AI Makes Clients Smarter. Here's the Context It's Still Missing.
Healthcare founders now show up to real estate decisions having already run their sites and leases through AI. That is a good thing. But AI answers with confidence and without your context. It doesn't have your drive-time data, your lease economics, or the specifics of your project. The work now is human-in-the-loop: take what the model produced and add the context it could never have. That is where the right call gets made.
A patient walks into an exam room having already run their symptoms through an AI chatbot. They arrive with a theory, a printout, and better questions. The good clinician doesn't dismiss any of it. They take the research seriously, then add what the model didn't have: the exam, the history, the labs, the judgment that turns a generic answer into the right one for this patient.
The same shift is happening in healthcare real estate. Founders come to me having run their site shortlist or their lease through Claude or ChatGPT. They show up sharper and more engaged than founders did three years ago, and I want that. It makes the work better. It also surfaces one thing quickly: AI is confident, and AI is missing your context.
What happens when AI grades a site it doesn't understand?
Here is the example that made the pattern obvious.
The site analysis that flipped
A client ran our shortlist of sites through Claude for a SWOT analysis. The output was confident and well organized. It trashed the site we were moving forward with and gave a glowing review to one we had already passed on.
He sent it to me with a fair question: are we sure about this?
Here is what the model never had. The site it disliked had stronger demographics inside the actual drive-time, and the lease economics were favorable. The site it praised had good demographics, but not as strong, and a lease that penciled out roughly 1.5x more expensive. Claude was grading on surface signals it could find. It never saw the drive-time analysis or the deal terms, because those live in our work, not on the internet.
I told the client to feed it that context and ask again. With the drive-time numbers and the lease economics in front of it, Claude reversed its assessment completely. Same model, opposite conclusion. The only thing that changed was the context.
The lesson:
- AI judges on what it can see; your best sites win on data it can't see.
- Drive-time demographics and lease economics are proprietary, not public.
- A confident answer and a correct answer are two different things until someone supplies the context.
Why does AI give generic real estate advice?
Another client uses Claude to teach himself development concepts before our calls. I am glad he does. He shows up fluent in the vocabulary, which makes our conversations faster and our decisions sharper.
But the advice the model hands back is generic, because it is working from the textbook, not from his deal. Some of it applies to our project. A lot of it doesn't. It doesn't know his jurisdiction's plan-check rhythm, his landlord's posture, the long-lead equipment driving his schedule, or the three constraints we already designed around. General development advice and the right move on this specific project split apart the moment the specifics enter the room.
AI knows healthcare real estate in general. It does not know your deal. The distance between those two is the work you hire an advisor to close.
Where does AI save real time, and where does it slip?
None of this means I am skeptical of the tools. I use them every day, and the speed pays off where it counts. They turn a stack of RFP responses into a clean summary in minutes. They pull clear themes out of interview and meeting notes. They give me a strong first draft of almost any synthesis I need.
Then I check it against the context, because the first draft gets pieces wrong. When I compare two general contractor or architect proposals, each firm formats its bid differently, and an AI summary smooths that over. What it misses is the scope gap: the work one firm priced and the other left out without saying so. AI compares what is on the page. It cannot flag what should have been on the page and isn't. My read on the project, on market norms, and on how these proposals usually play out is what catches the omission.
"So should I just have AI do it?"
Run the experiment. You'll get a fast, articulate, confident answer. Then ask what it was missing:
- The drive-time and demographic data that lives in your advisor's analysis, not online
- The lease economics and deal terms that change the whole calculation
- The jurisdiction, the landlord, and the project constraints specific to your build
- The scope gap in a proposal that no one wrote down
AI gives you the general answer in seconds. The work is supplying the context that turns it into the right answer for your project.
The clients getting the most out of AI use it to arrive with a stronger draft and sharper questions. The context that makes the call still comes from the room.
Key takeaways
- Founders now arrive having run sites and leases through AI. That is an advantage to build on, not a threat to resist.
- AI grades on what it can see. Your best sites often win on drive-time demographics and lease economics the model never gets to see.
- One client's site SWOT flipped completely once we fed Claude the drive-time data and the deal terms. Same model, opposite call, purely from context.
- AI gives generic development advice because it knows the field, not your deal, your jurisdiction, or the constraints you've already designed around.
- Use AI for speed on RFP summaries and synthesis, but a human catches the scope gap a proposal leaves unsaid. Context is the job.
The Bottom Line
AI made my clients smarter, faster, and more engaged, and I am glad it did. It also made the advisor's job clearer: hold the context the model can't have, and catch the moment a confident answer is a wrong one.
You can move fast and break things in software. In brick-and-mortar, breaking things costs $200K and six months. Use AI for the speed. Keep a human in the loop for the context. That combination is what protects the capital.
Frequently asked questions
Can AI evaluate a site for a healthcare clinic?
It can produce a fast, confident SWOT, but it grades on public, surface-level signals. It does not have your drive-time demographic analysis or the lease economics, which are usually what separate a strong site from a weak one. In one real case, a client's AI site analysis reversed completely once we supplied that context. Use AI for a first pass; the decision needs the data your advisor holds.
Why does AI give generic real estate advice?
Because it works from general knowledge, not your specific deal. It does not know your jurisdiction's process, your landlord, your schedule, or the constraints you have already designed around. Some of its advice applies and much of it doesn't. The value of an advisor is closing that gap between the general answer and the right move on your project.
Is AI useful for comparing contractor or architect proposals?
For speed, yes. It summarizes bids and pulls themes from meeting notes quickly. But each firm formats a proposal differently, and AI tends to miss the scope gap: the work one bid excluded without flagging it. A human with project and market context catches what is missing from the page, which is often where the cost overrun hides.
Will AI replace real estate advisors?
No. AI gives a confident, general answer in seconds, but it cannot supply the proprietary data, deal terms, and project context that determine the right call. The clients getting the most out of AI use it to arrive sharper and ask better questions, then rely on a human in the loop to add what the model cannot have.
Bring Your AI Analysis. Let's Add the Context.
Your team is already using AI on sites, leases, and proposals. Good. I help healthcare founders and operators add the context the models can't have: drive-time and demographic data, lease economics, jurisdiction and project specifics, and the scope gaps a proposal leaves unsaid. Bring what AI gave you and we'll pressure-test it together.
- Why a confident AI site analysis can still be wrong
- The drive-time and lease-economics context AI never sees
- How to use AI for speed without inheriting its blind spots
- Where a human in the loop changes the call
