Industry
Retail
Omnichannel CX and Intelligent Innovation – The Future of Retail
HIGHLIGHTS
- Merchandising intelligence has progressed from expert rules and offline planograms to KPI dashboards and AI models for predicting demand, elasticity, and affinity.
- While each step improved accuracy, few challenges such as decision latency and siloed optimization persisted.
- Agentic AI in merchandising shortens the sense–decision–action cycle, allowing retailers to respond at market speed while maintaining governance and control.
- Real world applications include demand substitution modelling, store clustering, and seasonal adaptation.
On this page
The need for speed
Buyers, range planners, and display managers work together to curate perfect assortments while considering space constraints, supply chain and market dynamics, business strategy, and store profiles. Their work is part science, part experience, balancing product availability, operational efficiency, promotions, and stock holding. However, today’s market moves faster than the tools that support them. Social media trends, for example, can upend a category overnight; a sunny day can spike salad sales.
With Walmart’s Sparky and Amazon’s Buy for Me artificial intelligence (AI) agents shopping on behalf of the customer, stores must respond with adaptive assortment and space strategies that match the fluidity of digital. These challenges, coupled with the complexity of operating across multiple channels and delivering seamless experiences, call for speed to keep up with competition, supercharge decision-making, and be more perceptive. Agentic AI in merchandising is a game changer; leading retailers are investing in agentic AI to transform their operations and offer hyper-personalised experiences that customers expect.
Agentic AI is enabling the pivot from predictive models to goal-seeking systems that perceive context, reason across constraints, run scenarios, and propose actions with human oversight.
The journey so far—evolution of intelligence in merchandising
Merchandising intelligence has progressed from simple rules and individual knowledge or experience to key performance indicators (KPI) dashboards and predictive models to determine demand, elasticity, and affinity. While each step improved accuracy, few challenges such as siloed optimization and decision latency still persisted: assortment, price, and space were all tuned separately, often on stale data, with limited visibility into trade-offs, impacting decision cycles.
Agentic AI changes the approach from models that only predict to goal-seeking systems that perceive context, reason across constraints, run scenarios, and propose actions with human oversight. It shortens the sense-decision-action cycle, allowing retailers to respond at pace to market changes while maintaining governance and control.
This means, range analysts and buyers can now be more perceptive. They can uncover the true substitutability of products, ensuring every range when cascaded into smaller format stores reflects customer intent, preserves choice, and safeguards sales. They can manage the unpredictability of the supply chain and quickly respond to a supplier factory fire or a delayed harvest, optimising the allocation for customers. Similarly, they can stock shelves with products based on customers’ behaviour patterns and not just static demographic details drawn from historical data.
Agentic AI can help merchandisers tackle the toughest, high-value use cases that directly shape range relevance, space efficiency, and drive sales and revenue growth. Retailers can unlock measurable gains such as fewer stockouts and write-offs, stronger gross margins, and faster turns. Top use cases for agentic AI in merchandising include:
- Demand substitution modelling using attribute and image embeddings to anticipate cross-elasticities when stock keeping units (SKUs) are added, shorted, or de-listed
- Dynamic store clustering that adapts to local missions and behaviour rather than static segments to respond in real time by optimising assortments, pricing, and inventory according to changing consumer preferences
- Seasonal and event-driven adaptation by pre-emptively shifting facings and secondary placements based on weather, festivals, and local events
- Price and promotion optimisation by simulating competitor moves and supply constraints to balance margins and market share
- Markdown optimisation to maximise sell-through while protecting brand equity
Agentic AI delivers exponential gains on its own, but when it is fused with digital twin platforms, its power compounds. Retailers can simulate store space layout and optimisation, new product or variant introductions, pricing shifts and planogram optimisation using digital twins before executing decisions in the real world, thus reducing risk and accelerating time to value. The result is a powerful decision-support ecosystem, overlaying the recommended assortment and planograms, promotions, and spatial planning with precision, helping enhance profitability and customer experience.
Building blocks of agentic AI-powered merchandising
An agentic AI system for merchandising is a network of autonomous software agents aligned to explicit business goals (for e.g., category growth, gross margin, availability). Each agent continuously listens to signals, infers the current state, simulates alternatives, recommends actions, and—where policy allows—acts, with human-in-the-loop controls and auditable guard rails. In plain terms, it is a digital team that works alongside planners, accelerating routine decisions while escalating exceptions that need judgement. With explainable AI offering decision makers the ‘why’ of every recommendation, key decisions are anchored in trust and confidence.
The building blocks of agentic AI-powered merchandising systems include:
- Listening agents that fuse point of sale (POS), inventory, planograms, supply events, weather, competitor signals, and local context to keep the ‘world model’ fresh
- Optimisation agents that run multi-objective simulations on assortment, space, price, promotions, and markdown, surfacing actions and quantifying trade-offs
- Continuous learning and feedback loop that captures outcomes and feedback, continually improving policies and results
Compared with rule-based or static models, this design is more responsive, scalable, and adaptable—and easier to govern because intent, constraints, and rationale are explicit.
By leveraging a multi-agent approach, a leading electronics retailer transformed their range planning into a strategic growth lever—moving from reactive execution to an intelligent, proactive capability. At the core was a goal-based range optimisation system, powered by proprietary, domain-specific AI models and multi-agent architecture. Specialised agents handled data ingestion and category‑specific store clustering while identifying item cohorts (comparable items inside each category), SKU‑to‑SKU transfer rates, and building consumer decision trees (CDTs). By continuously sensing market signals, simulating potential scenarios, optimising range per store and cluster against a goal function (profit vs inventory cost), governing workflows, and capturing human‑in‑the‑loop inputs, the solution recommended intelligent range choices, driving smarter assortment decisions. With store-wise SKU lists, display-ready views, and scenario-based projections for the upcoming merchandising period, the retailer realised improved profitability via optimised SKU mix per cluster and store and lower inventory carrying costs while maintaining availability.
The road ahead
Perceptive retailers will be able to thrive in a fast-changing world with foresight, precision, and agility. Over the next 12–36 months, leading retailers will shift from calendar-bound planning to continuous micro-planning. Assortment, facings, and secondary placements will refresh with event and demand signals; digital twins of stores will enable teams to A/B test planograms before execution; and autonomous pricing and markdown agents will rebalance margin, sell-through, and competitiveness in near-real time within policy guardrails. With agentic AI, merchandising, supply chain, and retail media will no longer operate in silos; inventory reality and audience reach can be woven together to orchestrate profitable and targeted promotions.
Operating models will evolve in parallel. Merchant teams will run a ‘mission control’ that supervises agents, sets objectives and constraints, and curates exceptions. They will pivot from manual number-crunching to scenario design, judgment calls, and vendor negotiations; obsolete KPIs will be replaced with ones that matter such as decision latency, explanation quality, and governance adherence to classic margin and availability metrics. With accurate product dimensions, customer need states, or live macro space, the improved accuracy of thousands of data points will underpin the effectiveness of agentic AI for space and assortment.
Guidelines for pivoting to agentic AI
The road to agentic AI is not without challenges. Success will hinge on robust foundational data (product, space, and event signals), clear policies, change management, and ethical guardrails. A disciplined playbook can help make the pivot easier and more effective.
- Ensure you have robust foundational data—accurate product and macro space dimensions, and customer need states—to influence quality decision-making
- Start with two or three lighthouse categories—identify high-impact, high-visibility projects, prove value, and then scale
- Set guardrails early—margin floors, safety stock limits, promotion caps—so automation stays safe
- Adopt a sandbox approach—run agents in ‘suggest-only’ mode before giving execution rights
- Ensure interoperability—agents need application programming interfaces (APIs) and event streams across legacy systems
- Invest in observability—monitoring, explainability, and audit trails build trust
The future of merchandising is not about faster decisions alone, it is about smarter, trusted ones. Retailers that embrace agentic AI are not just optimising today’s shelves, they are future-proofing growth. By empowering merchandisers, planners, and category managers with autonomous, explainable intelligence, agentic AI turns complexity into clarity, uncertainty into opportunity, and merchandising into a true competitive advantage.