01Find the high-value repetitive work
Workflows were reviewed for activities that were repetitive, document-heavy, error-prone, standardisable, and common across service lines. The first use cases: financial statement pre-reviews, trust deed summaries, tax return checks, document conversion, audit questions, timesheet preparation and report consolidation.
02Build reusable tools, not one-off prompts
Early prompts became durable AI skills, automations, templates and add-ins: automated financial statement and audit pre-reviews, audit mapping and lead-sheet indexing, compliance workpapers, tax and regulatory research tools, an R&D application workflow, a due diligence workflow run through Excel, and department-specific add-ins. When an early workpaper architecture proved too heavy, it was redesigned as a lighter skill: the process knowledge kept, the infrastructure cut.
03Embed AI where people already work
No new systems to learn. AI capability arrived inside Excel, Word, PowerPoint, Outlook, Teams and the firm's existing accounting and workpaper platforms. That removed the adoption friction that kills most rollouts.
04Build the capability, not just the tools
A structured enablement program ran alongside the build: AI champion meetings, firm-wide education, role-specific training for directors, starters and service teams, skill-share sessions, prompt and skill libraries, and training on hallucination reduction, source checking and responsible use. The champions became a network that drove adoption inside their own teams and surfaced the next opportunities.