Merchandising has always driven retail performance. It determines what customers see, what they buy, and how retailers create value. At its best, it aligns assortment, pricing, promotion, and inventory into a single strategy.
Today, that alignment depends largely on people. In the future, we expect AI agents to transform how decisions are made. Merchandising can then shift from a series of siloed processes to a more integrated, always-on system, with merchants focused less on assembling inputs and more on overseeing agentic outputs.
How Merchandising Works Today
For years, the category merchant for any given product line owned the decision. They pulled together sales data, competitor pricing, vendor terms, inventory levels, and margin targets and made weekly tradeoffs.
Over time, parts of this process have evolved. AI and machine learning now generate pricing recommendations based on predefined guardrails and targets. But those recommendations still pass through several review layers—from the category merchant to the chief merchant—before they are executed.
In parallel, space planning, promotions, and forecasting operate as separate processes. However, the category merchant remains responsible for stitching all the different elements together to create a final offer.
Then conditions change—a heat wave occurs or a competitor cuts prices—and the process has to reset. Assumptions shift, decisions are revisited, and the cycle begins again.
This model was built for stability. It is slow to sense change, to filter signal from noise, and to respond in real time.
Stay ahead with BCG insights on the retail industry
The Vision for Agentic Merchandising
In the future model, tasks that today are carried out by the category manager are expected to shift to a set of specialized AI agents. Each merchandising responsibility can then be managed by a dedicated agent that is provided with access to data and tools, business context, guardrails and decision rights, and evaluation criteria.
A pricing agent continuously scans for changes in competitor price, cost, elasticity, line structure, and category performance. When conditions change, it recommends the optimal price response, in line with operational and strategic constraints.
A promotion agent evaluates true net incrementality and calendar conflicts. An assortment agent assesses productivity per square foot, assortment overlap, and whitespace. An inventory agent monitors shipments and on-shelf availability, proactively flagging and resolving potential stockouts. Each agent is aware of the actions of the others: for example, when the inventory agent foresees a stockout, the promotion agent may delay a promotion.
Above them sits an orchestration agent that takes over the coordination and synthesis activities. It monitors the outputs of each agent, resolves conflicts, and ensures the combined outcome is aligned with the category strategy.
These agents monitor the external environment continuously. They detect competitor moves, weather shifts, and demand changes more quickly and more consistently than humans can. They evaluate tradeoffs across pricing, promotion, and assortment without bias or fatigue. They predict outcomes based on data rather than instinct.
Decisions that once took weeks can happen in hours. Core processes that overhaul categories, such as annual line reviews, that can take half a year from start to finish, can take weeks—and may disappear altogether as merchandising becomes an always-on process. The economic impact is set to compound across levers. Even small improvements in price realization, promotional effectiveness, and inventory turns can translate into material value at scale. The advantage of automated orchestration isn’t a single breakthrough but rather the steady elimination of value leakage across thousands of decisions and the speed at which changes can be rolled out to stores and end customers.
The advantage of automated orchestration is the steady elimination of value leakage across thousands of decisions.
Retail can become less about reacting to yesterday’s insights and more about shaping tomorrow’s outcome. While a few leading players are starting to build agentic capabilities, most haven’t even started. They need to act with urgency and prepare for the AI-led transformation in merchandising to remain competitive.
What Is Required to Bring the Vision to Life?
This future starts with new technology, but it doesn’t end there.
First, the strategy must be explicit. Agents execute rather than invent a differentiated strategy. Leaders must set priorities explicitly: growth versus margin, short term versus long term, how aggressive to be on price leadership, and what customer objectives to drive with promotions. Supported by explicit objectives and guardrails, agents can convert strategy into action.
Second, effective underlying quantitative engines are needed. Pricing, promotions, cost, inventory, and assortment tools must produce recommendations that are reliable and explainable. Our agents use underlying, proven algorithms that don’t hallucinate and have built-in safeguards. Weak engines, once connected, fail faster and create chaos at scale. Engines don’t need to be fully developed to provide AI benefits, but they need to be sufficiently advanced to act as a starting point for agentic merchandising.
Third, data and definitions should be standardized and structured for decisions. Category roles, margin definitions, incrementality, and price families must mean the same thing across the enterprise. Without a shared language, automation fails.
Finally, the operating model must evolve. Most merchandising organizations remain siloed by function. By contrast, agent-based systems can cut across pricing, promotion, assortment, space, and the supply chain. That requires clear end-to-end ownership, tight alignment between business and technology, and fast decision rights (for example, better alignment between the promotional, pricing, and marketing outcomes, metrics, and decisions). While software can coordinate and make decisions, the organization must put in place the foundation that enables those decisions to happen.
Software can coordinate and make decisions, but the organization must put in place the foundation that enables those decisions.
Incremental change is not enough. Merchandising organizations need to fundamentally rethink how teams are structured and reset their core business processes and rhythms.
The Merchant’s New Role
AI will not eliminate the need for merchants, though it is set to change their role. As agents take on time-consuming operational tasks, such as reporting and preparations for supplier negotiations, and handle tradeoffs, merchants will focus on more high-level, strategic activities.
We expect the merchant’s role to be shaped by the following trends.
Vendor Relationships. Negotiations, partnerships, and conflict resolution all depend on trust and context, and so can remain part of the merchant’s remit. However, the nature of the supplier-retailer relationship is set to change based on the pace at which both sides adopt AI tools. Once vendors also have agents, there will be an opportunity for retail and vendor agents to handle much of the work, elevating the role of humans to maintaining a strong relationship.
Brand Stewardship and Divergent Thinking. AI agents can detect trends but they cannot as yet define or develop an identity. The establishment of a retailer’s point of view—curating products, developing brand values, and choosing whether goods are sold online or in physical stores—still needs to be done by a human, especially for categories where taste matters. To prevent AI homogenization and help drive differentiation, merchants’ creative thinking will increasingly be in demand.
Portfolio Expansion. With agents responsible for monitoring and analysis, merchants can oversee more product categories and make investment and resourcing decisions across a broader portfolio. Senior merchants can spend more time on strategy and alignment among merchandising activities, while junior merchants can take over interrogating outputs and learning how to apply their judgment to ensure that data-driven recommendations achieve the best outcome. Retailers need to be on the alert to ensure merchants create value rather than slow decisions or add human bias, although both can be addressed via process design.
Merchandising leaders can achieve greater efficiency. But they should view this as a byproduct of better and faster decisions and not as the primary goal.
Making It Real
To bring agentic AI-enabled merchandising to life, retailers need a set of purpose-built agents, including:
- A value agent that continuously monitors price and promotion posture by reviewing external signals, such as competitor prices, weather, and consumer trends, and suggests real-time price and promotion changes within defined guardrails
- A cost and negotiations agent that manages cost changes, commodity movements, and funding and supports the ask generation and commodity analysis for negotiation scenarios
- An assortment, space, and inventory agent that balances SKU rationalization, planogram productivity, new product innovation, and capital deployment and that is aware of shipment lead times, supplier innovation schedules, and execution limitations and makes recommendations accordingly (for example, planogram development)
- A consumer sentiment agent that ingests search trends, social media signals, competitor moves, and external demand drivers to separate signal from noise
- A store execution agent that monitors execution success and surfaces store feedback as inputs for other agents
Above them is the orchestrator agent. It continuously monitors the recommendations across price, promotion, cost, space, inventory, and store agents to ensure that the portfolio outcome aligns with the strategy, risk appetite, and operational constraints. Instead of siloed optimizations, decisions are evaluated holistically by this agent.
Merchants engage with the orchestration agent through a unified interface. Rather than pulling reports, they see recommended actions, rationales for change, projected impacts, and flagged exceptions. The interface becomes less of a dashboard and more of a decision cockpit focused on intent, tradeoffs, and accountability. The actions consider the realities of physical retailers and abide by the associated limitations.
What takes many separate processes and handoffs today becomes a coordinated, continuously learning system—with humans governing strategy and differentiation and agents executing at scale.
Where to Start
Most retailers won’t move to an agent-enabled model all at once. In practice, implementation is likely to evolve incrementally.
- Align across functions on the future vision for merchandising and the use of AI agents to drive core decisions.
- Map the current-state operating model, talent, tech architecture, and data structures, identifying where change is likely to be needed to support agents and realize impact.
- Start with one agent that addresses a part of the merchandising skill set and targets tangible P&L impact. Prove the agent’s effectiveness by comparing outputs with those of current processes for a subset of categories.
- Once impact is proven, scale to more categories and merchandising processes, ensuring value delivery throughout the journey.
- Along the journey, prioritize change management to drive adoption and impact: redesign processes, upskill teams, and build trust.
The Bigger Picture
The future of merchandising is not about replacing people with machines. It is about redesigning how decisions are made so that retailers can enhance performance and act more effectively. Technology is advancing quickly. Companies need to move with urgency—and reset how they operate—to remain relevant and competitive for their customers.
