OnlineOnline version
Animated before → after diagrams, case videos and links to primary sources.
× Sergey SelivanovIndustry report · 2026
47 cases where AI in consumer goods has already paid off
What the largest food, beverage, beauty and household goods manufacturers measured in money, weeks and percentages
✓Link sent
Check your inbox
We sent the link to the report to . It should arrive within a couple of minutes.
No email? Check Spam and Promotions.
Three formats
Read it, download it or ask your AI
Leave your email and we'll send you a link to all three versions.Leave your email and all three versions open right away. Pick the one that suits you, or use them all.
Online
PDFPDF to print
To read offline, print for a strategy session or forward to your CEO.
Version for AI
The case base for Claude, ChatGPT or Gemini. Ask about your category: the answer comes with case numbers and sources.
Free. The link arrives by emailFree, right after your email
What's inside
Eight functions: from R&D to the store shelf
Each function gets its own case pages: how it was, what they did, which metric moved and where it was published.
47cases with a measured result
72more cases in the appendix
40+global manufacturers
8business functions, from R&D to sales
69pages with sources and reliability ratings
01
Content and creative
Unilever · L'Oréal · Kraft Heinz · Coty · Moët Hennessy · Coca-Cola−40…−90%production costs
7 cases02
Research and insights
Nestlé USA · Reckitt · Danone · Mars Wrigley · Barilla · Heinekenweeks → hoursresearch timelines
6 cases03
Media and creative effectiveness
Kellanova · Mars · Nestlé · P&G · Reckitt · Pernod Ricard · Mondelēz India · PepsiCo+11…+40%media ROI
8 cases04
E-commerce, CRM and AI advisors
Unilever · Nestlé Purina · L'Oréal · NIVEA · Clorox · Nestlé+9…+43%conversion and purchases
6 cases05
Planning and supply chain
AB InBev · General Mills · Kao · Kimberly-Clark · Suntoryup to −38%write-offs
5 cases06
Manufacturing and quality
Unilever · Hindustan Unilever · PepsiCo · Colgate-Palmolive · Diageo−85%line changeover time
5 cases07
Sales and retail execution
Heineken · Pernod Ricard · Coca-Cola · Henkel · Nestlé+1.5…+8%outlet sales
5 cases08
R&D, service and back office
Mondelēz · L'Oréal · Unicharm · P&Gup to −97%time per task
5 cases+ WEF Lighthouse factories: what full digitalization delivers+ appendix: 72 more cases in one table
Closest to proof
Numbers measured against a control
More than half of the numbers in public cases come from vendors or executives, and almost nobody publishes a control group. These four were measured against one or by a third party.
Nestlé Purina
+9.44%Conversion vs. a control group after product page photos were reordered by an AI score. No new content was shot
p. 33P&G
~12%Faster work in teams with gen AI. A controlled experiment with 776 employees
p. 62Coca-Cola HBC
+40%Sales per outlet vs. control in the pilot of outlet prioritization and promo evaluation, +10% after it
appendixNIVEA
+8.4 ppBrand recall after an AI skin advisor in WhatsApp, measured by Meta Brand Lift
p. 35Every other number carries a reliability mark, so you can see how far to rely on it.
Look inside
Every case on its own page
A before → after diagram, the key numbers, a linked source and three reliability squares.
60 more pages
Scroll sideways →
How we selected
Only published results
Forecasts, targets and “expected impact” did not make it in. What we checked and rejected is described in the appendix.
~900source pages: press releases, annual reports, trade press
120deployments found at manufacturers
47most telling ones in the main section
72the rest in the appendix
A case made it into the report only if the result was published and we saw the number in the source ourselves. A number in the report is the company's estimate, not an audited result. The rare cases with a control group are marked.
Source reliability
Three squares on each case page show how far you can rely on the number.
HighA primary source or independent press with a named quote and specific numbers
MediumNumbers are given, but the source is a vendor, there is no baseline, or AI's contribution is mixed with other changes
LowA single metric, no date, an anonymous quote, or questionable attribution to AI
The report is free
Get all three versions of the report
The online version, a 69-page PDF and a case base for AI assistants. We'll email you the link.They open right after your email.
