Project Case Study
TimeCapture Automation
A LegalTech automation stream focused on turning everyday work signals (calendar, email, documents) into accurate, reviewable time-entry suggestions—without adding admin burden.
Problem
Fee earners lose billable time because capture is manual and happens late. Signals exist in Microsoft 365, but they’re scattered and hard to translate into clean time entries. Teams also need governance: permission models, auditability, and predictable delivery for law-firm environments.
Approach
- Signal mapping: identified high-value sources (Outlook calendar, mail metadata, files) and defined “what counts” as a time candidate.
- Suggestion engine: generated draft entries with sensible defaults (matter context, durations, narratives) and kept a “human-in-control” review flow.
- Enterprise consent model: aligned Graph delegated permissions, admin consent, and tenant policies to fit law-firm expectations.
- Phased rollout: shipped incrementally (capture → suggestions → quality rules → reporting hooks) with adoption checkpoints.
- Quality + traceability: added audit trails, error handling, and safe fallbacks so automation never blocks users.
Key Decisions
- Delegated access first to match user context and reduce overreach risk.
- RAG-style retrieval where needed (policy/knowledge support), not “AI everywhere”.
- Keep suggestions explainable so users trust why a draft entry was created.
- Optimize for adoption: fast review, minimal clicks, consistent defaults.
What I’d Improve Next
- Better matter inference using firm-specific patterns (without exposing sensitive content).
- Rule tuning controls so admins can adjust thresholds and defaults safely.
- Stronger feedback loop: learn from accepted/rejected suggestions to reduce noise.