Fix the Slide: Teardown of a Zomato AI ROI Report
Objective: Review a realistic internal AI ROI slide before it goes to a CFO review, and identify the defects the lesson warns about: time-saved-only math, no baseline, no named output metric, and no miss-rate disclosure.
You're the marketing analyst at Zomato reviewing a colleague's draft AI ROI slide before it goes into next week's finance review.
Find what's wrong with the slide's math and its framing before finance does.
Before you start
What you'll need
Free path (everything below is enough to finish)
Enough to rebuild a two-column before/after metric table
The process
Specimens to review
This slide is about to go into a CFO review. Identify every defect that would make a sharp finance reviewer push back, and note what's actually fine about the slide as written.
SLIDE: 'AI Content Tool, Q2 ROI' Headline number: We saved 18 hours/week across the content team. At a loaded rate of ₹1,400/hour, that's ₹1,00,800/week saved, or ₹52.4L annualized. Tool cost: ₹18L/year. Net ROI: positive. Supporting slide notes: 'Team reports the tool feels faster.' No other metrics attached. No mention of pieces published, quality, or conversion impact. Every use case this quarter is reported as a win.
Specimen: synthetic, realistic
Final deliverable
An annotated copy of the slide flagging each defect by severity, plus a one-line fix proposing the output metric to report instead.
See a reference example
Swiggy, corrected ROI note (excerpt) Instead of 'saved 14 hours/week': content velocity went from 3.1 to 5.4 pieces/sprint (baseline: 6 weeks pre-rollout), held against the same editorial quality gate. One use case, personalization test copy, showed no measurable lift and is flagged as a miss.
Success criteria
You're done when you can:
- Identifies the time-saved-only headline as the critical defect, not a supporting detail
- Correctly separates real defects from the two distractors