Which merchandising tasks can AI agents take over, and which can they not?
Agents can monitor, analyse and propose. Merchandisers still own strategy, trade-offs and supplier relationships. Where the line runs in 2026, and what has to be in place first.
KEY TAKEAWAYS Summary by the editors
- AI agents in merchandising are software systems that monitor data, run analyses and propose or execute actions within limits set by people, rather than dashboards that wait to be read.
- McKinsey estimated in January 2026 that merchants spend about 40 percent of their time on low-value tasks and that agentic AI could help them reclaim up to 40 percent of their time.
- Tasks suited to agents include performance monitoring, data cleansing, replenishment and re-order exceptions, scenario testing and drafting supplier materials.
- Category strategy, assortment trade-offs, brand positioning, supplier negotiations and decisions with fairness or reputational consequences remain human responsibilities.
- The main barriers are organisational: McKinsey found that 71 percent of merchants reported limited or no impact from current AI tools and only 24 percent received adequate training.
AI agents can take over a large share of the repetitive analytical work in merchandising: monitoring sales and stock, flagging exceptions, preparing re-order and replenishment proposals, testing scenarios and drafting reports or supplier materials. They cannot take over category strategy, assortment trade-offs that define a brand, supplier negotiations or decisions that carry reputational risk. In 2026 the realistic model is agents that prepare and execute within guardrails, with merchandisers approving and setting direction.
What is an AI agent in merchandising?
An AI agent is a system that does more than answer a question. It can watch data continuously, decide which analyses to run, combine results from several systems and propose or carry out an action, such as adjusting a replenishment quantity, within rules people have defined. The difference from a dashboard is that the agent comes to the merchandiser with a finding and a proposal, rather than waiting for someone to find the problem.
McKinsey's January 2026 article 'Merchants unleashed' illustrates this with a fictional category manager (in grocery, not fashion) whose weekly cycle of reports and reviews becomes a daily cycle of agent-prepared decisions. The firm estimates that merchants spend about 40 percent of their time on low-value tasks, that agents could automate up to 60 percent of the example merchant's manual tasks, and that merchants could reclaim up to 40 percent of their time. These are estimates from a consulting firm, not measured outcomes across the industry.
Which merchandising tasks are agents suited to?
| Task | Agent role | Human role |
|---|---|---|
| Daily sales and stock monitoring | Monitor, detect anomalies, alert | Decide on response to significant issues |
| Data cleansing and attribute checks | Find and fix inconsistencies | Set rules, review edge cases |
| Re-order and replenishment exceptions | Propose quantities within limits, execute small changes | Approve larger or unusual orders |
| Stock rebalancing between stores or channels | Propose and execute within rules | Set rules and priorities |
| Markdown and promotion scenarios | Run scenarios, estimate effects | Choose strategy, protect brand price position |
| Trend signal screening | Match external signals to own attributes and sales | Decide which trends fit the brand |
| Supplier and vendor materials | Draft performance summaries and briefs | Negotiate and manage relationships |
| Range and category strategy | Provide analysis and options | Own the decision |
Why are re-orders and replenishment the obvious starting point?
Re-order and replenishment decisions are frequent, rule-based and measurable, which suits agents. An agent can compare current sell-through with plan, check stock at warehouses and suppliers, consider lead times and propose re-orders on fast-moving styles or reallocations from slow stores. Because the outcome (stock-outs avoided, excess stock reduced) can be measured, the agent's performance can be tested against human decisions before it is given more autonomy.
For fashion, the limit is that much of the range is seasonal and cannot be re-ordered at all. Agents add most value on carry-over and replenishment styles and on reallocating existing stock, and less on one-off seasonal buys where the decision is made once, months in advance.
Trend screening is a second candidate. An agent can match external trend signals with the brand's own attribute-level sales and flag where a rising attribute is already selling well in the range, which is a more actionable output than a generic trend report.
Which tasks should stay with merchandisers?
- Category and range strategy: which customer to serve, which price architecture to hold and how a category should evolve.
- Brand-defining trade-offs: keeping a slow-selling signature style because it carries the brand, or declining a profitable trend that does not fit.
- Supplier relationships: negotiation, long-term capacity commitments and handling disputes.
- Decisions with fairness or reputational impact: pricing across markets, allocation to key accounts and responses to public criticism.
- Setting the guardrails: the limits within which agents act, and when they must escalate.
McKinsey's April 2026 article 'From dashboards to decisions' makes a similar point, quoting the view that a human in the loop is needed long before anything close to full automation, and that agents are unlikely to be good at understanding category nuance and competitive differentiation.
What has to be in place before agents can help?
- Clean, connected data. Sales, stock, orders, product attributes and supplier data must be consistent across systems; agents amplify data errors.
- Codified rules. Pricing thresholds, minimum margins, allocation priorities and re-order limits need to be written down, not held in people's heads.
- Clear roles. Decide who approves what, and how merchandiser roles change when routine work is automated.
- Training. Merchandisers need to know how to brief, check and correct agents.
- Success metrics. Define in advance what improvement looks like, such as fewer stock-outs or less time on reporting.
The organisational gap is significant. In McKinsey's January 2026 research, 71 percent of merchants reported limited or no impact from current AI tools, 61 percent said they were unprepared to scale AI across merchandising and only 24 percent said they received adequate training. In its April 2026 article, McKinsey also says its Merchant AI Accelerator has helped retailers lift sales by up to 5 percent and improve margins by up to 3 percent; these are self-reported results from engagements rather than independent benchmarks.
What does this mean for fashion merchandising teams?
The near-term change is in how time is spent rather than in headcount alone. If agents absorb monitoring, data preparation and routine re-orders, merchandisers can spend more time on range architecture, trend interpretation and supplier and account relationships. The State of Fashion 2026 report by McKinsey and The Business of Fashion lists the rewiring of the workforce around AI among the industry's main themes for the year. Teams that begin with narrow, measurable tasks, and document what agents may and may not do, are likely to learn faster than those waiting for a general-purpose merchandising agent.
Frequently asked questions
What can AI agents do in merchandising?
AI agents can monitor sales and stock, detect anomalies, clean data, run markdown and replenishment scenarios, propose re-orders and draft supplier reports. They work best on frequent, rule-based and measurable tasks, with humans approving significant decisions.
Will AI agents replace merchandisers?
Current evidence points to agents taking over routine analysis rather than replacing merchandisers. Strategy, brand-defining trade-offs, supplier negotiations and decisions with reputational impact remain human responsibilities, and consultancies stress the need for a human in the loop.
How much time can AI agents save merchandisers?
McKinsey estimated in January 2026 that merchants spend about 40 percent of their time on low-value tasks and could reclaim up to 40 percent of their time with agentic AI. These are estimates, and actual savings depend on data quality and process design.
What do you need before using AI agents for re-orders?
You need consistent sales, stock and supplier data, written rules for re-order limits and margins, clear approval roles and metrics to judge results. Agents should start in recommendation mode and gain autonomy only where they prove reliable.
One edition every weekday morning. Read in five minutes. Free for industry professionals.
