Updated
Updated · ZDNet · Jun 18
PwC, NBCUniversal Outline 4 AI Agent Rules as Firms Balance Speed With Risk
Updated
Updated · ZDNet · Jun 18

PwC, NBCUniversal Outline 4 AI Agent Rules as Firms Balance Speed With Risk

2 articles · Updated · ZDNet · Jun 18

Summary

  • Four deployment rules emerged from PwC and NBCUniversal: keep humans in control, run fast experiments, fix data and workflows first, and match guardrails to each use case’s risk.
  • PwC said rapid one- and five-day AI test cycles help teams learn quickly, but warned many middle managers still struggle with the shorter development rhythm and broader process change.
  • NBCUniversal said AI can "blow up" broken workflows, making clean data, clear process ownership and repeatable tasks the best starting points for early agent rollouts.
  • Risk tolerance should shape oversight, the executives said: low-stakes internal tools may need little review, while customer-facing agents require tighter governance, intake checks and restricted data access.
  • PwC added that scaling safely depends on architecture built before widespread AI adoption, with about 1% of staff setting standards and 10% acting as distributed builders across the business.

Insights

As AI agents begin taking actions, is human oversight enough to prevent irreversible business mistakes?
With 'Shadow AI' surging, how can companies govern powerful tools they don't even know exist?
Is 'blowing up' entire processes for AI a necessary revolution or a needlessly disruptive risk?