Nitin KumarData Narrator
Meld u aan om de nieuwste inzichten en updates over technologie, AI & data-analyse, datawetenschap en innovaties van Polestar Analytics te ontvangen.
Vat dit blogbericht samen met:
AI-native RGM turns intelligence into action by putting pricing models and promotion models and elasticity models into live decision workflows instead of static dashboards. This way recommendations arrive before the market changes.
TL;DR Most CPG companies already have commercial intelligence. What they lack is the operating model to act on it before the decision window closes. AI-powered RGM bridges this gap by incorporating flexibility, incrementality, and scenario modeling into the actual process of making decisions on pricing and promotion, rather than into an analysis of those decisions.
Ajinomoto does not run pricing scenarios once a quarter and hope the market holds still. In a recent RGM roundtable hosted by Polestar Analytics, Harry Ergan, VP of Revenue Management at the company, described running pricing scenarios across 60 SKUs in 15 markets overnight, comparing revenue, volume, margin and mix impact before commercial teams walked into customer negotiations, not after. That single anecdote is a useful stress test for every CPG boardroom still debating whether AI in revenue growth management is a strategic priority or a slide in next year's roadmap.
Here is the uncomfortable part. The data mostly exists already. McKinsey found that 88% of organisations reported using AI in at least one business function in December, 2025. In CPG specifically BCG found in 2026 that 75% of companies are still stuck in pilots and exploration with only 18% scaling AI to significant impact. And yet global FMCG value-sales growth in 2025 came in at a thin 3.5%, with volume growth at just 0.9%, both slower than the year before. Adoption went up. Growth did not follow it up the stairs. That gap is the entire argument for AI-native revenue growth management, and it deserves more scrutiny than another "AI is transforming the industry" slide.
Most enterprises are not short on commercial intelligence. They are short on the ability to move on it inside the window where it still matters.
Trade data sits in TPM tools. Pricing lives in spreadsheets. Customer insight sits inside CRM. Sales, finance and marketing each hold a partial view of the same commercial reality, and by the time all three get reconciled into one number, the market has usually moved on. Only about 10% of CPG and retail companies have integrated AI agents across their workflows, according to BCG, which tells you the fragmentation problem is still the norm, not the exception.
This is also where the language gets sloppy. Commercial analytics and commercial intelligence get used interchangeably, but they are not the same job. Analytics explains what happened last quarter. Intelligence is supposed to change what happens next quarter. A company can be excellent at the first and still be losing money on the second, which is exactly what the trade promotion numbers below show.
AI-powered usually means a model got added somewhere in an existing process. AI-native means the process was redesigned around the model from the start, so the recommendation shows up inside the workflow where a pricing or promotion call gets made, not in a separate report someone has to go find.
Polestar Analytics' AI in Revenue Growth Management Playbook captures this well through a conversation with revenue leaders from Ajinomoto, Nestlé and Unilever. Built around what the piece calls the Four-Pillar RGM Scale Framework:
Skip a step and the AI layer becomes a very fast way to be confidently wrong.
In practice, AI-native RGM analytics looks like this:
Trade promotion is the answer, and it is not close. Trade spend is one of the two or three largest controllable lines on a CPG P&L, often 20 to 30% of gross sales. Yet only about 46% of trade promotions deliver a positive ROI. Run that math on a $5 billion business and you are looking at hundreds of millions of dollars a year riding on promotions that on average are a coin flip.
Pricing has a quieter version of the same leak. Pack-price inversions, where a smaller pack costs more per unit than a larger one, sit unnoticed inside portfolios for years because nobody is watching the ladder continuously. Polestar Analytics' work on Price Pack Architecture and its deeper piece on AI-enabled price elasticity both make the same underlying point in different ways: the model was never the hard part, the discipline to watch it continuously was.
| Where the leak hides | Traditional RGM analytics | AI-native RGM |
|---|---|---|
| Promotion effectiveness | Reviewed post-event, once the budget is spent | Uplift and cannibalisation measured continuously, mid-cycle course correction possible |
| Price elasticity | Portfolio average, refreshed annually | SKU x geography x channel, refreshed on live sales data |
| Pack architecture | Periodic consulting audit | Standing signal that flags inversion and whitespace as it emerges |
| Competitive response | Reacted to after share is already lost | Simulated before a response is committed |
| Decision speed | Weeks, bound by the next planning cycle | Days, bound by when the recommendation is trusted |
This is also the honest case for the two Pulse apps built specifically for this problem. PromoPulse reads incrementality and ROI early enough to matter, and PricePulse keeps elasticity and pack-price logic live rather than quarterly. Neither replaces a pricing or trade marketing leader's judgment. Both exist because that judgment deserves better inputs than a stale spreadsheet.
This is the part where the industry gets a little ahead of itself and it is worth saying plainly. Agentic AI for RGM, where systems flag under-delivering promotions, test pricing scenarios and recommend adjustments before a cycle closes, is a genuinely exciting direction. Capgemini estimates AI agents could generate up to $450 billion in economic value through revenue uplift and cost savings by 2028.
But Gartner also predicts that more than 40% of agentic AI projects will be cancelled by the end of 2027, largely because of unclear business value, escalating cost or weak risk controls. Both numbers are true at the same time, and CXOs should sit with that discomfort rather than resolve it too quickly. Poor data does not become strategic just because an agent sits on top of it.
Polestar Analytics' work on building autonomous revenue managers on Databricks and on orchestrating revenue growth agents with Databricks Mosaic AI both treat autonomy as the last mile, not the entry point, for exactly this reason. The architecture matters more than the ambition.
Not "buy an agent." Fix the sequence first, because sequence is the thing every failed AI rollout skipped.
Commercial intelligence has never really been the scarce resource in CPG. Acting on it inside the window where it still changes the outcome, that has always been the harder problem and it still is. AI-native RGM does not solve it by adding intelligence. It solves it by shortening the distance between knowing and doing.
No. A tool bolted onto an existing process is AI-powered, not AI-native. AI-native means elasticity and incrementality logic are built into the decision workflow itself, which is what Polestar Analytics' PricePulse and PromoPulse are built around.
Depends on the lever. Pricing shows margin impact within a quarter, since it closes an existing leak, which is why PricePulse is often the fastest deployment. Promotion and agentic workflows take longer, since they need a trusted data foundation first.
Start where the leak is largest and the data is cleanest. Usually that's trade promotion, given roughly half fail to break even, the gap PromoPulse targets. Pricing is often the faster win regardless.
Over de auteur
Data Narrator
The best ideas don't arrive. They wait for you to be ready.
Gerelateerde blog
Astha Chadha