AI Strategy: Where to Start
Artificial intelligence has moved from research novelty to practical business tool with remarkable speed. The challenge for most organisations is not whether to invest in AI but how to identify the right opportunities, avoid wasted investment on AI theatre, and build genuine capability. A practical AI strategy starts with problem identification, not technology selection.
Finding High-Value AI Use Cases
The best AI use cases share common characteristics: repetitive tasks with clear inputs and outputs; high volume (AI's value compounds at scale); tolerance for imperfect outputs (AI is probabilistic, not deterministic); access to relevant training data; measurable business outcomes. Examples: customer support deflection, document classification, predictive maintenance, personalisation, fraud detection.
The AI Maturity Ladder
- Exploratory: Using AI-powered SaaS tools (GitHub Copilot, ChatGPT, Notion AI) — no infrastructure investment, immediate value
- Implemented: Embedding AI API calls (OpenAI, Anthropic, Gemini) into products and workflows — moderate investment, customisable
- Fine-tuned: Fine-tuning foundation models on proprietary data for specific use cases — significant investment, differentiated capability
- Purpose-built: Training custom models on proprietary datasets — very high investment, maximum differentiation
AI Governance
AI use without governance creates risk: hallucinations presented as facts, biased outputs, GDPR violations in training data, IP exposure from data sent to third-party models. Establish an AI policy: approved tools, data classification rules for AI use, output review requirements for high-stakes use cases, and accountability for AI-generated outputs.