For all the technology in a modern warehouse, there is one part of the operation that can still look surprisingly manual: Deciding how much labour is needed, when, and where.
Most large operations have systems for forecasting demand, managing warehouse activity and scheduling people. They have dashboards and more operational data than they have ever had. Yet the actual workforce planning decision – how much labour will we need, where should we put it, what will it cost and what risk are we willing to carry – is often still assembled in spreadsheets and worked through by a relatively small number of experienced people.
There is a reason spreadsheets have survived. The decision they are being used to make is harder than it first appears.
A demand forecast is not a workforce plan. Nor is a schedule. Between the two sits a series of choices about workload, productivity, available labour, skills, agency, overtime, service levels and cost. Those factors are usually held in different systems and owned by different parts of the business, which means the person building the plan has to bring them together somehow. In many operations, Excel has become the place where that happens.
The challenge becomes more obvious when reality starts changing.
A promotion performs better than expected. Absence is higher. Productivity falls in one area. Work arrives in a different mix. Suddenly the plan created yesterday no longer works, and someone has to decide what to change.
Experienced planners are often very good at this. They have learned which assumptions to trust, where the operation has some flexibility and where a seemingly small change is likely to create a problem later in the shift. But there is a limit to what even the best planner can explore manually. Faced with a changing set of constraints, they may be able to compare a handful of plausible options; there could in reality be thousands.
That is where AI can change workforce planning in a useful way.
From building a few plans to evaluating thousands
The opportunity is not simply to generate another forecast or automate the spreadsheet. It is to evaluate many more possible labour plans than a person could realistically model, test them against the constraints of the operation and recommend the ones that produce the best balance of cost, capacity and service.
One option may protect service but require more overtime. Another may reduce agency spend while accepting a little more operational risk. A third may move labour between activities or make a different productivity assumption. None of those choices can be understood properly by looking at labour cost alone.
Our approach to Labour Decision Intelligence is designed around that point: As conditions change, the system can evaluate thousands of possibilities, recommend an optimised labour plan and then pass an approved commitment into the systems that execute it.
That is a meaningful step forward from spreadsheet-based workforce planning. But I think there is another capability that may prove just as important.
The business needs to remember why it decided
A recommendation will not always be accepted exactly as it is produced, and that is not necessarily a problem.
A planner may know something the model does not. There may be a local constraint that has not yet made it into the data, an unusual customer requirement or a good operational reason to carry more contingency than the optimisation recommends.
What matters is whether that judgement disappears once the decision has been made.
Today, much of that knowledge is informal. It sits with experienced planners as heuristics, preferences, rules of thumb and an understanding of how a particular site behaves. Conventional enterprise systems often capture the operational data around the decision without capturing that knowledge itself.
A better workforce planning system should therefore record more than the final number of people required. It should be possible to see what was recommended, which assumptions were used, what the planner changed, why they changed it and what happened afterwards.
Over time, that creates a learning loop. Plans, human interventions, actual outcomes and operational variance can all become evidence for the next decision rather than disappearing into another spreadsheet version or remaining in somebody’s head.
That may be one of the most important differences between workforce planning software and genuine decision intelligence.
The aim is not to eliminate experienced judgement. It is to give that judgement better options to work with – and then make sure the organisation learns from it.
Because the real limitation of spreadsheet-based workforce planning is not simply that spreadsheets are manual.
It is that, once the decision has been made, the knowledge of what made it a good decision is lost.