The Algorithmic Office: Why the future of workspace is written in code, not blueprints
By Rupesh Kumar, Chief Product & Technology Officer, WeWork India
For decades, the workspace has been designed the way most buildings are: a brief is taken, a blueprint is drawn, and the result is built to serve a workforce forecast months or years earlier. That model made sense when headcount was relatively predictable and teams worked together five days a week. It makes far less sense now. Hybrid work has made attendance patterns fluid, while teams can grow, shrink and reorganise within a single planning cycle. A floor plan fixed on paper is increasingly being asked to serve requirements that refuse to stay still.
This is why the next phase of workspace design will look different. Physical design will remain central, but it will increasingly be combined with code, data and computational intelligence. This is less about a programming language than a fundamental change in design philosophy: from creating a space once to building a system that can learn from how it is used. The workspace is no longer simply something we design and deliver; it is something we measure, test and improve.
Generative design is a good place to see this shift. Instead of a designer developing a handful of layouts and choosing among them, computational tools can generate and evaluate hundreds of floor-plan possibilities. Each can be tested against variables that matter to the people using the space, including access to daylight, circulation, acoustic comfort, opportunities for collaboration and density.
The value is not simply that technology produces options faster. It allows teams to understand the consequences of design decisions before those decisions become expensive physical realities. The designer’s question moves from “which option works?” to “given everything we know, which trade-off is right for this team?” That is a richer and better-informed question.
The quality of those answers, however, depends on the quality of the inputs. Workspace decisions have traditionally leaned heavily on predicted requirements: how many desks, how many meeting rooms, what ratio of collaborative to focus space. Spatial analytics adds another layer: evidence of how people actually use the workspace. It can show which rooms are consistently booked but unused, which zones experience peak demand on particular days, and where people naturally choose to collaborate. That matters because stated workspace preferences and actual workspace behaviour do not always align.
Across the workspaces we study, this gap between what teams say they need and how they actually move through a space shows up again and again, and it is precisely this gap that data can close. Designing from observed patterns, rather than assumptions alone, creates a feedback loop in which each workspace can inform the next. At WeWork India, operating 79 centres across eight cities and serving over 113,000 members, we see every day how differently teams use space from how they expected to.
The bigger shift is that this intelligence and feedback does not have to end once a workspace is built. Connected systems and IoT sensors can feed real-time information into building operations – adjusting lighting and HVAC based on occupancy, improving room availability and identifying persistently underused areas. Design and operations, once separate disciplines with separate timelines, start to run on the same evidence. The workspace moves from periodic intervention towards continuous optimisation, where operational data informs design, design changes behaviour and that behaviour generates new data. This is where the comparison with software becomes useful: the physical environment remains fixed, but the intelligence governing how it operates can keep evolving.
None of this means algorithms will replace architects and designers. A model can optimise for daylight, circulation or density, but it cannot decide what a space should feel like, how it should reflect an organisation’s culture or what will make people want to return to it. Those decisions demand judgement that no dataset can supply on its own. The role of technology is not to remove human judgement from design, but to sharpen it. By taking on repetitive modelling and pattern recognition at scale, computational tools free designers to spend their time where it matters most: the parts of workspace experience that were never going to be reducible to an equation.
That pairing also carries responsibility. The more a workspace learns from data, the more carefully that data must be handled. Occupancy and utilisation insights should serve better spaces, not surveillance of individuals. That means collecting only what is necessary, aggregating and anonymising data wherever possible, being transparent about what is measured and why, and building security into connected systems from the outset. At WeWork India, our spatial analytics platform works on anonymised computer-vision data. It measures how zones are used, not who uses them. A workspace cannot become more intelligent at the expense of the trust of the people using it.
The blueprint is not going away. Every workspace will still begin with a considered vision of how people should work together. What is changing is the idea that design ends when construction does. Increasingly, the blueprint is best understood as a hypothesis – a view of how a space should work, tested continuously against how people actually use it.
The organisations that get this right will treat physical space the way software teams treat a product: shipped, observed, and never quite finished. In that sense, the future workspace will be written as much in code as it is drawn on paper.