AI & Automation · AI Strategy · Medium · 25-35 min

GenAI to Deflect 30% of Call-Center Volume

GenAI to Deflect 30% of Call-Center Volume is a medium AI & Automation ai strategy case interview that runs 25-35 min. A company's contact center handles 8M calls a year at roughly $6 per call. A strong answer works through 5 phases: Segment call types; Build the savings case; Design the safety guardrails; Plan the rollout; State a realistic deflection range and what determines where it lands.

Last updated 2026-09-05

The brief

A company's contact center handles 8M calls a year at roughly $6 per call. The COO wants generative AI to deflect 30% of volume within 12 months. The CX team worries about hallucinated answers and brand damage; the CFO has approved a pilot budget but wants the business case. You are asked whether 30% is achievable and how to roll it out safely.

How to approach it

  1. Segment call types — which are simple/repetitive (good for AI) vs complex/emotional (keep human)
  2. Build the savings case — deflectable volume × cost per call, net of platform and integration cost
  3. Design the safety guardrails — retrieval over invented answers, confidence thresholds, human handoff, monitoring
  4. Plan the rollout — pilot on a narrow intent set, measure containment and CSAT, then expand
  5. State a realistic deflection range and what determines where it lands

What a strong answer does

  • Recognizes 30% is an outcome of which intents are automatable, not a target you simply set
  • Defaults to retrieval-grounded answers + confidence-based human handoff to control hallucination
  • Measures success on containment AND customer satisfaction, not deflection alone
  • Pilots narrow, instruments it, then scales — rather than switching 30% on day one
  • Nets platform, integration, and ongoing tuning cost against the gross savings

Red flags interviewers score down

  • Promises 30% deflection without analyzing call-type mix
  • Has no plan for hallucinations or escalation to a human
  • Optimizes deflection while ignoring CSAT and brand risk
  • Counts gross savings with no implementation or run cost

Cases are written in each firm's style, written and reviewed by working consultants; they are not the firms' own published cases.

More AI & Automation cases