Best AI for Retail Business

Demand forecasting vs operations, where retail AI pays first

A forecast tells a retailer what to order. Operations decide whether the order is placed, confirmed and delivered on time. For most small retailers, the second is where the stock-outs come from.

By The editors · Updated · 3 min read

A buyer in a small retail back office compares last season's printed sales sheets with a supplier catalogue, a pen in hand.
A buyer's best forecast is only as good as the order that follows it.
Contents
  1. The scale of the problem
  2. What forecasting tools do
  3. What operational AI does
  4. Where small retailers usually start
  5. A plan for the next reorder cycle
  6. Questions buyers ask
  7. Sources

Retail AI is often sold as better forecasting. For a large chain, more accurate forecasts are worth a great deal. For an independent or mid-sized retailer, the bigger loss is often simpler: the order that went out late, and the delivery nobody chased.

The scale of the problem

IHL Group's research, summarised by the British Retail Consortium in October 2025, puts the cost of inventory distortion across retail at $1.7 trillion. Empty shelves are the largest driver at $690.9 billion, which the summary links to failures in replenishment systems, and supply-chain disruptions cost $301 billion a year; misalignment between buying and planning is named among the top drivers⁠1. These figures come from large retailers, but the categories apply to a single shop: the wrong quantity, the late order and the late delivery.

What forecasting tools do

ToolForecasting, as published
Lightspeed RetailInsights with forecasting and order recommendations on every plan⁠2
Inventory PlannerSKU-level demand forecasting, replenishment, multi-location and budget planning⁠3
RELEXDemand planning and sensing, automated self-correcting replenishment and allocation for enterprise retailers⁠4
OpenLazarusForecasts from sales patterns, seasonality and trends inside its inventory manager⁠5

Forecasting tools answer the question of how much to order. They are most valuable where the catalogue is large, demand is seasonal and stock is spread across locations and channels.

What operational AI does

Operations start where the forecast ends. Lightspeed's AI Doc Import turns supplier documents into draft purchase orders⁠2. OpenLazarus describes an inventory manager that drafts actual purchase orders, not alerts, taking account of lead times, minimum order quantities and supplier performance; it sends them when told to, tracks delivery dates from its own inbox and flags delays⁠5. In its e-commerce example, which the company labels as invented, the purchasing teammate raises a claim on a faulty batch and drafts purchase orders⁠6.

A quick test of where your stock-outs come from

  • Take the last 20 stock-outs on products that matter.
  • For each, note whether the order was too small, placed late, or delivered late.
  • If most were too small, improve the forecast.
  • If most were placed late or arrived late without a chase, fix the operations first.

Where small retailers usually start

For a retailer with a few hundred products and one or a few stores, the forecast inside the POS is often adequate, and the cost sits in the hours between "running low" and "order confirmed". For a multichannel seller with thousands of SKUs, a dedicated forecasting tool earns its place sooner. For a chain, enterprise replenishment is a different category of purchase altogether.

A plan for the next reorder cycle

  1. Run the stock-out test above on the last 20 cases.
  2. Measure the gap from low stock to purchase order sent, and from order to confirmed delivery date.
  3. If the gap is the problem, give AI the drafting of purchase orders and the supplier follow-up, with a person approving orders above a set value.
  4. If quantities are the problem, turn on or buy forecasting, and compare its recommendations with what was actually ordered.
  5. Review after one full cycle and keep what closed the gap.

For tool-by-tool detail, see AI in retail by job and our main guide.

Questions buyers ask

What is the difference between demand forecasting and inventory operations?

Forecasting estimates how much of each product will sell and when. Operations turn that estimate into stock on the shelf, placing purchase orders, confirming them with suppliers, tracking deliveries and receiving goods. A good forecast with slow operations still produces empty shelves.

Do small retailers need AI demand forecasting?

Many already have basic forecasting in their POS, such as Lightspeed's Insights. A dedicated tool such as Inventory Planner helps sellers with many SKUs across several channels. If stock-outs come mainly from late orders or late deliveries, AI that handles purchasing and supplier follow-up is usually the better first step.

How do I know whether my stock-outs come from forecasting or operations?

Take the last 20 stock-outs on important products and check, for each, whether the order was placed too small or too late, and whether the supplier delivered late. If most orders were placed late or deliveries slipped without anyone chasing, the bottleneck is operational.

Sources

  1. Fixing inventory distortion: who’s winning, who’s failing, what’s working. British Retail Consortium (IHL Group research), 6 Oct 2025. Accessed 8 Nov 2026.
  2. Lightspeed Retail pricing. Lightspeed. Accessed 8 Nov 2026.
  3. Inventory Planner pricing. Inventory Planner by Sage. Accessed 8 Nov 2026.
  4. RELEX Solutions. RELEX Solutions. Accessed 8 Nov 2026.
  5. AI inventory manager. OpenLazarus, 26 Dec 2025. Accessed 8 Nov 2026.
  6. AI for e-commerce. OpenLazarus. Accessed 8 Nov 2026.

Updated 8 Nov 2026: First published.

Further reading