Eight AI use cases global FMCG brands run today, from pricing to agentic AI, and why they need one causal model.
Roughly 75% of consumer goods companies surveyed are still in pilot or exploration mode with AI, according to Boston Consulting Group's (BCG's) June 2026 research with The Consumer Goods Forum. The margin sits in the gap between a pilot and a scaled use case.
The usual diagnosis is a shortage of good use cases. What we see is the opposite: pricing, promotion, media, and distribution each run their own model on the same SKU (stock-keeping unit), and each model takes credit for the same sales.
Each of the eight AI use cases that follow comes from a global brand's own public account.
FMCG brands use AI to answer eight recurring commercial decisions: what to charge, which promotions to fund, which products to launch, and where stock is missing. The other four are which stores reps visit, where media earns most, how much to make, and which content and tasks to automate. Most brands run each decision on its own model.
Use case | Decision it answers | Global brand example | Data it needs |
|---|---|---|---|
Pricing & revenue growth management | What price and pack each SKU carries | Colgate-Palmolive | Prices, volumes, input costs |
Trade promotion optimization | Which promotions earn back their cost | PepsiCo | Promotion calendars, trade spend, baseline sales |
Portfolio & new product development | What to launch, reformulate, or retire | Mondelēz International | Recipes, costs, consumer tests |
Distribution & on-shelf availability | Where stock is missing, and which fix restores sales | Unilever (Mexico) | Shelf images, orders, store data |
Sales force & route-to-market execution | Which outlets reps visit, and what they sell in | AB InBev | Outlet sales, visit logs, routes |
Media & brand building | Where the next media dollar earns most | Procter & Gamble | Media spend, reach, sales |
Demand sensing & supply planning | How much to make, and where | Procter & Gamble | Purchase signals, shipments, capacity |
Generative & agentic AI | Which content and tasks run automatically | Mondelēz International, Unilever | Brand assets, workflows, commerce data |
The eight AI use cases in FMCG below cover the commercial decisions a brand owner makes on every SKU, from price and promotion to shelf, media, and supply.
Revenue growth management (RGM) uses AI to set price and pack decisions from measured price elasticity: how much volume a SKU loses when its price rises, and where that volume goes. Price pack architecture, the ladder of sizes and price points a brand offers, sits inside the same decision.
Colgate-Palmolive's 2025 Annual Report says its predictive models return pricing and promotion recommendations "in minutes."
At PepsiCo, the company's data science team describes PricingAI as estimating own-price and cross-price elasticities across the portfolio.
When inflation or input costs rise, the question becomes which packs can carry a higher price.
Decision answered: Which price and pack each SKU carries, in each channel, as costs move.
Trade promotion optimization uses AI to decide which promotions to run, when, and at what depth, and judges each one on the sales it adds beyond what would have sold anyway. The baseline is the hard part: a promotion that lifts a week's sales may only have pulled purchases forward, or discounted shoppers who would have paid full price.
PepsiCo's PromoAI optimizes promotional calendars across trade channels, according to PepsiCo's data science team. PepsiCo has not disclosed the system's financial results.
Nestlé's 2025 Annual Report says its sales teams use AI to optimize in-store execution and promotional spend.
Decision answered: Which promotions earn back their cost once you remove the sales they pulled forward or subsidized.
AI shortens new product development by generating and screening recipes and formulations before physical testing.
Mondelēz International says 60% of the biscuit recipes its AI generated performed better on nutrition, sustainability, and cost, as Reuters reported in July 2026.
In personal care, Unilever said in April 2026 that its Beauty & Wellbeing business cut formulation cycles from five or six rounds to one or two.
The open portfolio question is whether a new SKU grows the category or takes sales from the brand's existing lines.
Decision answered: Which products to launch, reformulate, or retire, and what each does to the rest of the portfolio.
AI detects out-of-stocks by reading shelf images alongside order and store data, so a brand sees where product is missing while there is still time to fix it.
In Latin America, Unilever's August 2026 account of its AI-powered supply chain says its AI-enabled customer connectivity model delivered 98%+ on-shelf availability in a pilot in Mexico.
In India, Hindustan Unilever's 2025–26 Annual Report describes Envision, an image recognition system that detects out-of-stock items in priority stores and triggers immediate replenishment.
Decision answered: Where stock is missing, and which fix restores sales fastest.
In the general trade of emerging markets, where sales run through small independent stores, AI tells each sales rep which outlets to visit and what to recommend in each one. This is where FMCG differs most from US planning around a few national retailers.
AB InBev's 2025 Annual Report says reps using its BEES Force tool spend, on average, 42% more time with retailers on key priorities since 2023.
Decision answered: Which outlets each rep visits, in what sequence, and with which recommended order.
Media AI in FMCG sets where the next dollar goes by estimating what each channel adds to sales, net of price, promotion, and distribution.
Procter & Gamble (P&G) is scaling AI tools "from creative development to media activation," the company said on its fiscal 2026 fourth-quarter earnings call.
Retail media belongs in the same budget decision, since it competes with brand media for the same dollar.
The open question is which spend caused the sales. Our guide to how AI is transforming marketing mix modeling explains how to measure it.
Decision answered: Where the next media dollar earns the most incremental sales.
Demand sensing uses near-real-time purchase signals to adjust short-term forecasts. On its fiscal 2026 fourth-quarter earnings call, P&G described its Supply Chain 3.0 as connecting purchase signals to production planning and material ordering.
In Hammam and colleagues' November 2025 study in Scientific Reports, a hybrid ARIMA (autoregressive integrated moving average) and XGBoost ensemble kept mean absolute percentage error below 13% on most datasets. Where demand was highly variable, it improved on ARIMA by up to 80%.
Decision answered: How much to make, and where to hold it.
Generative AI in FMCG produces content, and agentic AI in FMCG completes tasks without a person stepping in.
Mondelēz is using a generative AI tool to cut marketing content production costs by 30–50%, Reuters reported in October 2025.
Unilever's five-year partnership with Google Cloud, announced in February 2026, covers agentic commerce and agentic workflows. Nestlé's 2025 Annual Report says its agentic virtual sales assistant automates up to 40% of routine tasks in pilot markets.
An agent that reorders stock or books a promotion still needs a model that says which order or promotion is right. Our guide to agentic AI in marketing shows where agents fit.
Decision answered: Which content and routine tasks run automatically, within the limits the other seven decisions set.
FMCG AI use cases collide because pricing, trade promotion, media, and distribution models each plan the same SKU while assuming the other levers stay fixed. Each model learns correlations from a history in which the other levers were also moving, so each claims part of the same growth.
Take one 500g pack of biscuits, in one market, in one month. In this illustration, three models plan it independently, and the last two rows compare their sum with one model reading all three together.
Use case | Model's recommendation | What the model expects |
|---|---|---|
Pricing | Raise the list price 5% to cover input costs | Volume falls 2% |
Trade promotion | Run a four-week feature at 25% off | Volume rises 30% in promoted weeks |
Media | Put $1M behind the brand that month | Sales rise 6% |
All three, added up | Each plan run on its own | Volume rises 34% in promoted weeks |
One causal model | The same three levers, read together | Volume rises 20% in promoted weeks |
Each forecast assumes the other two decisions are not happening. The 25% discount more than cancels the 5% price rise on the shelf, and the media model will credit itself with volume the discount bought.
Causal AI models cause and effect across all the levers, so it can say what the price rise did while the promotion was running.
That is the job of Enterprise Decision AI: one causal model of what drives growth across the business, as our guide to Enterprise Decision AI in practice shows. POEM365 delivers it: DATA POEM's Large Causal Model (LCM), built on FOUNT's Large Causal Architecture (LCA).
The SKU collision check takes each model's recommendation for one SKU in one period. It tests the recommendations against each other in one causal model and shows which use case drove growth, so you can fund that lever and stop paying twice for the same volume.
To run the SKU collision check on your AI use cases in FMCG, ask three questions:
The third question is a WHY question, and DATA POEM’s Causal Clarity shows exactly what is driving the business to answer it.
Enterprise Decision AI in FMCG replaces eight separate models with one causal model that plans pricing, trade promotion, portfolio, distribution, sales force, media, and supply decisions together. That is how DATA POEM works with FMCG brands: POEM365 starts from a pre-trained causal model and fine-tunes it on a brand’s own sales, spend, price, and distribution data. The result is the brand’s Enterprise Growth Model, which reads every use case against the same baseline and shows the drivers behind each recommendation.
For an FMCG brand, those decisions run on three planning cycles. Strategic decisions, made each year or quarter, cover revenue forecasting across products, channels, and geographies, new product and market launches, and where to invest across brands, categories, and markets. Functional decisions, made each quarter, cover one marketing plan across paid, owned, and earned channels, demand forecasting that accounts for media, distribution, and competitors, and how a price or promotion on one product lifts or takes sales from another. Tactical decisions, made each week or month, cover moving media money mid-campaign, measuring the sales that retail media, search, and programmatic spend actually adds, and demand planning from 4 to 52 weeks out with the promotional calendar built in. The biscuit example above is a functional decision, and the same model carries it up to the portfolio and down to next month’s media plan.
Inside POEM365, four Decision Agents do the work across those cycles:
The generative and agentic tools in use case 8 then act inside that plan, so the agent that books a promotion books one the model has already tested against price, media, and stock.
The ROI of AI in FMCG is large on paper, and few companies capture it yet. Bain & Company estimated in October 2025 that consumer products companies that scale AI could expand EBITDA by 3–5 percentage points, yet 48% remain at the "exploratory stage" of AI maturity. The return depends on measuring what each use case caused.
FMCG companies leading on AI publish specifics: Unilever, P&G, PepsiCo, Nestlé, AB InBev, and Mondelēz each describe named systems and operating results in their own filings and calls. Most report time saved, availability, or cost; few publish the incremental revenue a use case added.
The difference between FMCG and CPG is mostly geography, since both describe fast-selling everyday products. FMCG (fast-moving consumer goods) is the standard term in the UK, the rest of Europe, and India; CPG (consumer packaged goods) is the US term. For AI in consumer goods, the practical difference is route to market, since FMCG playbooks include general trade.
Each use case of AI for FMCG already pays off on its own, in faster pricing, fuller shelves, and cheaper content. The larger gain comes when you plan the commercial decisions on each SKU together, so the price, the promotion, the media plan, and the stock position all work toward the same forecast.
Growth Planning builds one plan across the entire business, answering WHERE to invest next.
DATA POEM reads every SKU decision together in POEM365, its Large Causal Model for planning growth across the business.
Which of your SKUs has two models pulling in opposite directions? Ask the DATA POEM team to test it in one causal model.
Let's talk about how we can help you grow your business.
How CPG brands use causal AI to plan trade and retail-media investment across Amazon, Walmart Connect, and Kroger in one model, not three.
A CMO's playbook for reallocating an existing marketing budget by causal contribution, defending it to the CFO, and holding it in a downturn.
AI is reshaping marketing mix modeling, from open-source Bayesian models to causal structure learning. What's changing, and what it means for your budget.
See how Fortune 500 retailers use AI for demand forecasting, personalization, and supply chain, and where most retail AI investment still falls short of the P&L.