Cleaning is, for the most part, execution work on location: cleaning spaces, restocking materials, carrying out checks, reporting incidents. Around that sits a layer of planning work that is less visible but does determine costs: making rosters, arranging substitutes, handling complaints, scheduling quality checks and reporting to clients. Conditions on location strongly steer the outcome: one building has a fixed morning crew with few changes, another works with rotating teams, short contracts and many casual substitutes. That variation determines which part of the work can be captured in a pattern and which part remains dependent on what someone encounters on the spot.
These two layers — execution on location and planning at a distance — respond differently to AI. It is not that one layer is "smarter" than the other; it is that one lends itself better to being captured in data than the other.
Within the planning layer, part of the work has already shifted. Rosters based on fixed patterns, hours registration, invoicing and basic quality reporting are tasks that AI can take over once the underlying data is structured. Assessing an acute staff shortage due to illness, judging a complaint that does not fit the format, or negotiating a contract change with a client: that remains human work, or work in which AI makes a proposal and a manager approves or rejects it with reasoning.
On the floor, that is different. Sensor technology that reports when a space has actually been used and therefore needs cleaning is a form of AI support that already exists, but it requires investment in equipment and a building suited for it. Many locations do not have that. Cleaning itself, apart from robot deployment in specific environments, remains human work.
Why does this already work at one cleaning company and not at another? The difference rarely lies in the ambition itself — almost every company wants to plan more efficiently and lose less time on manual rosters. The difference lies in what underpins that ambition: does the company already have hours, sick leave and complaints in a system that an AI application can read, or is that still in the heads of planners and in separate spreadsheets per region.
A cleaning company that says "we want to automate rosters" means, at one organization: a system that independently generates a complete roster. At another: a tool that makes an initial proposal, after which a planner adjusts it. Those are two different ambitions with different requirements regarding data, the authority of planners, and the extent to which clients must agree to changes. As long as a management team has not discussed that difference with each other, it is talking about automation without knowing which variant it means.
That is exactly where an ambition structured in layers — from vision to a concrete target state — makes a difference: it forces the organization to specify what changes in who is allowed to make a decision, and who remains responsible for it. For comparison, in other sectors with a lot of execution work on location, a similar pattern plays out: those who read how that works out in construction or in the installation sector will see the same dividing line between planning work and execution work, with just different data sources.
The most frequently mentioned obstacle is not technology but record-keeping. Shift hours, absence, complaints and material use are often scattered across regional systems, paper forms and verbal agreements between planner and location manager. An AI application that takes over rosters or quality checks needs that data in a form that can be read. Without that foundation, an ambition remains a statement without a base.
In addition, there is the question of who is allowed to approve or reject an AI proposal. At companies where the planner currently has full authority over roster changes, an AI application that already rosters on its own effectively changes that planner's role into that of a controller. That touches on how positions are structured and on terms of employment; for decisions affecting personnel, separate statutory requirements apply and are not addressed here.
The question underlying all ambitions in this sector is not "can AI take over cleaning" but "which of our tasks are already recorded in a way that an AI application can read, and which are not yet". This differs per company, per region and even per client, and cannot be answered in general terms. What that means concretely per task for your organization is mapped out by the work scan from FTE TO AI: an overview of which work already lends itself to being taken over, which part requires oversight and which part remains human work. Those who first want to understand how an ambition translates into an operating model will find that explained in the target operating model in plain language, and those who want to know which capabilities a specific ambition requires can read that in the capabilities your organization needs for this ambition.
Formulating an ambition for AI in cleaning work begins with establishing where the organization truly stands, not where management hopes to stand. The free readiness check from FTE TO AI consists of eight short questions, one per dimension, and produces a picture of where the organization is furthest along and where the most steps remain to be taken. The full ambition assessment, with the four layers and the five confidence gates, is under construction.
Vertel wat u wilt bereiken, dan kijken we samen wat daarvoor moet staan.
Answers come from this site’s knowledge base. Not tailored advice, and not a scan of your company.