Use patterns in the current-state workflow to generate candidate opportunities.
High-signal patterns
Repetitive manual transfer
People copy information between email, spreadsheets, CRM or ERP systems.
Consider: integration, structured intake, rules-based automation.
High-volume drafting
Staff repeatedly produce similar emails, reports, proposals or summaries from known information.
Consider: AI-assisted drafting with review.
Knowledge search
People spend time finding answers across documents, policies or historical cases.
Consider: improved knowledge management or retrieval-augmented assistance.
Triage and classification
Teams manually sort requests, tickets, documents or transactions.
Consider: deterministic routing, ML/LLM classification or hybrid rules.
Rework and corrections
The same work returns because required information is missing or quality checks happen too late.
Consider: redesign the intake and validation process before automating downstream work.
Bottlenecked decisions
Work waits for a small number of people to interpret information or approve routine cases.
Consider: decision support, clearer rules, delegated authority or automated low-risk pathways.
Ask five questions for every candidate
- What business problem does this solve?
- What evidence shows the problem exists?
- What intervention is simplest?
- What data, systems and human controls are required?
- How would the client know it worked?
Do not confuse “AI can do this” with “the client should invest in doing this with AI.” Technical feasibility is only one part of the decision.
