The Mixed Agent Enterprise
← Publication ContentsThe enterprise will develop a mixed agent environment.
It will contain internally built agents, enterprise software agents, cloud platform agents, model provider agents, specialist services, open source components, and agents created by functions and teams close to the work.
This environment will grow through ordinary business activity.
Customer service platforms will add agentic capabilities. Engineering teams will build internal execution systems. Finance teams will introduce workflow agents. Software vendors will embed new forms of machine participation into their products. Business units will adopt specialist capabilities for their own operating needs.
The resulting enterprise will contain many computational actors.
Each will bring a different model, runtime, tool environment, identity pattern, memory approach, policy model, economic profile, and owner.
The challenge is to make this diversity coherent.
The enterprise can decentralize agent creation while maintaining accountability for enterprise outcomes.
Agent diversity follows enterprise reality
Large enterprises already operate through a diverse technology landscape.
They have systems of record, domain applications, shared platforms, external services, local tools, data products, workflow engines, and partner connections. Each exists because a function, process, regulatory condition, customer need, or technical decision created value through specialization.
Agentic AI enters this environment.
A single standard agent platform may support some forms of work. It will sit alongside the agents embedded in customer platforms, productivity suites, cloud environments, enterprise applications, and specialist services.
This diversity can create strength.
Agents built close to a domain can understand its workflows, vocabulary, measures, evidence, and operating conditions. They can serve a specific business outcome with greater precision. They can evolve with the people who own the work.
The enterprise gains value when it combines local domain expertise with shared control and architectural coherence.
Interoperability begins with clear boundaries
The mixed agent enterprise requires clear boundaries.
Each agentic participant needs a defined role. It needs an accountable owner. It needs a visible set of authorities, tools, data sources, and operating conditions. It needs a way to exchange work and evidence with other participants.
The enterprise can define a common work contract around these boundaries.
| Work contract element | Operating purpose |
|---|---|
| Outcome | Identifies the result the agentic capability exists to produce |
| Owner | Connects the work to accountable business and technical leaders |
| Authority | Defines the actions, thresholds, and escalation conditions |
| Context | Identifies the enterprise information that informs execution |
| Evidence | Captures the sources, actions, approvals, and resulting conditions |
| Interface | Enables the exchange of requests, responses, status, and completion signals |
| Economics | Connects the work to cost, capacity, and business value |
This contract enables coordination across different technical environments.
A customer service agent can request a supply chain assessment. A supply chain capability can return evidence, conditions, and available actions. An engineering agent can prepare a response to a production issue. A finance workflow can assess the commercial consequence.
Each participant contributes its domain capability. The work contract keeps the broader outcome coherent.
Ownership travels with the work
Ownership becomes especially important when agents coordinate across functions.
A delayed order may involve Customer Service, Operations, Finance, Supply Chain, and a carrier partner. Each function can own a valid part of the workflow. Each may operate through different systems and measures.
The agentic execution system needs to preserve this distribution of ownership.
The customer service function may own customer communication and remedy. Operations may own fulfillment conditions. Finance may own commercial thresholds. Technology may own execution reliability and system integration. A senior leader may own the final decision where the consequence crosses several functions.
The architecture can make these relationships visible.
It can show which agent carried which work, which authority applied, which function owned the decision, which systems provided evidence, and which outcome the enterprise achieved.
This avoids the common pattern in which an apparently unified experience hides fragmented accountability.
Shared controls support local execution
The mixed agent enterprise benefits from shared controls.
Identity, authority, observability, evaluation, cost attribution, incident response, lifecycle management, and intervention rights create a common operating language across the agent population.
These capabilities belong in the enterprise control plane.
They allow domain teams to build and adopt agents that fit their work while providing leaders with a consistent view of the operating environment.
A platform team can provide shared services for identity, policy, traces, evaluations, and cost. Functional teams can define the outcomes, workflow conditions, and acceptance criteria that give the agentic system business purpose.
This composition creates a balanced model.
Domain ownership remains close to the work.
Enterprise governance remains close to the shared risks, dependencies, and opportunities.
Coordination becomes a strategic capability
The mixed agent enterprise creates a new form of coordination challenge.
Agents can exchange requests, context, evidence, and work across systems. They can identify dependencies, prepare decisions, carry actions across functional boundaries, and route exceptions toward the appropriate owners.
The quality of this coordination depends on the quality of the underlying enterprise architecture.
Clear business definitions, reliable data, consistent identities, visible authority, strong interfaces, and accountable ownership create the conditions for productive machine participation.
The enterprise gains a more adaptive operating environment.
Work can move toward the best available participant, whether that participant is a person, a deterministic service, a domain agent, or a coordinated group of agents and humans.
The objective is coordinated execution around an outcome.
The executive question
The executive question is:
“What must remain coherent as agentic capability expands across our enterprise?”
The answer includes outcomes, authority, ownership, evidence, interfaces, economics, and intervention.
The mixed agent enterprise will thrive through this coherence.
It will allow specialized capabilities to grow close to the work while preserving the shared operating principles that give the enterprise confidence in its own execution.
That is the path from agent diversity to enterprise capability.