By Ed Muna, President, National Lease Advisors
Barely a week goes by without another headline asking whether artificial intelligence has gotten ahead of the people trying to control it.
Recently, the news reported on an Anthropic researcher who resigned publicly because of concerns that AI was becoming too powerful to keep under control. A few years ago, that kind of story would have sounded like science fiction. Today, it barely raises an eyebrow.
Lawyers have also faced sanctions for filing briefs that included AI-invented case law that sounded perfectly plausible. Meanwhile, regulators in the U.S. and abroad are racing to create rules for a technology that continues to evolve faster than they can regulate it.
That is a safety conversation, and it is a much bigger one than we are qualified to referee. However, it has put AI on everyone’s mind. It has also highlighted a theme that matters enormously in our business.
The common thread in many of these stories is not simply that AI got something wrong. The bigger concern is that AI got something wrong while sounding completely certain that it was right.
At National Lease Advisors, we work at the intersection of real estate, accounting, and contract law. Every day, our team reviews leases, processes CAM reconciliations, validates landlord billings, and manages portfolios where a single misread clause can cost a client six figures.
So when people ask how we think about AI, they are often surprised by our answer. We are not afraid of it, and we are not caught up in the hype. Instead, we have a healthy respect for what the technology can do and, just as importantly, for what it cannot do.
We Use AI. We Just Don’t Trust It Blindly.
Let’s be clear: we are not AI skeptics.
We use these tools, and we are excited about where they are headed. AI is genuinely useful for certain tasks. For example, it can pull structured data from scanned documents, flag anomalies across large data sets, and create a first-pass summary of a long contract.
When used well, AI makes our team faster. We expect those capabilities to keep improving, and we plan to adopt tools that help us work more efficiently.
However, faster and right are not the same thing.
Lease administration depends on accuracy. A CAM reconciliation with a plausible-looking number that is actually wrong is not a minor inconvenience. It is real money leaving a client’s pocket, sometimes year after year, because no one caught the error.
We keep seeing that same failure mode with AI more broadly. The problem is often not obvious nonsense. Instead, AI can produce a confident, polished, well-formatted answer that is completely wrong.
In lease administration, “looks correct enough” is exactly where costly billing errors can hide.
Data Extraction Is the Easy Part
Here is something people outside our industry often do not realize: pulling dates, rates, and clauses from a lease is the easy part of lease administration.
Any reasonably capable system, whether human or machine, can extract data.
The real value comes after extraction.
Commercial lease language is rarely clean or simple. Defined terms often cross-reference other defined terms. Leases may include escalation clauses that attorneys drafted years ago for buildings that have since changed ownership multiple times. Operating expense provisions can also reflect complicated compromises between landlords and tenants.
Interpreting that language correctly requires judgment, context, and experience.
It also requires an understanding of how lease disputes play out in the real world.
An AI system may be able to summarize a clause, but it does not know that a particular landlord routinely includes questionable charges in CAM reconciliations. It also does not know that the parties heavily negotiated a specific “permitted charges” provision three amendments ago, changing how the tenant should interpret it today.
That context matters.
Human Judgment Goes Beyond the Lease Language
The relationship work matters too.
Our team does not simply process a lease and move on. We pick up the phone, call landlords and property managers, and work through discrepancies directly. When a charge does not belong, we push back. When an issue requires negotiation, we advocate for our client.
Our team also challenges line items that might pass through most accounting departments and many AI systems without a second look because they look legitimate on paper.
Knowing that a charge is improper is not just a data problem.
It requires understanding what the lease actually permits, recognizing patterns across a portfolio, and questioning a landlord when something does not look right.
Sometimes that also means having an uncomfortable conversation with a landlord who expected no one to look closely.
That is not a prompt.
That is expertise built over years and applied by people who know what they are looking at.
Where We Land
AI is here, and it is not going away.
We believe it will make lease administration, and most knowledge work, more efficient over the next several years. We also plan to adopt tools early when they genuinely help our team work faster without sacrificing accuracy.
What we are not interested in is a version of efficiency that hands important judgment calls to a system that can deliver the wrong answer with complete confidence.
Our clients do not hire us simply to extract data.
They hire us to protect them.
That means navigating ambiguous lease language, identifying charges that do not belong, and catching quietly incorrect answers that might otherwise go unnoticed.
People still provide that protection, and we believe they will for a long time to come.
We are embracing AI as a tool.
We are not confusing it for a colleague.