success measures together.