Managed AI
Managed AI solution types

GPT, agent, or automation?

Compare how each solution starts, works, connects, acts, and involves people.
Compare byDecision point
InteractiveGPT / ChatbotPrompt → response
AdaptiveAgentGoal → judgment → next step
RepeatableAutomationTrigger → rules → action
01Primary role
Assists a person
Works toward a defined goal
Executes a defined process
02How it starts
A person submits a prompt, question, or file
A person assigns a goal, or an approved event creates a task
A schedule, status change, submission, message, or other trigger
03How it works
Responds using instructions and approved knowledge
Interprets information and selects the next approved step
Follows predetermined steps, rules, and branches
04Decision-making
Creates a response for the user to evaluate
Uses judgment within defined boundaries
Applies rules established before the process runs
05Connections
Optional and often knowledge-focused
Uses the specific sources and tools required for the goal
Uses defined source and destination systems
06Actions
Usually produces information, analysis, or content
May research, coordinate, update, prepare, or complete approved actions
Completes the actions built into the process
07Human role
Directs the interaction and reviews each result
Sets the goal and reviews approvals, exceptions, and sensitive actions
Defines the process and handles approvals, failures, and exceptions
08Best fit
Interactive knowledge, drafting, and guided task support
Variable, multi-step work requiring analysis or judgment
High-volume, repeatable work with stable rules
Detailed comparison

What each solution type includes.

On-demand assistance

GPT / Chatbot

A guided assistant used on demand to answer questions, create content, or complete a defined task from approved knowledge.
User provides
A question, instruction, document, or other context
Solution provides
An answer, draft, summary, analysis, recommendation, or structured document
Common boundaries
Approved knowledge, defined instructions, user access, and source requirements
Examples
Policy Q&AProposal draftingCompliance guidanceIncident response support
Browse GPTs in the Solution Directory
Reasoning and coordination

Agent

A connected solution that interprets information, applies rules, and coordinates multi-step work across approved systems.
User provides
A goal, case, record, assignment, or defined outcome
Solution provides
A completed analysis, coordinated work product, prepared action, updated record, or escalated exception
Common boundaries
Permitted tools, accessible data, allowed actions, approval gates, and escalation rules
Examples
Account researchRFQ intakeCampaign analysisMatter research
Browse Agents in the Solution Directory
Repeatable execution

Automation

A repeatable process that starts from a schedule or trigger and completes approved actions without manual step-by-step execution.
Process provides
A trigger, sequence, business rules, source systems, destinations, and approval points
Solution provides
A routed request, notification, generated document, synchronized record, or completed process step
Common boundaries
Predetermined steps, action permissions, failure handling, logging, and human approvals
Examples
Meeting follow-upEmail triageEmployee onboardingApproval routing
Browse Automations in the Solution Directory
Choosing a solution type

Start with how the work operates.

Choose a GPT / Chatbot

When a person should initiate and guide the work.

  • The work begins with a question or instruction
  • The primary result is information or content
  • The user should review each response
  • Independent system action is not required
Choose an Agent

When the path changes based on what the solution finds.

  • The work has a goal, not one fixed sequence
  • Several sources or tools may be required
  • Information determines the next step
  • Exceptions and approvals can be defined
Choose an Automation

When a defined process should run consistently.

  • The same event starts the work
  • The steps and rules are predictable
  • The action should run without manual initiation
  • Failures and approvals have named owners
Combined solutions

One process may use all three.

Each part retains separate access, authority, and approval requirements.
01

Automation starts

A defined event collects information and creates the work item.
02

Agent evaluates

The agent checks approved sources and determines the next step.
03

Human approves

The reviewer approves, rejects, revises, or escalates the proposed action.
04

GPT supports

Employees review the result, ask questions, or prepare related content.
Managed AI configuration

What ITS defines and manages.

Business purpose

Approved need, intended result, users, owners, and success requirements

Knowledge

Policies, procedures, templates, records, and other approved sources

Connections

Specific systems, data, tools, and actions required for the solution

Instructions and rules

Required process, output format, decision boundaries, and prohibited behavior

Human oversight

Approval points, assigned reviewers, exception owners, and escalation requirements

Ongoing management

Testing, access reviews, source updates, configuration changes, and monitoring
Every connected solution needs boundaries.
  • Only approved users and data
  • Only required system access
  • Only defined actions
  • Human review for sensitive activity
  • Documented failures and escalation
  • Ongoing access and performance review

Explore the Solution Directory.

Browse Solution Directory