Why Category Management Is the Last Frontier of Commercial Planning, and Why AI Changes Everything

Most enterprise planning functions have been through at least one major wave of digitization. Demand planning has its models. Supply chain has its control towers. And yet category management, the discipline that connects consumer insight to pricing, assortment, and promotional investment, still runs largely on spreadsheets.

That is not an oversight. It is a structural problem. And in the operating environment of the mid-2020s, it is becoming an increasingly costly one.

The final frontier

Category management is the powerhouse of the business. No other planning process a retailer or consumer goods company runs shapes revenue, margin, and consumer experience as directly, or drives as much overall value. But that power comes with a uniquely demanding position: category management sits at the sharp intersection between two competing forces, with consumers on one side and suppliers on the other, and every decision has to reconcile the two. Category managers are the CEOs of their product groups, treating each category as a strategic unit and owning the assortment, pricing, and promotion decisions that determine commercial outcomes. Yet the tools they work with have not kept pace with that responsibility. 

Despite all the technology investment that has flowed into enterprise planning over the past decade, an estimated 90% of grocers still manage category management in Excel. Not because the data does not exist, but because data, decisions, and departments have remained disconnected. Analytics teams generate insights. Commercial teams build plans. But each of these activities happens in isolation, and the effort required to align them is slow, manual, and chronically out of step with live market signals.

The result is that category managers spend a disproportionate share of their time on data gathering and scenario building, the least strategic parts of the role, rather than on supplier relationships, portfolio shaping, and consumer trend navigation.

This is the core argument for digitization. But if that argument has been clear for years, what has changed is the urgency.

A cost environment that cannot wait

The macroeconomic backdrop of the mid-2020s has fundamentally altered the stakes for category planning. Supply chain costs are rising sharply ahead of inflation. Tariff-driven import cost increases have hit grocery and consumer goods particularly hard, with research tracking retail price responses showing that prices began adjusting within days of policy announcements and kept moving for months afterward. A KPMG survey of C-suite executives found that half of retail respondents reported a gross margin decline of 1 to 5% directly attributable to tariffs, the highest figure across all sectors included in the study.

For category teams, this creates a specific and acute problem. Assortment decisions that were historically made on a seasonal or annual cadence were built around stable cost structures. That stability no longer exists. A SKU that was margin-accretive in the first quarter may be margin-dilutive by the third if its supply chain has been disrupted or its import cost has risen sharply.

This is the crux of the matter. The only way to manage that dynamically, across hundreds of categories and thousands of SKUs, is through connected digital platforms that ingest updated cost data the moment it changes and immediately recalculate the commercial implications across pricing, promotion, and distribution. No spreadsheet, and no sequence of manual handoffs between analytics, commercial, and finance teams, can keep pace with cost inputs that move within days.

Every hour that a category manager spends rebuilding a model by hand is an hour in which the underlying costs may have already moved again. When margin can swing from accretive to dilutive on a single SKU inside a quarter, the ability to see the full commercial impact of a cost change in real time, and to act on it before it erodes the P&L, is not a competitive advantage to aspire to. It is the baseline requirement for protecting margin at all.

The question of whether to absorb a cost increase, pass it through to the consumer, offset it through promotional mechanics, or rationalize the assortment to protect margin cannot be answered correctly by a category manager working from a week-old data export. It requires live cost visibility, real-time demand modeling, and the ability to run and compare multiple scenarios against a live P&L. This is not a future capability. It is a present requirement.

AI In Category Management: Creating Clarity From Complexity

Category Management stands as a critical yet underexplored area, one of the last planning processes to undergo full digitization.

This white paper explores its potential, highlighting how machine learning (ML), large language models (LLMs), and agentic AI can transform it.

Download now

AI is not one thing, and getting this wrong is costly

Here is where most enterprise AI programs go sideways. Organizations tend to conflate "AI" with large language models. ChatGPT, conversational assistants, natural language queries. These tools are visible, easy to demonstrate, and genuinely impressive in the right contexts. The problem is that for category management, leading with LLMs before the quantitative foundation is in place produces systems that feel capable but cannot be trusted when commercial decisions depend on them.

AI in category management is better understood as a layered capability spectrum with three distinct tiers, each dependent on the one beneath it.

The first tier is machine learning. This is the quantitative engine of the stack. ML models operate on the numerical and tabular data that category management generates at scale: scan data, promotional uplift histories, elasticity coefficients, inventory movements, basket analyses. A price elasticity model trained on three years of a retailer's own data, calibrated to their specific consumer base and promotional mechanics, produces materially more accurate predictions than any generic model applied cold. The same logic applies to promotion uplift forecasting, distribution gap detection, and supply-sensitive demand sensing. This is not interchangeable with language models. It is the substrate on which everything else depends.

The second tier is large language models, but deployed correctly. An LLM fine-tuned on retailer-specific data and commercial planning language, and grounded in the verified outputs of the ML layer beneath it, becomes a genuinely powerful interface. It can explain why a promotion underperformed. It can synthesize the cross-category implications of a tariff-driven cost increase. It can draft a supplier negotiation brief grounded in real elasticity data. Research has found that generic LLMs show a 20 to 35% accuracy drop on specialized enterprise tasks without domain fine-tuning, which is a significant gap when the outputs are informing pricing and margin decisions.

The third tier is agentic AI: systems that observe, plan, execute, and learn autonomously in pursuit of defined commercial goals. An agent in a category management context might monitor sales performance against target, detect a distribution gap in a key SKU, simulate the demand impact of a corrective promotion, identify the supply constraint that limits its viability, draft a supplier brief, and escalate the margin implication to the finance team. All without a category manager initiating each step. Leading organizations deploying this complete stack correctly are capturing 5 to 20% additional value and 15 to 30% efficiency gains from agentic commercial workflows.

The critical point is sequence: ML first. Domain-trained LLM second. Targeted agents third. Shortcuts do not save time. They defer failures to higher-stakes moments.

What practical, AI-driven category management actually looks like

Consider a real scenario. A consumer business enters the year with stretching revenue targets and a plan that already reveals gaps. Mid-year, a tariff-driven cost increase hits a key imported category. With connected, AI-powered planning, the response is immediate and multi-dimensional. The system recalculates the category P&L under the new cost structure, models consumer demand responses across different pricing scenarios, identifies promotional mechanics that can offset volume risk without sacrificing margin, flags the assortment SKUs where the cost increase makes distribution unviable, and proposes a supplier negotiation position grounded in elasticity data.

What previously required weeks of manual scenario analysis across disconnected teams can be resolved in hours.

This is the practical value of an initiative-based framework that connects customer needs, demand signals, supply realities, and live cost inputs simultaneously. AI-enabled decisioning for promotions alone has been shown to boost gross profit by 2 to 5% per market. When category managers are freed from data gathering and scenario building, they can focus on the work that actually requires human judgment: diversifying portfolios, strengthening supplier relationships, negotiating better terms, and reading long-term consumer trends.

It is worth noting that 79% of retailers and manufacturers now prioritize AI investments in category management. The question for most organizations is no longer whether to invest, but how to do it in the right sequence.

The compound advantage

There is a version of this journey where organizations get the sequence right, build the quantitative foundation first, layer in domain-trained language capabilities second, and introduce targeted autonomous agents third. In that version, each tier builds on the reliability of the one beneath it, and the whole system compounds in value over time. Forecasts get more accurate. Scenario planning gets faster. Decisions get better.

And there is a version where organizations skip steps because LLMs are immediately accessible and easy to demonstrate, deploy them on top of disconnected data and generic models, and find that the outputs sound authoritative but cannot be trusted when margin is on the line.

The operating environment does not leave much room for the second version anymore. Volatility is no longer a disruption. It is the default. And category management, properly digitized and properly AI-powered, is one of the most significant levers available for turning that volatility into a source of competitive advantage rather than a recurring source of margin pressure.

For a deeper look at the full AI capability framework, the role of each tier in category management transformation, and the practical applications that connect simulations to real commercial outcomes, the complete analysis is available in our new white paper.

Why Grocery Assortment Planning Has Fallen Behind — and What Intelligent, Integrated Decision-Making Can Do About It

Traditional grocery assortment planning is failing. Discover how an integrated decision layer connects assortment, pricing, promotions, and space to provide a single, commercially grounded view that protects your margins and drives retail growth. Download our whitepaper.

Download now