DataFramed #352 AI Agents at Work: What Actually Breaks (and How to Fix It) with Danielle Crop, EVP Digital Strategy & Alliances at WNS
55 snips
Mar 23, 2026 Danielle Crop, EVP of Digital Strategy & Alliances at WNS and former Chief Data Officer at American Express and Albertsons, guides a pragmatic take on AI agents. She discusses ring-fencing experiments, avoiding hallucinations, balancing curiosity with critical checks. Conversations cover data access choices, token costs, tone controls, governance for unstructured data, and tying AI plans to P&L.
AI Snips
Chapters
Transcript
Episode notes
Start RingFenced Pilots To Tune Data Access
- Try feeding data into agents in a ring-fenced proof of concept to learn which inputs are valuable and which create risk.
- Crop recommends dialing data access up or down based on measured organizational risk appetite.
Control Token Spend Early
- Monitor token and compute costs and consider restricting per-user usage once adoption grows.
- Danielle Crop warns token spend can become significant and organizations should plan access and cost constraints.
Strategy Needs Business And Data Fluency
- Insight: Effective AI strategy blends business savvy with data/tech expertise to choose partners and simplify stacks.
- Crop argues leaders must connect dots across cloud, SaaS, LLMs and client needs to avoid complex ecosystems.