Choosing an AI stack
AI on IBM i isn’t about choosing a single technology stack — it’s about matching the right approach to your business problem. This guide helps you identify which AI journey aligns with your organization’s needs and directs you to the solutions that make it practical.
AI Use Cases for IBM i
Section titled “AI Use Cases for IBM i”Different business challenges require different AI approaches. The following journeys represent the most common ways IBM i organizations are applying AI today.
Agentic AI
Section titled “Agentic AI”What it is: AI systems that can autonomously plan, reason, and take actions across your IBM i environment — querying data, monitoring systems, and orchestrating workflows without constant human intervention.
Why it matters for your business:
- Automate complex, multi-step processes that currently require manual coordination
- Enable business users to access IBM i data through natural language queries
- Reduce operational overhead by having agents monitor system health and flag issues proactively
- Accelerate decision-making by giving agents direct access to decades of transactional data
Common use cases: Business process automation, conversational data analytics, system operations monitoring, security auditing, performance management
Developer Experience
Section titled “Developer Experience”What it is: AI-powered coding assistants that help developers write, understand, and modernize IBM i code faster — from RPG and CL to SQL and Python.
Why it matters for your business:
- Accelerate application development and reduce time-to-market
- Lower the barrier for new developers learning IBM i languages
- Speed up modernization efforts by generating boilerplate and explaining legacy code
- Reduce technical debt by making code refactoring more accessible
Common use cases: Code generation, code explanation, modernization assistance, automated documentation
Machine Learning
Section titled “Machine Learning”What it is: Training algorithms on your historical IBM i data to make predictions and identify patterns — without being explicitly programmed for each scenario.
Why it matters for your business:
- Turn decades of transactional data into predictive insights
- Identify fraud and anomalies before they become costly problems
- Optimize inventory and reduce waste through demand forecasting
- Improve customer retention by predicting churn before it happens
- Reduce downtime through predictive maintenance
Common use cases: Fraud detection, demand forecasting, customer segmentation, churn prediction, predictive maintenance
Deep Learning
Section titled “Deep Learning”What it is: Neural networks that automatically learn features from unstructured data like images, documents, video, and time-series — without manual feature engineering.
Why it matters for your business:
- Process invoices, purchase orders, and forms automatically with OCR
- Analyze video feeds for quality control, security, or operational monitoring
- Detect subtle anomalies in high-dimensional operational data
- Forecast complex temporal patterns in production and sales data
- Extract insights from unstructured text in customer service records
Common use cases: Document processing, image/video recognition, time-series forecasting, anomaly detection, natural language processing
Generative AI
Section titled “Generative AI”What it is: AI systems that produce new content — text, code, summaries, answers — by reasoning over ambiguous inputs and synthesizing information from multiple sources.
Why it matters for your business:
- Give non-technical staff direct access to IBM i data through natural language queries
- Reduce time analysts spend writing reports by automating summarization
- Help developers work faster by generating boilerplate and explaining unfamiliar code
- Surface insights buried in unstructured data like emails and service notes
- Build conversational interfaces to business applications
Common use cases: Natural language queries over Db2 for i, code generation, document summarization, conversational chatbots
How to Choose
Section titled “How to Choose”Start by identifying the business problem you’re trying to solve, not the technology you want to use:
-
Do you need to automate complex workflows or give users conversational access to data? → Start with Agentic AI
-
Are you trying to speed up development or modernize legacy code? → Start with Developer Experience
-
Do you want to predict outcomes from structured transactional data? → Start with Machine Learning
-
Are you working with images, video, documents, or complex time-series data? → Start with Deep Learning
-
Do you need to generate text, summaries, or enable natural language interfaces? → Start with Generative AI
Each journey page includes:
- Detailed use case descriptions
- Platform and deployment options
- Integration patterns with IBM i
- Getting started guides and examples
Choosing Deployment Model
Section titled “Choosing Deployment Model”Once you’ve identified your AI use case, the next decision is where to run your AI workloads. There are typically two options: self-hosted (on-premise) solutions and cloud-based SaaS offerings based on data sovereignty, compliance requirements, and operational preferences.
When to Choose On-Premise (Self-Hosted)
Section titled “When to Choose On-Premise (Self-Hosted)”Best for organizations that:
- Must keep sensitive data on-premise due to regulatory requirements
- Want full control over infrastructure, security policies, and data governance
- Already have significant investment in Power hardware and want to maximize utilization
- Need to minimize network latency between AI workloads and IBM i data
- Prefer predictable costs over variable cloud consumption pricing
On-premise solutions include:
- IBM i Native AI — Run AI directly in IBM i PASE
- Red Hat OpenShift AI — Enterprise Kubernetes platform for AI/ML on-premise
- Red Hat AI Inference Server — Deploy and serve LLMs on your own infrastructure
- Wallaroo AI Platform — Production ML deployment optimized for Power
- Wallaroo AI Starter Kit — Managed ML deployment service
- Equitus Video Sentinel — Video/image AI running on Power hardware
- Equitus Knowledge Graph Neural Network — Graph-based AI on Power
- RocketGraph XGT — High-performance graph analytics on Power
- Python Ecosystem for IBM Power — Run AI workloads directly on Power systems with MMA acceleration
When to Choose SaaS/Cloud
Section titled “When to Choose SaaS/Cloud”Best for organizations that:
- Want to start quickly without infrastructure setup
- Prefer operational simplicity and managed services
- Need elastic scaling for variable workloads
- Want access to the latest models and features without manual updates
- Are comfortable with data leaving the on-premise environment (with appropriate security controls)
SaaS/Cloud solutions include:
- watsonx.ai — IBM’s managed AI studio for training and deploying models
- watsonx.data — Managed data lakehouse with AI integration
- IBM Bob — Cloud-based AI coding assistant
Deployment Considerations
Section titled “Deployment Considerations”Regardless of which AI journey and deployment model you choose, you’ll need to consider:
- Data access: How will your AI system connect to Db2 for i?
- Compute location: Will AI workloads run on IBM i, on adjacent Power systems, or in the cloud?
- Hardware acceleration: Can you leverage MMA or Spyre?
- Governance: What controls do you need around model deployment, data privacy, and compliance?
- Data sovereignty: Where can your data legally reside and be processed?
- Network connectivity: What bandwidth and latency requirements exist between IBM i and AI systems?