Silhouette of an executive looking at a screen full of green indicators

Oct 29, 2026

AI's First Job Should Not Be Writing Ads, but Finding Lost Margin

When AI is discussed, most companies first think of image generation, ad copy or a chatbot. These can be useful. But in cosmetics and retail companies, the bigger opportunity often sits in the invisible corners of operations.

AI

Data

Commercial Management

Insight

Good technology doesn't have to be visible. Sometimes the best transformation is the stock crisis that never happened or the wrong discount that was never given.

Which product is overstocked in which store? Which SKU has high sales but low net contribution? On which marketplace are returns eroding the real profitability of advertising spend? Which distributor sends a risk signal before placing a reorder? In which production order does a delivery delay keep recurring?

Analysis

Data-driven forecasting and early-warning systems can help find answers to these questions sooner. But before any AI implementation, product codes, cost definitions, channel-level order data and reporting discipline must be reliable. A fast system built on bad data only speeds up the wrong decision.

For me, the success measure of a technology project is not how many new tools were installed. It is how much cash was released from stock, which class of errors decreased, how much decision time shortened and how net commercial contribution changed.

An old calculator on financial printouts

High shelves in a large distribution warehouse

AI's First Job Should Not Be Writing Ads, but Finding Lost Margin

When AI is discussed, most companies first think of image generation, ad copy or a chatbot. These can be useful. But in cosmetics and retail companies, the bigger opportunity often sits in the invisible corners of operations.

AI

Data

Commercial Management

Insight

Good technology doesn't have to be visible. Sometimes the best transformation is the stock crisis that never happened or the wrong discount that was never given.

Which product is overstocked in which store? Which SKU has high sales but low net contribution? On which marketplace are returns eroding the real profitability of advertising spend? Which distributor sends a risk signal before placing a reorder? In which production order does a delivery delay keep recurring?

Analysis

Data-driven forecasting and early-warning systems can help find answers to these questions sooner. But before any AI implementation, product codes, cost definitions, channel-level order data and reporting discipline must be reliable. A fast system built on bad data only speeds up the wrong decision.

For me, the success measure of a technology project is not how many new tools were installed. It is how much cash was released from stock, which class of errors decreased, how much decision time shortened and how net commercial contribution changed.

AI's First Job Should Not Be Writing Ads, but Finding Lost Margin

When AI is discussed, most companies first think of image generation, ad copy or a chatbot. These can be useful. But in cosmetics and retail companies, the bigger opportunity often sits in the invisible corners of operations.

AI

Data

Commercial Management

Insight

Good technology doesn't have to be visible. Sometimes the best transformation is the stock crisis that never happened or the wrong discount that was never given.

Which product is overstocked in which store? Which SKU has high sales but low net contribution? On which marketplace are returns eroding the real profitability of advertising spend? Which distributor sends a risk signal before placing a reorder? In which production order does a delivery delay keep recurring?

Analysis

Data-driven forecasting and early-warning systems can help find answers to these questions sooner. But before any AI implementation, product codes, cost definitions, channel-level order data and reporting discipline must be reliable. A fast system built on bad data only speeds up the wrong decision.

For me, the success measure of a technology project is not how many new tools were installed. It is how much cash was released from stock, which class of errors decreased, how much decision time shortened and how net commercial contribution changed.

AI's First Job Should Not Be Writing Ads, but Finding Lost Margin

Related reading

AI's First Job Should Not Be Writing Ads, but Finding Lost Margin

Related reading

AI's First Job Should Not Be Writing Ads, but Finding Lost Margin

Related reading