The Big Bet on Data-Informed Design

Every office designed over the next few years is a decade-long bet on how people will work. That is what property is: capital committed once, lived with for a long time. Right now, across most of the industry, that bet is being placed blind.

This is my actual passion point, and it is not analytics for its own sake. We design today, build tomorrow, and live with the result for the next ten years. Evidence only changes the outcome at the first step. Miss that window, and the only lever left three years after occupation, when the fit-out budget is spent and the lease is signed, is retrofitting what should have been right from the start. Front-load the decision with evidence, and you spend once. Skip it, and you spend twice: once to build it wrong, again to fix what you should have known the first time.

Not for lack of ambition. For lack of a foundation. Ask a real estate or workplace leader what evidence informed this year’s space plan, and you will get benchmarks, utilisation surveys, an architect’s brief built on the last cycle’s assumptions. In the conversations I have had, ask what evidence exists for how the AI era specifically should change that plan, and that is where it stalls.

I work at the intersection of data and corporate real estate, and this is the gap I keep coming back to. This is Part 2 of two pieces. Part 1, “Are We Designing the Workplaces of the Future?”, asks the harder question: whether the paradigm the industry is designing towards is even the right one. This piece is the follow-on: I think the data to inform a decision like that already exists, sitting inside a platform most real-estate teams have never thought to treat as a design instrument.

The stakes: a decade-long bet either way

Whatever the industry eventually decides about the shape of the AI-era office, more enclosure or less, it will not be a decision anyone gets to revisit next quarter. That is worth establishing before anything else.

Property is a slowly changing dimension

For the data people reading this, I mean that precisely. Property is a Slowly Changing Dimension, Type 2. It does not update in place. It changes state by adding a new row: a new lease, a new fit-out, a new floor plan, and the old row stays valid for years after it was written. You live with what you signed.

A commercial building is typically committed on a lease of ten years or more, and that number is not really a real estate habit, it is an accounting one. Fit-out costs get capitalised as leasehold improvements and amortised over the shorter of their useful life, commonly assumed at around a decade, or the remaining lease term, under both IFRS 16 and US GAAP. Exit early and you write off whatever is left unamortised as an immediate loss, so finance sets the floor on the lock-in before real estate even opens the brief. I see it at the bank I work in, and it holds the same way across most large organisations, not just banks.

The space is then fitted out on a five to ten year cycle and designed against a standard someone wrote for the way people worked when the brief was set. Hold that against the clock the technology runs on: the interface shift we are living through is measured in quarters. What we draw on the plans this year is a bet on how people will work in 2035, placed with assumptions from 2024. The two clocks are wildly out of sync, and property is the slow one. Get it wrong and you do not patch it next sprint. You carry it for a decade.

I made the case for what I think that decision should actually weigh in Part 1: that voice-first, human-plus-agent collaboration is more likely to need more enclosed space than less, backed with real evidence rather than just conviction. Whichever way that argument lands for you, it depends on data that mostly does not exist yet inside a typical real-estate team. This piece is about where that data already sits, hiding in plain sight.

The opportunity: the big bet

Here is the how, not the point. The point is getting the design right before the capital is spent. What follows is simply the clearest answer I have found to where the evidence for that already sits.

My bet is Microsoft Places, or more precisely, the stack Microsoft has assembled around it. Not a product endorsement. A data thesis, and one I want to state precisely, because the imprecise version of it does not survive scrutiny.

How people work

Microsoft owns the richest behavioural signal of that in existence: Microsoft Graph, the model underneath Microsoft 365, mapping meetings, documents, collaboration, and the flow of work across an organisation. This is not speculative. Microsoft already markets Graph as a source of workplace insight, down to analysing meeting requests to understand conference-room utilisation. On this specific axis, the depth of the organisational-collaboration graph itself, nobody else is close: no space-analytics vendor holds anything approaching it. That does not settle the wider question. It settles this one.

Where people work

In November 2024 Microsoft brought Microsoft Places to general availability: a workplace app that layers occupancy and utilisation data, interactive floor plans, and portfolio-level space insight on top of that same Microsoft 365 fabric. But Microsoft did not build that spatial layer natively, and it is worth being precise about that, because it changes what the bet actually is. It partnered the spatial layer in. Archilogic markets itself as Microsoft’s official global mapping integrator for the product; Mappedin and Pointr convert floor plans into Places-ready maps at similar scale. VergeSense feeds occupancy sensing and predictive space planning straight into the floor plans customers already manage inside Places, no separate onboarding required.

Licensing changed too, and it is worth calling out because it changes who the bet actually reaches. Until 1 April 2026, most of what makes Places useful sat behind Teams Premium, a per-user add-on priced around £7.70 a month, so adoption was gated by how many seats an organisation chose to license. From 1 April 2026, the core of Places, work schedules, colleague availability, the Places card, hybrid RSVP, is bundled into standard Microsoft 365 licences at no extra cost, and the advanced layer, desk booking, check-in, utilisation analytics, moves to a new per-space licence, Teams Shared Space, priced against the desks and rooms being managed rather than the people using them. Microsoft’s own framing is that desks, not users, are the scarce resource in a downsized hybrid estate. Whatever the motivation, the effect is structural: Places stopped being an opt-in add-on for whichever seats a company chose to pay for, and became close to the default state of a Microsoft 365 tenant. The addressable footprint just went from however many Teams Premium seats an organisation had bought to essentially every Microsoft 365 tenant there is.

Why not a specialist, or build it yourself?

The honest answer is that a specialist vendor can beat Microsoft on spatial precision. Archilogic’s floor-plan intelligence and VergeSense’s occupancy sensing are both genuinely strong products, sold to anyone, Places customer or not. VergeSense goes further and does its own version of behavioural fusion, reading calendar data directly to catch no-shows and ghosted bookings. If the question were purely “who has the best geometry and the best sensor data,” a specialist stack might win.

But that is not the question that matters for design decisions locked in for a decade. The organisational-collaboration depth only exists inside Graph, and building it yourself means re-implementing what these vendors already sell, plus carrying the hardware, integration, and data-science headcount to do it, without ever reaching Graph’s view of actual collaborative work. Worth noting too: the most credible challenger in this space is not competing with Places, VergeSense spent 2024 to 2026 plugging deeper into it instead. That could just be normal go-to-market, integrating with the dominant platform is what any enterprise vendor would do. But it is at least consistent with where the leverage sits, not evidence against it.

The join

Put the two together and the picture is still striking, just not for the reason I first reached for. This is not Microsoft uniquely sitting on a data mine nobody else can touch. It is Microsoft being the only party assembling the full stack: genuine collaboration-behaviour depth it owns outright, a best-of-breed spatial layer bought in through partnership, and the productivity-suite seat penetration to make the combination the default rather than an integration project. The mapping partners feeding Places have digitised well over fifteen billion square feet of interior space between them, across their whole businesses, not a Places-only figure Microsoft has ever published, but real evidence of the scale they operate at. No architect, no REIT, no facilities team assembling its own stack has anything remotely like that combination.

That is the data that could inform the designs of the future, an AI future. Not opinions about what a good office feels like. Signal.

The catch, before someone else says it

Say the obvious risk out loud. The same join that could design a better building can just as easily run a worse one. Fuse how-you-work with where-you-are and you have a surveillance engine as readily as a design tool, and the criticism has already started: journalists have flagged the newer presence-tracking features as a return-to-office enforcement mechanism dressed as convenience. Microsoft’s current answer is that the feature is off by default and requires individual opt-in consent. But the signal it produces is person-level, and whether it is ever aggregated to the building rather than used to monitor individuals is a choice, not a technical guarantee. That distinction is exactly the one that matters.

Aggregate signal, used to design better places for people, is a gift. Per-person signal, used to monitor them, is a trap. Design leadership does not get to be neutral about which one it is building towards. If we are going to use this data to shape the workplaces we lock in for a decade, we have to insist, loudly, on the first and refuse the second.

The conclusion

Let me separate what I am sure of from what I am not.

I am sure the office is a slow, capital-locked asset, and that we are designing right now for a voice-first, human-plus-agent way of working that is arriving faster than the property cycle can absorb. I am sure that we are right-sizing our portfolios against the world that is leaving and not the one that is coming. And I am sure the data to do this properly largely already exists, sitting in a stack most real-estate teams have never thought to treat as a design instrument.

What I am less sure of is the answer itself. Whether the future floor needs more enclosure, different ratios, a reversal of the densification of the last ten years. I have strong instincts, but instincts are exactly what this decision cannot run on, which is the whole point of this piece.

The real work is unglamorous, and it comes back to who is actually looking at this, and how. Leadership that owns the question, not delegates it to whoever happens to hold the AI brief. Standards rewritten for how people are actually going to work. And data, used honestly rather than exploited quietly, to inform decisions we only get to make once a decade.

The clock is already running. We design today, build tomorrow, and live with it for a decade, whether or not anyone used the evidence at the only point it could have mattered. The only real question is which version of that sequence we choose: deliberate, with the evidence that already exists, or found out the hard way, three years after occupation, retrofitting what should have been right the first time.

I want to connect with people already working through this: not the theory, the practical implications of how these spaces get redesigned and operationalised. If that is you, reach out.

Frequently Asked Questions

Mark Cunningham is the founder of Insights². He has spent more than a decade building data and analytics products at the intersection of data and corporate real estate. He writes at insights-2.com about how to make the meaningful measurable.