What Automation Actually Does to Your Headcount

Jingyuan Supply Chain in Xi'an cut its picking team from 20 people to 7 when it automated sortation. It also tripled its sorting efficiency, went live in three days and returned the cost of the system inside a year. That result is from a named site, and it is the number every operations manager reads first and trusts least.

Trust it exactly as far as it goes. It shows one very manual site needing fewer pickers after automation. It does not show that your site will lose a third of its team, and it does not say the 7 who stayed were doing the same jobs as before. This article covers where the hours go today, which of them a machine removes, which jobs change instead of disappear, and where people are still needed. Mushiny's shelf-to-person system is the clearest example, so it comes up often.

Where do the hours in a manual warehouse actually go?

Mushiny's shelf-to-person page puts it simply: about half of a picker's time in a manual warehouse goes on non-productive walking, not picking. Half of what you pay for is movement between the stock and the order.

The rest of the day goes on the work that touches the stock: picking, scanning, putting items into the right order, packing and clearing. In a batch-picking operation one of those tasks is the sortation step, where someone reads each item and finds its slot on a putwall. It is slow and repetitive, and a misplaced item becomes a wrong order going out the door.

Which hours does automation remove?

Two kinds, from two different machines.

Walking comes out with shelf-to-person. A robot lifts a whole shelf pod and carries it to the picker. Mushiny puts the gain at up to 4 times the picking rate against manual pick-and-walk lines per hour. That is the top of the range. The gain is largest where pickers cover long distances and smallest where the pick face is already dense. Mushiny's own line is that you do not add people, you take out the walking.

The sorting decision comes out with a Minisorter or a 3D Sorter MAX. The operator scans an item and places it on the machine, and the machine carries it to the destination for its order at 99.9 per cent sort accuracy. Nobody reads a slot. That removes the putwall job and the headcount that grew with it.

What happens to each task?

Two rows need a closer look. Clearing completed orders matters as much as sort speed, because a full destination nobody has emptied is one the next batch cannot use. Packing is the other. Mushiny puts the practical gain from the PolyPacker at triple the productivity of a manual bench, but the operator still loads every item. The machine removes the handling around the item, not the person.

Which jobs change instead of disappear?

Most of the people in that table keep a role. The role is different.

A picker in a shelf-to-person system stands at a station and picks from a pod that arrives. Knowing the aisles matters less. Speed and accuracy at the station matter more.

A putwall worker becomes an inducer. Mushiny describes the Minisorter induction job as the whole operator job. It is smaller than the one it replaces, which is why the headcount falls.

A packer becomes a machine operator. The PolyPacker is one machine, one operator and one press.

Supervisors get new work. Someone has to watch the destinations fill, confirm orders are being cleared and deal with the items the machine rejects. A smaller floor team still needs people who understand the process.

Where are people still needed?

At the stations. A robot does not pick. If picking labour is your constraint, shelf-to-person changes how much each picker gets done. It does not remove the picker.

At induction. A sorter is only as fast as it is fed. The Minisorter runs two induction ports and a 3D Sorter MAX runs up to 10, and each port needs someone standing at it. Under-staff induction and the machine runs below its rate.

At the end of the flow. Orders still have to be cleared, packed and handed to the carrier, and the machine delivers them faster than a manual process did.

At the edges. A Minisorter takes items from 50 x 50 x 5 mm up to 750 x 500 x 250 mm depending on size class, and a 3D Sorter MAX takes up to 450 x 320 x 320 mm. Anything outside stays on a manual line. If the oversize end of your range carries a large share of your volume, a large share of your labour stays where it is.

What does each of those positions cost?

The Australian benchmark is the Storage Services and Wholesale Award, MA000084. From 1 July 2026 Storeworkers are paid $27.08 an hour at grade 1, rising to $29.61 at grade 4, before on-costs. One position at the grade 1 rate across a 38 hour week is roughly $53,500 a year before superannuation, leave loading, penalty rates or agency margin.

That tells you what a position costs. It does not tell you how many you will stop needing, and nobody can say without your order data: lines per order, shared stock, the share of volume that fits the machine and the hours it will run. Mushiny's public position is that after analysis and sizing, goods-to-person and sortation systems typically land under a three year return.

What happens to peak casuals?

If your site brings in extra people over peak, the putwall is the job that scales most easily by adding bodies, so it is often where casual staff end up.

Mushiny's position on the Minisorter is that the induction job is simple enough that casual staff over peak are not a risk to accuracy. A sorter also holds 99.9 per cent however far into the shift it is, which a manual putwall does not. The number of casuals you need can fall, because the machine does not scale with headcount the way a putwall does.

Peak does not disappear, though. A machine has a rate. If peak orders exceed it you still need a manual route for the excess, so decide in advance what happens above the rate. Capacity can be added later: GPA Logistics in the USA doubled its sorter from 80 destinations to 160 after it was running.

What does a smaller team mean for training and cover?

Training gets shorter for some roles and longer for others. A new starter is productive on a goods-to-person station in about half a day. Reading why a destination is blocked or why a robot has stopped takes longer. That points to a small number of people who need proper training instead of a large number who need a little. Ask us for a training plan in writing as part of the quote.

Cover matters more. In a team of 20, one absence is a small gap. In a team of 7 it is a visible one. Train more people on each station than you strictly need, so one absence does not stop a line.

The system needs cover too. Mushiny's shelf-to-person robots work as a swarm with no conveyor, so if one drops out the others continue, and a 24/7 service hotline backs the site. That reduces the risk from equipment, not from people. Every shift still needs someone who knows what to do when something stops.

Will your people be redeployed or let go?

The Jingyuan figures do not say, and Mushiny's public material does not record what happened to the 13 who left the picking team.

What the evidence supports is narrower. Automation reduces the labour needed for walking and sorting. For a growing site, volume usually rises without the picking team rising with it. For a flat site, the same machine means fewer positions. Which one you are depends on your volume plan, your attrition and what else you need people to do. If you lose people each year anyway, a smaller requirement may be met by not backfilling. If volume is growing, it may go through the same team. Settle that before the project starts, not after go-live.

How do you find out for your own site?

Walk the floor for one shift with a stopwatch and record how much of each role's time goes on walking, sorting, waiting and packing. If pickers are walking, shelf-to-person is a candidate. If they are waiting at pack, it is not.

Then bring your lines per order, SKU range, peak picks per hour and labour cost to the Chatswood Tech Centre. It is the only place in Australia with Mushiny's full range running live under one roof, and it is free to visit. The reference sites page shows how other sites run.

Frequently asked questions