Body
Purpose
The AI Practice is an Enterprise Architecture practice that helps the University of Arkansas adopt, operate, and evolve artificial intelligence technologies through shared knowledge, documented patterns, and reusable architecture guidance.
The AI Practice exists to help the university learn from AI implementations and apply those lessons broadly rather than repeatedly solving the same problems.
Mission
Provide practical architecture guidance that helps university units use AI capabilities in a secure, supportable, cost-conscious, and sustainable manner.
The practice supports AI adoption by documenting standards, patterns, decisions, lessons learned, and reference architectures.
Why the AI Practice Exists
Artificial intelligence is becoming easier to access, configure, and incorporate into university work.
Faculty, staff, researchers, students, vendors, and university units can rapidly acquire, create, or enable AI capabilities. These capabilities may process university data, connect to enterprise systems, take actions, generate content, or become part of ongoing university operations.
Existing requirements for university data continue to apply when AI technologies are used. Data obligations follow the data regardless of the platform, model, application, or service used to process it.
The challenge is not whether AI technologies will be used.
The challenge is ensuring that the university can:
- Reuse successful approaches
- Avoid unnecessary duplication
- Understand ownership and operational responsibilities
- Manage costs and consumption
- Protect university data
- Connect AI solutions to existing systems and services
- Support AI capabilities over time
What the Practice Does
The AI Practice serves three functions.
Curate
Maintain reusable AI knowledge, including:
- Standards
- Patterns
- Reference Architectures
- Architectural Decision Records (ADRs)
- Lessons Learned
- Roadmaps
Connect
Connect AI consumers with:
- Existing services and platforms
- Subject matter experts
- Architecture guidance
- Data owners and stewards
- Institutional requirements
Accelerate
Help teams move from idea to implementation by providing:
- Proven approaches
- Reusable patterns
- Architecture guidance
- Platform recommendations
- Early identification of dependencies and requirements
The goal is to reduce friction, improve consistency, and help useful AI ideas move forward with fewer surprises.
Scope
The AI Practice focuses on architecture and guidance related to:
- AI platform architecture
- Generative AI applications
- Copilots and conversational assistants
- AI agents and agent actions
- Model access and hosting approaches
- Retrieval and grounding architectures
- Knowledge source integration
- Identity and access integration
- Application and API integration
- Data classification and handling
- Monitoring and observability
- Cost and consumption management
- Ownership and lifecycle models
- AI service patterns
Out of Scope
The AI Practice does not:
- Approve every use of AI
- Operate all AI services
- Build or support every AI solution
- Own institutional AI strategy
- Own business priorities
- Own academic integrity decisions
- Determine research goals or methods
- Own university data
- Replace procurement processes
- Replace security, privacy, compliance, or policy authorities
The practice provides architecture guidance and stewardship.
Business decisions, data decisions, policy decisions, and operational ownership remain with the appropriate stakeholders.
Deliverables
Standards
Examples:
- AI platform standards
- AI identity standards
- AI data handling standards
- AI logging and monitoring standards
- AI ownership and lifecycle standards
Patterns
Examples:
- Public knowledge assistant pattern
- Authenticated knowledge assistant pattern
- Retrieval-augmented generation pattern
- Agent action pattern
- Human approval pattern
- AI-enabled application pattern
Reference Architectures
Examples:
- AI knowledge assistant architecture
- Enterprise AI application architecture
- Copilot architecture
- Agent architecture
- AI platform architecture
Architecture Decision Records
Examples:
- AI platform decisions
- Model access decisions
- Agent identity approaches
- Knowledge grounding approaches
- Integration approaches
- Cost allocation approaches
Roadmaps
Examples:
- AI platform evolution
- AI service maturity
- Shared AI capabilities
- Strategic AI architecture needs
Relationship to Services
The AI Practice and AI services are related but distinct.
The AI Practice stewards architecture knowledge.
Services provide operational capabilities.
| AI Practice |
AI Service |
| AI platform standards |
AI Platform Service |
| Copilot patterns |
Copilot Service |
| Agent architecture guidance |
Agent Platform Service |
| AI adoption guidance |
AI Enablement Service |
| Knowledge integration patterns |
Enterprise Search or Knowledge Services |
| Identity guidance |
Identity Services |
| Integration guidance |
Integration Services |
| Cost allocation guidance |
Cost Reporting Services |
A practice may inform one or more services.
A service may be influenced by multiple practices.
Practice stewardship does not imply operational ownership of an AI service.
Operational ownership of an AI service does not automatically imply stewardship of the AI Practice.
The same individual may participate in both.
Relationship to Other Practices
The AI Practice works across other Enterprise Architecture practices.
| Practice |
Relationship to AI |
| Cloud Practice |
Provides hosting, platform, cost, networking, and operational patterns used by AI capabilities. |
| Identity Practice |
Provides authentication, authorization, ownership, and identity lifecycle patterns for users, applications, and agents. |
| Integration Practice |
Provides API, connector, event, and system interaction patterns used by AI applications and agents. |
The AI Practice does not replace these practices. It brings their guidance together for AI-enabled solutions.
Stewardship
The AI Practice is maintained through the Enterprise Architecture stewardship model.
The practice may include:
- A Practice Steward
- Practice Contributors
Contributors may be drawn from teams responsible for cloud, applications, data, identity, integration, research computing, security, academic technology, and other relevant capabilities.
The practice may be active, emerging, or unassigned depending on organizational need.
Participation does not require a dedicated team or full-time role.
Success Measures
The AI Practice is successful when:
- AI decisions become easier to make.
- Teams can find clear guidance for common AI use cases.
- Reusable patterns reduce repeated effort.
- AI solutions have clear ownership and support expectations.
- Teams engage the appropriate stakeholders earlier.
- Lessons learned are documented and reused.
- Duplicate platforms and solutions are reduced.
- AI implementations are easier to integrate and support.
- Useful AI ideas move from experimentation to sustainable operation more effectively.
- AI adoption becomes more consistent across the university.