Thanks for the welcome. I’ve read AutomatedBuildings for years and I’m glad to finally be contributing. I’ll skip the part where I tell you my journey has been incredible and get to what I actually care about, because that is probably more useful to you.
Who Am I? The Guy Who Reads the Data-Ownership Clause First
I have spent my whole career on the vendor side of this industry. Two decades selling technical software, the last thirteen years at CopperTree Analytics, currently as CRO. So when I tell you the most important line in a building technology contract is the one almost nobody negotiates, understand that I am implicating myself.
Here is the line: what happens to your data the day you fire the vendor?
For most portfolios the honest answer is brutal. The operational history vanishes, or it comes back as an unreadable proprietary file no other system can ingest. The owner paid for the sensors, generated the data, and still ends up renting access to their own building’s history. If your platform does not give you full, usable data portability the day you terminate the contract, you did not buy software. You bought a mortgage on your own building’s history. You own the concrete, the chillers, and the ductwork. You do not, in any practical sense, own the data.
This is a contract and valuation problem at least as much as a technical one. I am not a lawyer, though I have negotiated enough contracts to argue like one, since legal review, drafting, and negotiation have been part of my job for years. I eventually took a contract law course through Harvard, “From Trust to Promise to Contract,” because I kept running into this in practice and wanted to understand the mechanics instead of guessing at them. The clause that decides who owns a building is almost never the one anyone negotiates. It is the one nobody reads twice.
Why I get to have this opinion
Most of my thirteen years at CopperTree has been spent on the unglamorous foundation: fault detection and diagnostics, an independent data layer the owner actually controls, and energy information systems that turn raw meter data into something an operator can use. More recently I have been in the room from the start as we built automated commissioning and automated system optimization, from requirements gathering through deploying the technology into real buildings and watching it succeed and fail on live sites. That is where my views come from. Not a deck. Job sites.
Before CopperTree I sold signal processing software, then subsea engineering systems. Different industries, same lesson: the hard problem is never the technology, it is getting clean, trustworthy data out of messy operations and into a decision someone will actually trust. Signal processing is the discipline of pulling meaning out of noise, and buildings are mostly noise.
The gap where most value dies
Most building technology value dies in translation. The engineer knows exactly what is wrong and cannot get it funded. The executive controls the budget and cannot follow the diagnosis. They are talking past each other, and the fix sits unfunded in the gap between them.
I have spent my career in that gap. I can sit with an engineer and talk static pressure, delta-T, and why the sequence of operations is fighting itself, then sit with a CFO and talk NOI, payback, and why the cheap chiller is the expensive one over thirty years. Not because I am special, but because I came up through both the technical side and the business side and never got to specialize into one. A lot of what I write here will be about closing that gap.
A word on “independent,” since everyone suddenly loves the word
I am a vendor. I want to be honest about that up front, because I am about to be hard on other vendors. What I sell takes data out of platforms the big traditional players built, the same players who spent thirty years with a very real commercial interest in keeping that data siloed inside their own boxes. So I have watched this from a specific seat.
Here is what is funny about this moment. The incumbents who built the walled gardens are now, all at once, announcing that they believe in openness. Everyone has an “independent data layer” this year. Everyone is throwing a party for a guest they spent decades locking out. And the wall is not really coming down. It is moving up the stack. These players have worked out that software and SaaS carry far better margins than hardware ever did, so the new moat is the recurring software subscription wrapped around the same system. The garden did not open. It just started charging rent by the month.
So ask the obvious question, and ask it of me too. The easy version is whether your “independent” layer is actually independent or still wired to one vendor’s cloud and renewal cycle. But the harder question is not about the data at all. It is about who gets to tell you the truth. If the vendor who designed the system, installed it, and profits from it is also the one grading how well it performs, how honest do you expect that report card to be? Are they going to hand you an analysis that says the sequence they wrote is wrong, the commissioning was rushed, and the fix is warranty rework that comes out of their own margin? Almost nobody falls on their own sword for free. This is the oldest problem there is, the fox reporting on the state of the henhouse, and the industry has spent years politely not mentioning it.
The only way an owner gets the real answer is to own the data and have the means to analyze it independently of whoever built the thing. That is the entire reason independent analysis exists, and it is the test I would apply to any of us making the “independent” claim, including me. Can you see the truth about your building even when the truth is inconvenient for the vendor who sold it to you? If not, you do not have independence. You have a nicer brochure.
Where I think this is all going
I am genuinely excited about AI in buildings, and I am going to write about it a lot here. But here is my bias, straight from a career spent on data foundations: AI is only ever as good as the data underneath it. Point a sophisticated model at inconsistent sensor data, undocumented equipment changes, and a decade of missing history, and you do not get intelligence. You get automated hallucination at scale, wrong answers produced with total confidence, faster than a human could catch them.
I am not an AI skeptic. I am a skeptic of doing AI on a broken foundation. The buildings that win with it will be the ones that did the boring work first: clean data, owned by the operator, structured so a model can reason about it. Most of the industry is racing for the exciting part and skipping the part that makes it work.
Why I am here
I have helped sell the industry’s software for a long time, and I would rather spend the back half of my career arguing for the version where the owner keeps control of their own building, and where AI gets built on something solid instead of something hyped. Ken built the place where this industry actually talks to itself, so this is one of the rooms where I most want to have that conversation.
I will try to be blunt, specific, and occasionally wrong in public. If I write something you disagree with, tell me. The best conversations in this business happen in the comments.
Keith La Rose
Follow me on LinkedIn, where I post a few times a week on this stuff.
