A verdict you can trace back to a clause
A written read on the deal — what supports the price, what the risks are, what to negotiate — grounded in the rent roll and the documents it was built from rather than in a summary of what you typed.
Not a score. An argument you can disagree with.
A number between one and ten tells you nothing you can act on or push back against. What is useful is a stated position, the evidence under it, and the specific things that would change it.
- A position, and whyWhat the income supports, what the vacancy is worth, and whether the price accounts for the things that could go wrong.
- The risks, named and sizedTenant concentration over threshold, rollover clustering inside the hold, assumptions that look optimistic — each with the figure attached.
- What to negotiateThe specific asks the analysis thinks the findings justify, which is the part you can actually take into a conversation.
Below-market in-place rents across 162,000 leased SF support the asking price, and the 22,500 SF vacancy is the upside rather than the risk. The concentration and the 2027 roll are what the price has to account for.
Key risks
- Cardinal Freight is 48% of base rent — over the 25% concentration threshold
- 31,600 SF rolls in Nov 2027, inside the hold period
Negotiation opportunities
- Price the vacancy at market lease-up cost, not at stabilised value
Why it can say anything worth hearing
Most AI analysis in this category reads a form. You type a price, a cap rate and an NOI, and the model writes paragraphs about the numbers you just told it. It cannot find anything you did not already know, because you supplied everything it has.
This one reads a completed underwriting. The rent roll came out of your leases, the NOI came out of the rent roll, the cash flow came out of the NOI, and the seller’s claims were captured separately without ever becoming inputs. So the model can compare things that were derived independently — which is the only way an observation can be news.
That is also why every judgement carries a citation. An AI verdict you cannot trace back to a lease is worth precisely nothing, and the clause behind every term is what makes tracing it possible.
Decision support, and we mean the word support
The product's stated purpose is judgement, context and decision support. It is worth saying plainly what that excludes, because the category is full of claims that do not survive contact with a real deal.
- It does not decideIt raises what a careful reader would raise, with the evidence. Whether to buy remains a judgement involving things no document contains.
- No accuracy percentage, because we have not measured oneA published figure we could not stand behind would undermine the one thing this feature is for. Instead the reasoning is shown so you can check it.
In-place income covers debt service comfortably and the price is defensible on current rents. The difficulty is that most of the modelled return arrives at sale rather than during the hold.
Key risks
- 68% of total return sits in the exit, at a cap rate 40bps tighter than entry
- No replacement reserve in the expense stack
Negotiation opportunities
- Test the exit at entry cap before committing to the price
Questions people ask
What does the AI analysis actually read?
The completed underwriting — the rent roll built from your documents, the cash flow derived from it, the debt, the exit — plus the documents themselves. It is not summarising a form you filled in. It is reading a model whose every number traces back to a lease or a statement.
Does it give a buy or pass recommendation?
It gives a written verdict, with the reasoning and the risks it rests on. Whether you buy is your decision, and the honest framing is that this is decision support rather than a decision. What it is genuinely good at is making sure the obvious objection has already been raised before somebody else raises it.
How do I know it is not making things up?
Every judgement cites what it rests on — the clause, the suite, the line in the cash flow. A verdict you cannot trace back to a document is worth nothing, so the citation is not a nice extra, it is the mechanism that makes the output checkable at all.
Is the analysis accurate?
We do not publish an accuracy percentage, because we do not have a measured one and inventing a number would undermine the point of the feature. What we do is show the working. The model states the risks it identified and the evidence behind each, so you are reviewing an argument rather than trusting a score.
Does it replace underwriting the deal yourself?
No. It reads an underwriting you have already built and tells you what it sees in it — concentration, rollover clustering, assumptions that look aggressive, the gap between the seller's claim and your model. It is fastest as a second opinion and worst as a substitute for having one.
Where this goes next
Stated vs derived check
The arithmetic finding that often drives the verdict.
Deal audit
Errors in your own model, checked separately.
Cash flow modeling
The model the analysis actually reads.
How to analyze a CRE deal
The framework, done by hand.
All capabilities
Everything else the product does.
Decide whether to buy
The job this capability exists for.
Get a second read before you commit a week.
Build the deal from your documents and DealWise reads the finished underwriting — what supports the price, what the risks are, what to negotiate, each with the clause behind it. Free plan, no credit card.