Hybrid AI Workflows: Choosing the Right Pattern for AI and Workflows - Workflow Automation CoE

Lisa Holenstein · ‎01-09-2026

Workflow Automation CoE > Hybrid AI Workflows

Goal

In this Workflow Automation Center of Excellence article, you will learn how to choose the right automation pattern when combining AI capabilities with deterministic workflows. You will understand when to build AI-first workflows that leverage flows as tools versus when to build deterministic workflows that call AI Agents for specific steps.

Technology

Workflow Studio is our central, low-code workflow automation builder that consolidates flow, subflow, playbook, decision table, and action design into a single environment.

The following product features enable hybrid workflow automation:

Feature	Description
Execute Skill action	Flow action that invokes a published skill from Now Assist Skill Kit
Use an AI Agent action	Flow action that invokes a published AI agent from AI Agent Studio
Use AI agent as an activity	Playbook activity that invokes a published skill built in Now Assist Skill Kit
Use an AI skill as an activity	Playbook activity that invokes a published AI agent built in AI Agent Studio
Agentic Playbooks	AI agents help research and complete playbook activities
AI Agent Studio	Design environment for creating and managing AI agents
Now Assist Skill Kit	Design environment for creating custom GenAI skills

Understanding the Two Patterns

ServiceNow now offers two approaches for combining AI with workflow automation. Understanding when to use each pattern is critical for successful implementation.

Pattern 1: AI-First Workflows (Agent as Orchestrator)

In this pattern, an AI Agent serves as the orchestrator, reasoning through requests and calling flows and subflows as tools to accomplish tasks.

How it works: The AI Agent receives a request, reasons about which tools (including flows) to use, creates a plan, and executes by calling those tools. After each tool invocation, the agent can modify its plan and repeat until the desired outcome is reached.

Characteristics:

Aspect	Description
Orchestration	AI Agent decides what to do and when
Flexibility	High; agent adapts based on context and results
Human involvement	Agent decides if and when to prompt a human
Predictability	Lower; outcomes may vary based on reasoning

Pattern 2: Deterministic Workflows with AI Steps (Workflow as Orchestrator)

In this pattern, the deterministic workflow remains the orchestrator, with AI Agents or Skills called at specific steps where adaptive intelligence is needed.

How it works: The workflow defines the sequence of steps, conditions, and human interactions. At designated points, the flow calls an AI Agent or Skill for tasks requiring reasoning and/or non-deterministic behavior, then continues with deterministic logic.

Characteristics:

Aspect	Description
Orchestration	Workflow controls sequence and decisions
Flexibility	Targeted; AI flexibility only where specified
Human involvement	Workflow explicitly defines where humans participate
Predictability	High; workflow structure ensures consistent paths

Academy Video

If you want to see the deterministic-first hybrid AI automation pattern in action, check out the AI Academy session on YouTube!

Decision Framework: When to Use Each Pattern

Use the following questions to determine the right pattern for your use case.

Choose AI-First Workflows (Agent as Orchestrator) When:

The task is undefined or highly variable
- Example: "Automate a timesheet based on a week's activities"
- Example: "Determine and execute follow-ups from a business dinner"
- The agent needs to reason about what steps are even required depending on the situation
Context determines the path
- The "right" approach cannot be predetermined
- Each execution may require different tools or sequences
- Discovery of information changes what actions are needed
Conversational interaction is primary
- Users interact through natural language
- The agent needs to ask clarifying questions
- Response generation is the main output
Speed of implementation over governance
- Rapid prototyping or experimentation
- Internal productivity tools with lower risk
- Controlled user populations who can validate outputs

Choose Deterministic Workflows with AI Steps When:

Rules and compliance drive the process
- Regulatory requirements mandate specific steps
- Approvals must follow defined hierarchies
- Audit trails must be explainable and reproducible
- Example: Provisioning a laptop to a new employee
- Example: Processing an expense report approval
The process is well-defined with targeted AI needs
- 80% of steps follow known business rules
- Specific steps benefit from AI (summarization, classification, drafting)
- AI enhances the process; it does not define it
Existing flows should be enhanced, not replaced
- Teams have invested in proven workflow automation
- Rebuilding as agentic workflows would duplicate effort
- AI capabilities should augment current investments
End-to-end visibility is critical
- Teams need to understand the complete automation
- Troubleshooting requires clear execution paths
- Separate agents "listening" for events creates complexity
Human touchpoints are fixed by policy
- The workflow must define exactly when humans participate
- Certain decisions cannot be delegated to AI reasoning
- The organization requires "human in the loop" at specific stages

Real-World Scenario: Financial Services Operations

This example illustrates why many enterprise use cases benefit from deterministic workflows with AI steps.

The Challenge

A financial services team has a flow that processes ACH disputes. They want to add AI capabilities to accelerate resolution. They have two options:

Option A: AI-First (Agent as Orchestrator) Build an AI Agent that handles ACH disputes end-to-end, using flows as tools.

Option B: Deterministic with AI (Workflow as Orchestrator) Enhance the existing flow to call an AI Agent at specific steps.

Why Option B Was Chosen

The team chose to embed the AI Agent within their existing flow for these reasons:

Factor	AI-First Approach	Deterministic with AI
Compliance	Agent decides dispute handling sequence	Flow enforces required regulatory steps
Visibility	Agent and flow are separate entities	Complete end-to-end process visible in Workflow Studio
Maintenance	Two separate automations to maintain	Single flow with AI step to maintain
Predictability	Agent may handle similar disputes differently	Consistent handling with AI-enhanced intelligence
Existing Investment	Would require rebuilding current automation	Enhances proven flow with new capability

The Implementation

With the Use an AI Agent action in Workflow Studio, the team:

Kept their existing dispute intake and validation steps (deterministic)
Added an AI Agent step where reasoning about dispute resolution is needed
Continued with deterministic steps for approvals and notifications
Maintained full audit trail with clear AI contribution visibility

Hybrid Patterns in Practice

Many implementations combine both patterns based on the specific requirements of different process phases.

Example: Investigative Case Management

In investigative case management, different phases have different requirements:

Phase	Recommended Pattern	Rationale
Case Intake	Deterministic with AI	Validation rules are fixed; AI helps with document extraction and case narrative drafting
Lead Triage	Deterministic with AI	Escalation criteria are policy-driven; AI assists with entity matching and data enrichment
Active Investigation	Hybrid consideration	Dynamic nature may benefit from AI-first for certain investigative tasks
Reporting	Deterministic with AI	Report structure is defined; AI generates content within the template
Post-Investigation	Deterministic	Archival, retention, and legal integration follow strict rules

Key Insight: It Is Not Either/Or

The most effective implementations often use deterministic workflows as the primary orchestrator with AI agents called for specific capabilities:

Summarization: AI Agent summarizes case notes, emails, or documents
Classification: AI Agent determines category, priority, or routing
Content generation: AI Agent drafts responses, reports, or recommendations
Reasoning: AI Agent analyzes data to suggest next steps (human reviews)

Playbooks: AI Agent and AI Skill Activities

Pattern 2 also extends to Playbooks, ServiceNow's guided, stage-based process automation. Just as a flow can call the Use an AI Agent action or Execute Skill action, a playbook activity can call an AI agent or AI skill directly within a stage, so the playbook — not the AI — remains the orchestrator.

Use an AI agent activity: Adds an existing AI agent to a playbook stage. The author selects the agent, writes an objective, defines expected outputs, and sets who the agent runs as (the active user, the user who triggered the playbook, or the user who completed a prior activity). The agent gathers context and performs the work inside the playbook, without opening the separate Now Assist panel, and its outputs are available to the next activity.

Use an AI skill activity: Adds a lightweight, single-purpose AI task, such as summarization or classification, to a stage. Skill inputs load dynamically based on the skill selected. Outputs are referenced by later activities through the data pill picker, and multiple skill activities can be chained within the same playbook, each producing an independent output that downstream steps or branch conditions can use.

Playbook Activity	Description
Use an AI agent	Adds an existing AI agent to a playbook stage to gather context, perform work, and produce outputs for later activities
Use an AI skill	Adds a published AI skill to a playbook stage for a lightweight, focused task; its output feeds later activities or branch conditions

Agentic Playbooks use AI agents to assist the process worker in completing playbook activities, such as questionnaires and form activities. The agents perform research or AI-search tasks, can read knowledge base articles, and attempt to slot-fill any fields displayed in the activity. They can run collaboratively, surfacing results for the process worker to review and approve before the activity completes, or autonomously, completing the activity without human intervention. In either mode, Agentic Playbooks operate within the playbook's existing stages, business rules, and approvals: AI accelerates specific activities while the playbook governs the sequence, and required human touchpoints stay intact, extending the deterministic-with-AI pattern to guided, human-facing processes.

Implementation Guidance

Starting with Deterministic Workflows and Adding AI

If you have existing flows and want to add AI capabilities:

Identify steps where AI adds value
- Which steps require interpretation, summarization, or reasoning?
- Where do fulfillers currently spend time on manual cognitive tasks?
Choose the right AI capability
- For targeted tasks (summarization, classification): Use a Skill
- For complex reasoning with multiple tool calls: Use an AI Agent
Configure the AI step appropriately
- Define expected outputs if subsequent steps need the data
- Set execution mode (autonomous vs. supervised) based on risk
- Specify support user for supervised agents
Maintain governance
- AI operates within the workflow's governance structure
- Audit trails capture both deterministic steps and AI contributions
- Maintain human touchpoints at policy-defined locations

Transitioning from AI-First to Hybrid

If you have AI agents that would benefit from more structure:

Identify predictable sequences
- Which steps does the agent perform consistently and in the same order?
- Where are rules being "re-derived" that could be enforced?
Extract deterministic logic
- Build flows for the predictable portions
- Keep the agent for reasoning-intensive portions
Use the agent as a step, not the orchestrator
- Workflow controls the sequence
- Agent provides intelligence at specific points

Governance and Trust Considerations

When deciding between patterns, consider the governance implications:

Auditability

Pattern	Audit Trail
AI-First	Agent conversation logs capture reasoning, but path varies
Deterministic with AI	Workflow execution history shows exact steps; AI contributions are logged within the workflow context

Explainability

For regulated environments, consider who needs to understand decisions:

Auditors and regulators: May require step-by-step explainability that deterministic workflows provide
Technical teams: Can interpret agent reasoning logs
Business users: Benefit from predictable workflow structures

The Trust Threshold

As AI capabilities mature, the threshold for when to use AI-first patterns may shift. Currently, most enterprise use cases benefit from deterministic orchestration with targeted AI enhancement. Over time, as AI explainability and reliability improve, more processes may be candidates for AI-first approaches.

Conclusion

The introduction of hybrid workflow automation creates new possibilities for combining the best of deterministic and AI-driven approaches. The key takeaways are:

Choose based on requirements, not technology preference: Governance, compliance, and predictability needs should drive pattern selection
Deterministic workflows provide the "guardrails": AI operates inside the workflow, ensuring governance and auditability
Existing investments can be enhanced: You do not need to rebuild proven flows; you can add AI capabilities where they provide value

Resources

Workflow Automation Center of Excellence

Getting Started with Playbooks
Jumpstart your Workflow Automation: Playbooks (Impact Offering)
Platform Fundamentals Academy: Use Workflows as Tools for AI Agents

Getting Started with Agentic Playbooks

Identifying the right use case(s) for Agentic workflows

Design considerations - traditional vs agentic workflows

Training

Workflow Automation Training on ServiceNow University
Now Assist AI Agents Deep Dive on ServiceNow University
AI Builder Fundamentals on ServiceNow University

Academy Sessions

AI Academy: Hybrid AI Automation (Recording)

Center of Excellence Navigation

Workflow Automation - Center of Excellence
Workflow Automation Migration Considerations
What's new for Workflow Automation?
Resource Hubs
FAQ
Checklist
Training
Workflow Academy