‍The case for demand-driven, AI-powered staff allocation in terminal operations

Every morning at airports around the world, a supervisor opens a spreadsheet. They check the roster - who is in, who has called in sick, what tasks need covering - and begin manually distributing staff across the dozens of roles that keep a terminal running: customer assistance posts, trolley collection, taxi and bus rank support, check-in floor coverage, and break rotations timed to comply with regulatory requirements.

It is a process that works. Experienced supervisors do it competently, often daily, and with a level of contextual knowledge no system has yet fully replicated. But it is also a process that was designed for a simpler operating environment - one where flight volumes were lower, passenger expectations were different, and the volume of variables a planner needed to hold in their head was manageable.

That environment no longer exists. And the gap between what Excel-based manual planning can deliver and what modern terminal operations actually require has quietly become one of the most significant sources of operational inefficiency and service risk in the industry.

What Manual Workforce Planning Actually Involves

To understand why the manual approach has reached its limits, it helps to be precise about what it requires. A daily task allocation plan for a Terminal Service Delivery team is not a simple rota - it is a multi-constraint optimisation problem, assembled manually, and reassembled multiple times before and during a shift.

A supervisor preparing the plan must simultaneously account for:

1. Staff availability. Who is rostered? Who has confirmed? Who typically calls in on short notice? The roster, often exported manually from a Time and Attendance system as a CSV file, is a starting point - not a reliable picture of who will actually arrive.

2. Task coverage requirements. Which posts are mandatory and which are desirable? What is the minimum staffing level for the CBP pre-clearance facility this morning? How many trolley collectors are needed across the terminal at peak departure time?

3. Passenger demand patterns. How many passengers are expected, when, and through which parts of the terminal? The morning transatlantic bank has different staffing implications from the early afternoon European departures - and both differ from a day with a special event or a delayed peak.

4. Break compliance. Breaks must be scheduled within regulatory, contractual, and operational constraints. Inserting them into an already complex plan without creating coverage gaps requires careful sequencing - done manually, for every member of staff, every day.

5. Last-minute changes. Absences notified on the morning of the shift, delayed flights shifting peak demand timing, equipment failures requiring additional coverage - each requires the plan to be revised and re-communicated, typically verbally or via informal messaging.

This is a genuinely demanding cognitive task. The fact that experienced supervisors do it, often under time pressure and with limited tooling, is a testament to their expertise. It is also precisely the kind of task where the gap between what is achievable manually and what data-driven optimisation can produce is largest.

The Specific Failure Modes of the Spreadsheet Model

The limitations of manual, spreadsheet-based workforce planning in airport terminal operations are structural, not incidental. They do not stem from poor planning - they stem from the fundamental constraints of the tool being asked to do work it was not designed for.

Static Plans in a Dynamic Environment

A spreadsheet plan is a snapshot. It reflects the world as it was when it was built - typically the evening before, or the morning of the shift. By the time the first departure bank begins, the world has moved. A staff member has called in sick. A flight is delayed by 90 minutes, pushing its passenger wave back into a period that was planned as low-demand. A special handling request has come in that requires redeployment from another area.

None of these changes are exceptional. They are the normal rhythm of airport operations. But each one requires the plan to be manually revised - a process that takes time, introduces error risk, and draws the supervisor's attention away from the operational environment they should be managing.

Demand Disconnect

The most consequential limitation of manual planning is the difficulty of truly aligning staffing to real-time passenger demand. Experienced supervisors develop strong intuitions about demand patterns over time - they know which flight banks drive trolley demand, which routes generate customer assistance requests, which days require additional presence at the bus and taxi ranks.

But intuition, however well-developed, is not data. Actual passenger volumes vary day-to-day based on booking patterns, flight load factors, connection traffic, and external events. A plan built on historical experience is a reasonable approximation - but it is an approximation, and the gap between approximation and actuality is where both overstaffing costs and service failures accumulate.

Matching staffing to demand is not a planning problem that can be solved once. It is a continuous optimisation challenge - one that requires live data inputs and the computational capacity to process them.

The Invisible Variance Problem

Perhaps the most operationally consequential failure mode of manual planning is one that almost no airport currently measures: the gap between the plan and what actually happened.

When a supervisor modifies the daily allocation plan during a shift - covering an absence, redeploying staff to a developing queue, adjusting break timing in response to a delay - those changes are typically made verbally or informally. They are not recorded against the original plan. At the end of the shift, the official record shows what was planned, not what was executed.

This means there is no systematic data on variance from plan. Patterns that would reveal structural staffing mismatches - recurring shortfalls at specific tasks during specific time windows, chronic overstaffing in low-demand periods, regular last-minute deployments to the same locations - remain invisible. There is nothing to analyse, no baseline from which to improve, and no mechanism for turning operational experience into organisational learning.

Why Passenger Flow Is the Variable That Changes Everything

The central driver of terminal workforce demand is not the roster - it is passenger flow. And passenger flow in a modern airport is not a fixed curve that supervisors can memorise. It is a dynamically varying pattern shaped by flight schedules, load factors, connection traffic, seasonal events, weather-driven delays, and real-time operational conditions.

Consider a concrete example. A terminal operating morning transatlantic departures will see a very different staffing requirement across Customer Assistance, check-in floor support, and the US Customs and Border Protection pre-clearance facility than the same terminal handling a mid-morning European bank. Both differ again from the demand pattern generated by a large sporting event producing hundreds of passengers with the same destination and a narrow departure window.

A manual planner can account for these differences in broad terms. What they cannot do is model the precise hour-by-hour demand curve for each task type, across each terminal zone, for each specific day - incorporating the actual flight schedule, confirmed load factors, and real-time adjustments as they emerge. That requires data integration, computational modelling, and continuous updating that no spreadsheet can provide.

30–60 minutes  lead time that AI-driven passenger flow forecasting can provide airports before congestion forms - enabling proactive, not reactive, staffing decisions (Copenhagen Optimisation, 2025)

That window - 30 to 60 minutes of forward visibility - is the difference between deploying staff before a queue builds and responding to one after it has formed. It is also the difference between a passenger experience that feels controlled and one that feels chaotic.

The Compliance Dimension

Terminal workforce planning is not only an operational challenge - it is a compliance one. Break scheduling must comply with regulatory working time requirements, contractual entitlements, and operational coverage minimums simultaneously. In a manual planning environment, this compliance is managed by the supervisor's knowledge of the rules, applied individually to each team member, each day.

The risk is twofold. First, in a complex, high-pressure planning session, compliance errors occur. A break window is miscalculated, a contractual entitlement is inadvertently missed, coverage drops below minimum at a critical point. Second, when changes are made during the shift - the most common scenario - the original compliance logic is disrupted and manually re-applied, with all the error risk that entails.

An automated planning system encodes compliance rules explicitly. They are applied consistently, automatically, and visibly - every time a plan is generated or revised. The supervisor can see at a glance that all breaks are compliant and that coverage minimums are maintained. The cognitive burden of compliance management is removed from an already demanding planning task.

What Demand-Driven AI Planning Delivers

The shift from manual, spreadsheet-based workforce planning to AI-driven, demand-aligned allocation is not primarily a technology change. It is an operational model change - from reactive scheduling to proactive, continuously updated deployment.

In practice, a mature AI workforce planning system for terminal operations delivers across four dimensions:

Optimised Plan Generation

The system ingests the day's roster data - directly from the Time and Attendance system, eliminating the manual CSV export step - alongside the flight schedule, passenger forecasts, and any known constraints or special requirements. It generates a recommended daily task allocation plan that balances demand coverage, task prioritisation, compliance requirements, and staff availability. What currently takes a supervisor 30 to 60 minutes of manual work is produced in seconds, as a validated starting point for their review.

Continuous Demand Alignment

As the day progresses, the system updates. A delayed flight pushes its passenger wave back two hours - the system identifies the coverage implication and flags it. A late absence notification creates a gap in the early morning trolley team - the system recommends which other tasks can be de-prioritised to cover it, based on live demand data. The supervisor makes decisions; the system ensures those decisions are informed by current data rather than this morning's plan.

Priority-Weighted Shortage Management

When staff shortages occur - as they will, routinely - the system applies a defined task prioritisation framework to recommend how available staff should be deployed. Mandatory posts are protected first; desirable posts are reduced or temporarily vacated based on live demand. Supervisors retain the authority to override these recommendations, but they do so with a clear picture of the trade-offs involved.

~20%  reduction in overtime requirements achieved through AI-driven workload balancing at London Heathrow, which also reduced team utilisation rate variance from approximately 35% to under 14% (Research, 2025)

Actuals Capture and Performance Intelligence

Critically, AI-driven systems close the loop that manual planning leaves open. Changes made during the shift are recorded against the plan - creating, for the first time, a reliable Actual vs. Plan record. Over time, this data reveals patterns: which tasks are consistently under-resourced, which demand peaks are systematically underestimated, which operational events reliably require contingency staffing. That intelligence feeds back into planning quality, creating a continuous improvement cycle that no manual process can replicate.

What the Wider Industry Is Doing

Airport operators globally are accelerating investment in AI-driven workforce management, and the direction is consistent. The global airport operations market reached USD 7.9 billion in 2026 and is projected to grow at over 21% CAGR through to 2035 - driven in large part by the recognised cost of operational inefficiency and the demonstrated ROI of data-driven optimisation.

Amadeus reports that 94% of airport operators increased IT budgets in 2025, with operational resilience and digital passenger experience cited as the primary drivers. For front-line terminal operations, that investment priority translates directly into workforce planning and demand management tools.

At Heathrow, AI investment across operations has been linked to measurable performance improvements - including British Airways recording 86% on-time departure performance from Heathrow in Q1 2025, its best result on record, following a £100 million investment in operational resilience technology that included AI-driven decision support across ground operations.

Smaller and mid-sized airports are following. The emergence of cloud-based workforce management platforms designed specifically for aviation environments has lowered both the implementation cost and the technical barrier to adoption. The case for change is no longer restricted to hub airports with large IT functions - it is accessible to any airport operating at sufficient scale to have a material workforce planning burden.

The question for terminal operations leaders is not whether AI-driven workforce planning is better than the spreadsheet model. The evidence for that is clear. The question is how quickly the organisation can build the data foundations to deploy it effectively.

What Good Implementation Looks Like

For airport operators considering the transition, a few implementation principles consistently differentiate successful deployments from those that stall:

1. Start with data quality. An AI workforce planning system is only as good as the data it ingests. Rostering data, passenger forecasts, task definitions, and compliance rules must be clean, current, and systematically maintained. Investing in data infrastructure before or alongside the planning tool is not optional.

2. Encode operational knowledge explicitly. The constraint sets and priority rules that experienced supervisors apply intuitively must be formalised and built into the system. This process - working with operational teams to articulate what they currently do instinctively - is itself valuable, independent of the technology.

3. Design for supervisor adoption, not supervisor replacement. The most effective implementations position the AI as a planning assistant that supervisors validate and adjust, not a system that generates plans they are expected to follow unquestioningly. Trust in the system's recommendations builds gradually, through demonstrated accuracy.

4. Build Actuals capture from day one. The performance intelligence value of the system depends entirely on capturing what actually happened during each shift. This capability should be designed in at the outset, not retrofitted after implementation.

5. Plan for extensibility. Terminal Service Delivery is typically the first workforce planning use case - but the same core approach applies to car park operations, retail staff allocation, security deployment, and cleaning teams. Architectures designed for a single use case create rework costs when the scope expands.

The Time for Proof of Concept Has Passed

Excel-based workforce planning in airport terminal operations is not wrong. It is simply insufficient - for the volume of variables it needs to process, the speed at which those variables change, the compliance rigour it needs to maintain, and the performance intelligence it needs to generate.

The technology to do better exists, has been deployed at scale, and is producing measurable results. The airports that have invested in demand-driven, AI-assisted workforce planning are not running pilots - they are running operations, with quantifiably better outcomes, from the same staff base.

For airport operators still building daily task allocation plans in spreadsheets, the path forward is clear. The first step is not selecting a tool - it is quantifying the true cost of the current approach: the planning hours, the unrecorded variances, the service failures during demand peaks, the compliance risk, and the institutional knowledge that walks out the door when an experienced supervisor moves on.

Once those costs are properly counted, the business case for change tends to make itself.