The shift from planning tools to Agent Teams

For twenty years, retail planning software asked the same question in different ways: can we give planners a better dashboard, a cleaner workflow, a more sophisticated model? The answer, for a long time, was yes, and the industry built a dense stack of merchandise financial planning tools, assortment platforms, allocation engines, markdown optimizers, and demand forecasting systems to prove it.

What that stack never changed is the loop. A planner still pulls a report, runs an analysis, decides what to do, types the decision into another system, and waits for next week. The AI copilots layered on top in 2023 and 2024 made the report prettier and the analysis faster, but the loop stayed the same. In most retailers, that loop runs on a weekly or quarterly cadence, and the plan built in Q4 is broken by Q1.

Agentic retail merchandising is a different operating model. It inverts the loop. Instead of giving planners dashboards to act on, specialized agents own distinct decisions and run them continuously. The agent reads the signal, makes the recommendation, explains the reasoning, and lets the planner approve, adjust, or override in plain language. The resulting action is pushed into execution systems automatically. Nothing gets re-typed. Nothing waits for a weekly cadence.

What is agentic retail merchandising?

Agentic retail merchandising is an AI operating model that automates the merchandising lifecycle through specialized AI agents organized into teams. Each team owns a complete business domain (demand analytics, financial planning, assortment, in-season execution, pricing) and runs continuously against live data. Agents recommend; planners approve; execution flows into the systems of record automatically.

Three characteristics separate agentic merchandising from the AI-copilot generation of planning tools:

1. Agents own the loop, not planners. A planner approves and strategizes; agents run the forecasts, allocation passes, OTB reconciliations, and markdown decisions continuously. The loop runs every hour, not every week.
2. Decisions, not dashboards. The output of an Intelo agent is an action, approved and pushed into the OMS, ERP, or planning system. Reports are a side effect, not the product.
3. Compounding intelligence. Every planner override, every approved recommendation, every rejected scenario is a signal the system learns from. The recommendations get sharper and more tailored to the retailer the longer they are used.

Why the old model is breaking

Three forces are converging to make the planning-tool model untenable for premium retailers specifically:

1. The cadence problem

A weekly allocation run misses the accelerating hero, the slowing launch, and the stockout before the weekend. A quarterly plan broken in Q1 is debated for the rest of the year. The business runs faster than the plan: luxury houses launch capsules on social cadence, specialty retailers chase seasonal demand that peaks for days, and every retailer absorbs macro shocks (tariffs, weather, press cycles) that the quarterly model never saw.

2. The handoff tax

Most retailers stitch merchandising across four or five vendors: item planning here, clustering there, OTB reconciliation in a spreadsheet, buy planning in another system, markdown optimization in a fifth. Every handoff adds a reconciliation meeting, a version war, and a chance for the plan to drift from reality. The handoffs are where the margin leaks.

3. The headcount constraint

Merchandising teams are being asked to move faster with the same or fewer people while carrying more risk from unpredictable demand. There is no 20% more planners hiding somewhere to run the weekly cadence twice as often. The leverage has to come from a different operating model, not more capacity on the old one.

The five Agent Teams

Intelo's agentic merchandising platform is organized as five Agent Teams, each a cohesive network of specialized agents that collectively own a merchandising domain. Teams can be purchased individually and deliver standalone value, or run together as a compounding intelligence network where every team learns from the others.

Merchant Analytics: the analytical spine

Merchant Analytics forecasts demand at SKU-store-week granularity, diagnoses assortment performance, builds size curves, predicts newness performance, and quantifies lost sales from stockouts and substitution. It is the layer every other team learns from. Most Intelo customers land with Merchant Analytics or MFP first, because the forecasts and diagnostics are the foundation downstream decisions need to be smart.

MFP: the financial operating system

The MFP Agent Team builds, stress-tests, and continuously reconciles the merchandise financial plan. Four agents (Merchandise Financial Planning, Business Planning, Open-to-Buy Manager, Scenario Planning) replace the quarterly cycle with continuous reconciliation. Ask "what if we shift $8M of receipts from Q2 to Q3?" and get a full P&L impact without a two-week modeling sprint.

Assortment Planning: line list to committed buy

Assortment Planning owns the end-to-end workflow between the financial plan and the committed buy. Five agents (Item Planning, Store Clustering, Assortment Reconciliation, Cluster/Store Level Assortment, Buy Planning) replace the spreadsheet choreography from line list through purchase orders. Store clustering uses actual demand and sell-through behavior, not zip codes or manual tiers.

In-Season Management: execute the plan, course-correct automatically

In-Season Management turns the plan into decisions every day: allocation, replenishment, rebalancing, and real-time inventory optimization. In public deployments, the agents push transfer orders straight into Fluent Commerce. Psycho Bunny went live with Replenishment at the start of 2026 and has six agents live across a multi-hundred-door US and Canada footprint, positioning 1 to 5 weeks of supply in a single allocation run with zero manual re-entry.

Pricing & Promotion: lifecycle pricing

Pricing & Promotion owns markdown timing and depth, promotional design, and revenue maximization across the product lifecycle. Three agents (Markdown Optimization, Promotional Strategy, Revenue Maximization) simulate every price decision against the live financial plan before it is surfaced to a merchant. The planner sees the recommendation, the projected margin impact, and can approve by cluster or channel.

How agentic merchandising runs in practice

The operating model centers on three mechanics: planner approval in plain language, automatic execution into systems of record, and a memory layer that compounds the retailer's own history.

Approval in plain language. A planner opens the workspace and sees what the agents recommend doing next. They can accept, adjust, or override in a conversation layer. "Hold the replenishment on these 12 SKUs through the end of the month" is a sentence; the agent interprets it, applies the constraint, and re-recommends. No form fields. No workflow steps.

Execution into systems of record. Approved recommendations push straight into the retailer's execution stack: OMS (Fluent, Manhattan, Oracle), planning platforms (SAP, Oracle Retail, custom), PLM (Centric and others), and the data warehouse. Nothing gets re-typed. The audit trail follows the recommendation end-to-end.

Memory layer. Every decision and outcome is captured and fed back into the system. The agents learn what works for this retailer: which overrides signal a brand floor, which clusters respond to markdowns differently, which categories need capsule-specific logic. Recommendations get sharper every season.

Agents do not replace planners

This is the question every merchandising team asks first, and the answer is honest: no. The repetitive, spreadsheet-heavy work (allocation runs, replenishment calculations, variance analysis, markdown passes) moves to the agents. Planners move to higher-leverage work: strategy, exceptions, cluster logic, vendor conversations, and the calls that need real human judgment.

The agents surface their recommendations and wait for approval before anything moves. The merchant stays in control of what gets shipped, discounted, allocated, or moved. The team becomes smaller only if the retailer explicitly re-scopes it. In most deployments, planners redirect capacity to work that was previously impossible: more strategic clustering, deeper vendor negotiation, better launch choreography.

What customers measure

Public outcomes retailers have shared from Intelo deployments include:

• Psycho Bunny with Fluent Commerce: 1 to 5 weeks of supply positioned in a single allocation run, with zero manual re-entry. Six agents live across a multi-hundred-door US and Canada footprint. Public session at RetailClub AI Festival, Huntington Beach, September 2026.
• Shipment consolidation (Fluent Commerce partnership): Split-shipment rate reduced from around 17% to below 10%, cutting freight spend and improving customer experience.
• Scenario modeling: Scenarios that previously took a two-week modeling sprint ("what if we shift $8M of receipts across quarters") now run in minutes against the live plan.
• Time to go live: One quarter for the first Agent Team against real data. Expansion to additional teams without a new sales cycle.

How to get started

The deployment pattern is Land, Expand, Compound. Start with the Agent Team closest to your sharpest current pain. If demand accuracy and lost sales are the problem, land with Merchant Analytics. If the financial plan is the broken artifact, land with MFP. If allocation and replenishment are consuming the team, land with In-Season Management.

Prove value in a quarter against live data. Expand to adjacent teams as the intelligence network compounds. There is no requirement to buy the full suite, and there is no new sales cycle to add a team once the first is live.

Frequently asked questions