MODULE 17 — AI STRATEGY & DEPLOYMENT

7

17.1 What is AI Strategy?

AI Strategy is a long-term plan for using AI effectively within an organisation or project to achieve measurable outcomes.

Why Have an AI Strategy?

AI strategy can help organisations:

• Improve productivity
• Reduce costs
• Increase innovation
• Gain competitive advantage

Without a proper strategy, organisations may waste resources, increase security risks and lose trust in AI systems.

17.2 Components of an AI Strategy

A. Business Goals

First determine why AI is needed.

Examples:

• Customer-support automation
• Faster data analysis
• Personalised recommendations

B. Data Strategy

AI depends heavily on data quality.

Organisations need to:

• Collect relevant data
• Store it securely
• Maintain data quality

C. Technology Selection

Choose suitable:

• AI models
• Cloud platforms
• Databases
• Tools

D. Team & Skills

AI projects may require:

• AI engineers
• Data scientists
• Developers
• Domain experts

Continuous upskilling is important.

E. Governance & Ethics

AI must follow relevant ethical, privacy and security requirements.

17.3 AI Deployment Lifecycle

6

Stage 1 — Problem Identification

Understand the exact problem AI will solve.

Stage 2 — Data Collection

Gather relevant, high-quality and diverse data.

Stage 3 — Model Development

Train, tune and optimise the AI model.

Stage 4 — Testing & Validation

Evaluate:

• Accuracy
• Fairness
• Performance
• Safety

Stage 5 — Deployment

Release the AI system for real-world users.

Stage 6 — Monitoring & Maintenance

Continuously track performance and improve the system.

17.4 Types of AI Deployment

7

Cloud Deployment

AI systems are hosted on cloud platforms.

Examples include:

• AWS
• Google Cloud Platform
• Microsoft Azure

Advantages

• Scalable
• Lower upfront cost
• Global accessibility

Challenges

• Internet dependency
• Privacy concerns
• Ongoing cloud costs

On-Premise Deployment

AI runs on an organisation's own servers.

Examples:

• Private data centres
• Company servers
• Local GPU clusters

Advantages

• Greater data control
• No internet requirement
• Lower latency

Challenges

• High setup cost
• IT maintenance
• Limited scalability

Edge Deployment

AI runs directly on devices near the data source.

Examples:

• Smartphones
• IoT sensors
• Autonomous vehicles

Advantages

• Very low latency
• Offline operation
• Privacy benefits

Challenges

• Limited computing power
• Model-size constraints
• Update complexity

17.5 Challenges in AI Deployment

Major challenges include:

High Costs

AI infrastructure can require significant investment.

Data Privacy

Sensitive information requires appropriate protection and compliance.

Model Inaccuracies

Real-world data may differ from training data.

Infrastructure Requirements

AI needs suitable hardware, networking and monitoring.

User Adoption

Users may distrust or misuse AI, making training and change management important.

17.6 AI Scaling & Monitoring

6

AI Scaling

Scaling means expanding an AI system to handle:

• More users
• Larger datasets
• More tasks
• More features

AI Monitoring

After deployment, organisations should continuously monitor:

Error Detection
Identify incorrect outputs.

Bias Surveillance
Detect emerging unfair patterns.

Security Monitoring
Identify threats.

Performance Tracking
Ensure accuracy and speed remain acceptable.

17.7 Future of AI Deployment

Future developments include:

• Autonomous AI systems
• Real-time AI monitoring
• Expansion of Edge AI
• Personalised AI services

The overall direction is toward AI systems that are faster, smarter and closer to users.

Module 17 Activity

AI Deployment Planning Exercise

Choose an AI project and define:

1. Business problem
2. Data requirements
3. AI technology
4. Deployment type
5. Testing plan
6. Monitoring requirements
7. Security and privacy requirements