A voice picking warehouse system (pick-by-voice) is one in which pickers receive instructions through a headset and confirm each step by speaking. The picker hears the location, product and quantity, then reads out the check digits and quantity picked. With hands and eyes kept on the job, the aim is to cut screen time and the risk of error compared with paper lists and handheld terminals.
Why picking is so error-prone
In many warehouses, order picking accounts for a large share of the labour and of the mistakes. A picker moves between shelf and carton with a paper list or handheld terminal in hand; on every line they look at the screen, read the location, find the item, scan the barcode and key in the quantity. At peak times it is easy to reach into the wrong slot, confuse similar packaging or short-pick without noticing. We cover handheld terminals on the counting side in handheld terminal stock counting.
The error surfaces after dispatch: the customer reports a missing or wrong item, a return and re-shipment follow, and the stock record no longer matches reality. With date-coded goods, picking the wrong batch is a further risk; the rules are covered in FIFO and FEFO inventory methods. Voice picking aims to cut the picker's need to look at a screen and type data in by hand. Beyond picking, other AI uses in warehousing and dispatch are gathered in AI in logistics and warehousing.
The cost of doing nothing: bad data grows along the supply chain
Data errors are more common in manual processes than most managers assume. In a 2018 retail supply chain study by the Auburn University RFID Lab and GS1 US, 69% of orders shipped from brands to retailers contained data errors when RFID was not used. The study measured the effect of RFID, but the lesson is general: processes that rely on human eyes and manual entry are prone to error.
The total cost is not trivial. Data quality expert Thomas C. Redman, writing in MIT Sloan Management Review in 2017, estimated the cost of bad data at 15% to 25% of revenue for most companies. In a warehouse that cost shows up as returns, excess stock, count discrepancies and lost customers.
In many Turkish businesses the root problem is missing systems:
According to TurkStat, only 28.3% of enterprises in Türkiye with 10+ employees used ERP software in 2025. (TurkStat ICT Usage in Enterprises Survey 2025)
Voice picking does not work without a warehouse management system (WMS), or at least software that holds location and stock data. So the first question is not "should we use voice?" but "can we trust our stock and location data?" We cover the foundations in our warehouse management system guide.
How a voice picking warehouse system works
A typical pick line runs as a short dialogue:
- The system gives the location: "Aisle 12, slot 04."
- The picker reads the check digits: they say the two- or three-digit check number on the slot label ("37"). If it is wrong, the system warns them, catching a wrong-slot pick on the spot.
- The system gives product and quantity: "Pick three."
- The picker confirms the quantity: "Three." If stock is short they say "two" and the system records the shortfall.
- The system moves to the next line. As soon as the line is complete, stock and order status update in the WMS.
On the hardware side, the picker wears a headset with microphone and a belt-mounted or wearable terminal. On the software side, speech recognition is tuned to recognise a limited command set (digits, "repeat", "short", "damaged" and so on) quickly and reliably in a noisy environment. That is a different job from transcribing free speech; the wider picture is in Turkish speech to text for business.
Voice picking vs handheld terminals vs pick-to-light
| Method | Strength | Weakness | Best suited to |
|---|---|---|---|
| Paper list | No investment | Error-prone, no real-time stock view | Very low volumes |
| Handheld terminal (barcode) | Every item scanned, clear record | One hand and both eyes tied to the screen | General-purpose warehouses |
| Voice picking | Hands and eyes free, check-digit verification | Noise and language tuning, needs a WMS | High line counts, case and crate picking, cold rooms |
| Pick-to-light | Very fast at fixed stations | Hardware in every slot, low flexibility | Small parts, fast-moving zones |
One method is rarely enough. Where serial numbers or batch tracking are required, a barcode scan step can be added to the voice flow. In cold chain warehouses, where staff wear gloves and screens are awkward to read, the advantage of voice becomes much clearer.
Pilot and roll-out steps
- Measure where you are. Lines picked per hour, mis-pick rate, time for new staff to get up to speed. Without a baseline, no improvement can be shown.
- Clean the data. Locations, check digits and item records must be complete; weed out duplicates and errors through a data quality exercise.
- Test noise and the command set. Try recognition with real staff in areas with forklifts, compressors and conveyors. Commands should be short words that do not sound alike.
- Pilot in one zone. Start with a group of aisles and one shift, and compare against a zone still using handheld terminals over the same period.
- Connect the systems. Data flows between voice terminals, WMS and ERP are built through API integration; decide in advance how terminals behave if the connection drops.
- Bring staff with you. Explain clearly whether voice is recorded and which performance data is kept. Under KVKK, Türkiye's data protection law, employees must be given a privacy notice, and the data collected should be limited to what the purpose requires.
Which KPIs should a pilot track?
A pilot is judged not on a single figure but on several indicators read together:
- Mis-pick rate. Wrong item, wrong quantity or wrong batch, counted alongside customer returns and pre-dispatch check findings.
- Lines picked per hour. The voice zone is compared with the handheld zone over the same period and a similar order profile.
- Time to full speed. How many shifts does a new starter need to reach full pace? This matters most for warehouses that take on seasonal staff.
- Recognition issues. How often the system asks for a repeat and where it struggles; use this to fix noise sources or command choices.
- Staff feedback. Headset comfort, battery life and how natural the flow feels. A system that goes unused on the floor is not a success, whatever the numbers say.
Write these indicators down before the pilot starts and compare the results with thresholds agreed in advance. If the thresholds are missed, narrow the scope or change the method; that decision is as valuable as the pilot itself.
How we do this at Digital Bridge
- Discovery and site assessment. We visit your warehouse and map the picking process, location structure, noise levels and current software. We tell you honestly whether voice picking makes sense at your volumes, and set out scope, phases and cost in a written proposal.
- WMS foundation. If location and stock tracking are missing, we first implement warehouse and inventory management (WMS) software, or prepare your existing system for a voice flow.
- Voice layer. On the voice recognition side, we tune the command set to Turkish and to your warehouse's noise conditions, then measure it in the pilot zone.
- Hardware. For points where barcode scanning or a handheld is still needed alongside voice, we design and build industrial data terminals in-house. Software and hardware come from the same team, so responsibility stays in one place.
- Sector scenarios. We build the dispatch, batch and cold-chain requirements of logistics and transport and food and beverage warehouses into the flow.
Warehouse security and shifts: SmartPass
To read picking productivity properly, you need to know who actually worked each shift, and warehouse entry control can run on the same card infrastructure. Our own product, SmartPass, combines card access control, time and attendance and canteen meal counting in a single system: one card read opens the door and creates the attendance record. The warehouse team belongs to an access group that is linked to the Warehouse zone, optionally with a weekly time profile, and visitors receive a temporary card at the entrance.
In SmartPass, timesheets are calculated against the shift plan; a closed month is locked and a monthly per-person summary is exported to the ERP. That lets you set the line data from voice picking beside the staff who actually worked each shift. Warehouse entry control is covered in detail in warehouse and logistics access control.
Next step
For one week, log picking errors and lines picked per hour. Then contact us and we will review your warehouse together and choose the best zone for a pilot. For other operational uses of AI, read where to start with AI in business and browse our Artificial Intelligence hub.