AI in warehouse picking and putaway, starting with the walking

6 days ago
3 min read

Updated: 4 days ago

Introduction

In a small warehouse, most of the labour is walking. Picking a ten-line order can involve several hundred metres of travel, and the difference between a good pick path and a bad one is frequently half the time. The layout that produced the bad path was usually set when the unit was first occupied and has never been revisited.

This is a well-understood problem with a cheap solution, and it is a good example of a case where the analysis matters more than any technology. Knowing which items are picked together, and how often, tells you where to put them, and that knowledge is sitting in your order history.

1. AI in warehouse picking and putaway starts with slotting, not equipment

Layout before automation.

Placing the fastest-moving items nearest the packing area, at accessible heights, removes more time than any device. This is a data exercise on your own order lines and it costs nothing but a weekend of moving stock.

2. Rank items by pick frequency, not by value

The relevant measure.

An inexpensive item picked forty times a day deserves the best location. A valuable item picked monthly does not. Firms slot by product family or by supplier, which is convenient and wrong.

3. Put items that are ordered together near each other

The second largest gain.

Analyse which lines appear on the same orders and cluster them. This is genuinely analytical work across thousands of order lines and it produces layout decisions nobody would reach by intuition.

4. Sequence the pick path deliberately

One pass, not several.

A pick list ordered by location rather than by product code turns a route that doubles back into a single sweep. This is the cheapest improvement available in most operations and it requires only that the list is sorted.

5. Batch small orders into one route

Where volumes justify it.

Picking five small orders in a single pass and sorting them at packing removes four journeys. This adds a sorting step and is worth it above a certain order volume, which is a calculation rather than a judgement.

6. Put away with pick location in mind

Putaway decides picking cost.

Stock placed wherever there was space creates a permanent picking penalty. Directing putaway to the right slot, or at least to a sensible zone, is where the layout discipline is actually enforced.

7. Fix your location accuracy first

Nothing works without it.

If the system says an item is in a location and it is not, every optimisation fails and pickers revert to searching. Cycle counting and a rule that stock is never moved without a system record are the precondition.

8. Measure lines picked per hour and pick accuracy together

Both, always.

Speed alone rewards shortcuts that produce wrong shipments, which cost far more than the time saved. The two numbers reported together are what tell you the operation is genuinely improving.

9. Consider whether the problem is the range

Sometimes the answer is fewer lines.

A warehouse struggling with picking frequently has hundreds of slow-moving lines occupying the best space. Rationalising the range is often more effective than optimising around it, and it releases capital too.

Be careful about introducing devices before the layout is right. A scanner or a voice system in a badly slotted warehouse makes an inefficient process faster to execute rather than making it a good process.

Conclusion

Attack the walking first, because it is most of the labour in a small warehouse.

Rank items by pick frequency rather than value and slot the fast movers nearest packing, cluster the lines that appear on the same orders, sort every pick list by location so the route is a single sweep, batch small orders where the volume justifies it, direct putaway rather than allowing stock to go wherever there is space, fix location accuracy before optimising anything, report lines per hour alongside pick accuracy, and consider rationalising the range instead of optimising around it.