Many automation projects start with a product demo. A supplier shows an autonomous mobile robot, a goods-to-person system or a high-density storage solution. Everyone’s impressed. Then the conversation turns to where it might fit in the warehouse.
That’s backwards.
The first question should be: what’s actually stopping this operation from performing? Maybe pickers spend too much of their shift walking. Maybe you’ve run out of pallet positions, despatch is congested, or nobody fully trusts the inventory count. Each of those needs a different fix.
A Malaysian FMCG distributor handling mostly full pallets has a different case from an Indonesian ecommerce operation processing small, volatile orders. Putting the same technology into both is expensive guesswork.
So we like to pin down the operational change you need before anyone talks equipment. Fewer picker kilometres. More orders through the same despatch doors. Higher storage density without losing accessibility. Those are outcomes you can measure, and they’re what an investment decision should rest on.
Automation doesn’t always mean a new building. Modular systems can increase capacity inside an existing site, while avoiding the cost and service risk of a full relocation.
That makes retrofit attractive, particularly where the current location still suits customers and transport routes. But an old building doesn’t become suitable just because the equipment is modular.
Check the physical constraints early, because they have a habit of surfacing late. Clear height, floor flatness, slab loading, column spacing and despatch access can all restrict the design. Fire protection, emergency egress and power capacity may also need work. Any one of these can move the project cost a long way.
The operating profile matters just as much. Dense storage can release valuable space, but it may reduce direct access to individual pallets. For some inventory profiles, that’s a fair trade. For others, it’s a real problem.
Phasing is often the real advantage of retrofit. An operator can automate one zone, prove the process and expand after demand materialises. If the business case only works when the whole system is running flat out from day one, look at it again, harder.
AI can act as an orchestration layer across orders, workstations, robots and conventional equipment. In plain terms, it helps decide which task should happen next and which resource should perform it.
That can be useful where order priorities change frequently or several automated systems share the same workflow. The goal isn’t a fancier dashboard. It’s less idle time, fewer blocked stations and better flow through the building.
The dependency is data quality. If inventory locations are wrong, product dimensions are incomplete or order statuses arrive late, AI will make decisions using faulty inputs. It just makes bad decisions faster. It won’t repair the underlying process.
Before adding AI, test whether the warehouse management system already produces reliable inventory, task and order data. If your supervisors still run the real plan off a spreadsheet, there’s groundwork to do first.
Matrix storage and other dense systems reduce the space given to conventional aisles. They can suit sites where property costs are high, expansion is difficult or storage capacity is the main constraint.
The assessment should cover more than storage positions. Look at retrieval sequencing, replenishment, peak throughput and access to slow-moving stock. A dense system that hums along at average volume can start queuing the moment replenishment and customer orders compete for the same equipment.
There’s also a resilience question. When a forklift breaks down in a conventional warehouse, you grab another one. A specialised storage system may depend on a smaller number of critical components. Spares, maintenance response and manual recovery procedures belong in the design, not in a later commissioning checklist.
Humanoids and mobile manipulators could eventually handle tasks in buildings designed around people. That would reduce the need to rebuild the environment around fixed automation.
For now, treat them as something to test, not something to plan capacity around. A controlled pilot can establish whether the technology handles real packaging, damaged cartons, shifting priorities and human traffic.
Don’t let the core business case depend on performance nobody has seen in your own operation.
Labour savings are where most business cases get optimistic. Automation may remove tasks without removing an equivalent number of paid positions, particularly where staffing is needed across several shifts or work areas.
Calculate the paid hours that will actually disappear or be avoided. Include overtime, temporary labour and recruitment pressure where relevant. Then confirm whether supervisors and support roles will still be required.
The cost side needs the same discipline. Software integration, controls, cybersecurity, maintenance, spare parts, training and periodic upgrades all affect payback. So does the extra inventory or outsourced capacity needed during implementation.
Model at least three operating conditions:
The third case is the one people skip. A system that meets service targets only when every asset is available may leave the operation exposed during faults or maintenance.
The useful measure isn’t cost per order at full capacity. It’s whether service remains affordable and recoverable when conditions are less favourable.
Regional forecasts are helpful, but automation economics are always site-specific. Malaysia and Indonesia show how different the brief can look from one market to the next.
In Malaysia, an ecommerce warehouse in Kuala Lumpur preparing for a mega-sale peak may prioritise flexible picking and packing capacity. An FMCG operation with a seasonal full-pallet peak may have a stronger case for improving pallet flow, replenishment and despatch control.
In Indonesia, network geography can change the design brief. A facility serving customers across several islands may need to protect despatch reliability against variable line-haul connections. Faster picking won’t solve that.
Ask about local technical support before you sign, not after. Ask who will diagnose a controls fault, where critical spares will be held and how quickly a technician can reach the site. A cheaper equipment price gets expensive fast if every fault means waiting for a specialist to fly in.
Safety must be assessed with the operating layout, not just the machine specification. Local safety, fire and building requirements still need separate review.
Before you ask suppliers for proposals, start with a short operational diagnosis.
Confirm the constraint. Measure travel, queues, storage use, labour hours and despatch performance. Separate the primary bottleneck from symptoms elsewhere in the building.
Stabilise the process. Correct location control, master data and basic slotting problems. Automate an unstable process and you usually just get a more expensive unstable process.
Define the operating envelope. Document SKU dimensions, order profiles, peak periods, growth assumptions and service cut-offs. Suppliers need this information to design comparable solutions.
Test failure and recovery. Establish what happens when a robot, conveyor, interface or network connection fails. Include manual workarounds and the labour required to sustain them.
Compare more than one level of automation. Sometimes better layout, slotting and conventional equipment provide enough capacity at lower risk. Keep that option on the table.
The decision in front of you isn’t which robot to buy. It’s which operating constraint justifies capital, and how much automation the site can absorb without creating a new dependency.
If automation is on the table, that’s exactly the question a Warehousing Health Check is built to answer, before you commit to equipment or a new building.
Talk to LCA Savills. We have been designing, justifying and delivering logistics facilities across Asia since 1999, as operators, not report writers. We look at the operation, the building and the commercial case together, so the automation decision is based on how the facility will actually run.
Start with a Warehousing Health Check at lca.asia.
The first step is to identify the operational constraint that automation must remove. Measure picker travel, labour hours, storage utilisation, workstation queues and despatch performance. Then establish whether layout, process or inventory changes could solve the problem at lower cost. Equipment selection comes after that, not before. Starting with a preferred robot or storage system can produce an investment that increases capacity in one area while leaving the real bottleneck untouched.
Payback should compare total project and operating costs with savings that the business can actually realise. Labour savings should use paid hours removed or avoided, not a theoretical reduction in activity. Include integration, software, maintenance, spares, training, utilities and implementation disruption. Test the calculation at normal and peak volumes, and with part of the system unavailable. A business case that depends on immediate full utilisation carries more risk than one that works under phased demand.
AI is not necessary for every automated warehouse. Conventional warehouse management and control systems can run many operations effectively. AI becomes more relevant when numerous orders, workstations and mobile resources require continuous reprioritisation. Its performance depends on accurate inventory, product and order data. First, confirm that existing systems provide reliable inputs and that the operation has a coordination problem AI can measurably improve.
Retrofit automation is often preferable when the existing location still supports customer service and transport requirements. It can reduce relocation risk and allow phased investment. But the building still has to support the proposed system. Clear height, floor condition, structural loading, fire protection, power and despatch access can all affect feasibility. A retrofit is not automatically cheaper once building modifications, operational disruption and future expansion limits are included.
Humanoid robots may eventually handle tasks in facilities designed for people, but operators should validate them carefully before relying on them for capacity. A useful pilot should test real packaging, damaged items, variable workflows, human interaction and fault recovery. The trial should also measure intervention rates and support requirements. Humanoids can be assessed as an emerging option without making them essential to the current warehouse investment case.
