AI has made it much easier to produce an answer that looks finished. It has also made it much harder to know whether that answer deserves your confidence.
A polished investment memo can make a recommendation feel further along than it is. The numbers tie. The market discussion sounds reasonable. There are sensitivities, a risk section, and a clear recommendation. You can read the whole thing and still be unclear about what would actually have to be true for the investment to work.
I spend a lot of time teaching students how to build and interpret financial models. I also spend a lot of time thinking about what, exactly, I want them to be able to do with one. Producing the spreadsheet matters. So does being able to explain which assumptions are carrying the return, why those assumptions are reasonable, what would change the answer, and what happens if they are wrong.
Increasingly, I’m finding that the best way to evaluate that understanding is through oral 1:1 evaluations without any screens. It’s old-school, but effective, and definitely better than deciphering 90 blue books of chicken scratch for each assignment.
AI has made the distinction between producing the work and understanding the work much harder to ignore. I use it myself. It can help organize information, work through an analysis, test an idea, and get something onto the page much faster. That is genuinely useful, particularly with limited headcount.
But when a finished-looking answer is easy to produce, the burden shifts. The harder part is deciding what deserves attention, what the answer depends on, what information is still missing, and when more analysis is no longer going to make the decision for you.
Domain expertise matters more because you need enough command of the subject to question the output and distinguish what is actually correct and useful from what is technically plausible but not actually useful.
That is what I think expertise is for now, and where judgment and discernment become that much more valuable.
Expertise changes where you look
Consider an apartment acquisition where the rent roll shows rents below the surrounding market. The obvious opportunity is to raise rents as leases turn over, and a model can calculate the resulting income and returns very neatly.
Before getting excited about the upside, I would want to understand the comparison. Are those competing properties actually comparable? Are their advertised rents what residents are paying after concessions (effective rents rather than asking rents)? Does the subject property’s unit mix, condition, location, or resident profile explain part of the gap? What happens to occupancy if you try to close it?
You can do a great deal of analysis on a rent increase that was never available to you in the first place.
That is one of the places experience matters. It changes which questions you ask first, which information you distrust, and where you spend the next hour. It can also help you see an opportunity a standard comparison misses. Maybe the better investment is in common areas, a different leasing strategy, or an operating change that has very little to do with renovating apartments.
The value is not just having more information. It is knowing where to look.
That applies well beyond real estate. A leadership team can commission a detailed compensation study when the harder issue is that nobody has agreed on what the role is actually responsible for. The study may be excellent. It still leaves the decision they are avoiding untouched. This is the kind of problem I work through with leadership teams in Organizational Advisory.
An accurate answer to a secondary question can absorb a surprising amount of time.
Expertise helps you open up the assumptions
A lot of professional shorthand compresses decisions that somebody once had to think through.
Market rent. Stabilized occupancy. Normal staffing. Reasonable growth.
Each sounds like an input you can look up. In practice, each may contain several judgments about what is comparable, what is sustainable, and what the future is likely to look like.
Two analyses can be internally consistent and still answer different questions. One may describe the income supported by current leases. Another may assume those leases turn over at higher rents. Both can be mathematically correct. The distinction matters enormously if you are deciding what to pay today.
A useful expert should be able to open those assumptions up and explain them in language that lets other people participate in the decision. If the recommendation only holds together while everyone accepts the vocabulary, I start questioning how much we all actually understand.
There is also a question of consequence. Being wrong about a minor expense item is different from being wrong about when a building can open. A delay can affect revenue, financing, staffing, contractual commitments, and sometimes the entire capital plan at once.
Experience helps you see those connections sooner. It should make the analysis more specific about what could happen and what you would do about it.
It does not make you immune to getting it wrong. In fact, expertise creates its own risks. A familiar pattern can send you in the wrong direction if the conditions around it have changed. The fact that something worked before can make it easier to stop asking whether it still applies.
That is why expertise should make you better at questioning an answer, including your own.
Expertise helps distinguish a missing fact from a real decision
Sometimes the next step is straightforward: get the missing information. Read the lease. Call the operator. Check the source. Ask the person who actually has to deliver the work.
Other times, the information is already there.
People are disagreeing about acceptable risk, the purpose of the investment, how much uncertainty they are willing to absorb, or whose interests should take priority.
More analysis can clarify the trade-off. It cannot make that choice on its own.
A team considering an expansion may know the likely cost, the staffing requirements, and the effect on near-term earnings. What it has not decided is whether it wants to absorb that disruption to pursue the opportunity. Another forecast may narrow the range of outcomes a little. Eventually someone still has to decide which risk the business is willing to take.
Knowing when you have reached that point is part of professional judgment too.
So is recognizing when you have reached the edge of your own knowledge and need somebody who understands the part you do not.
I think this is a much larger part of what people are asking for when they ask for expertise. Help me understand the decision. Tell me what I am overlooking. Explain which uncertainty matters. Help me distinguish a missing fact from a choice I have to make.
The fundamentals matter more, not less
One of the stranger conclusions I have reached as AI gets better is that the fundamentals matter more, not less.
If a tool can produce a plausible market analysis, financial model, investment memo, strategy deck, or recommendation in minutes (all beautifully formatted, of course), you need enough command of the underlying work to know when the answer is wrong, incomplete, or answering the wrong question entirely.
You need to know what good looks like before you can use faster tools well.
That is why I still care about students learning how a model works rather than only how to get one produced. It is why I want them to understand the operating logic underneath the inputs, not just where the inputs go. And it is why I think experienced professionals still have an important job even when a machine can generate something that looks very polished.
Information is getting cheaper.
Plausibility is getting cheaper too.
Judgment is not. In fact, I would argue that judgment becomes more valuable with each passing day.
The part that remains difficult is deciding what matters, what the evidence really supports, where the answer is fragile, and when you know enough to act.
I am less interested in whether someone can produce an impressive answer than in whether they can show me why it is a useful answer to this particular problem.
Sometimes that means building a better model. Sometimes it means finding the person who knows something the model leaves out. Sometimes it means challenging the premise of the analysis altogether.
And sometimes it means acknowledging that we have enough information, the trade-off is real, and it is time to decide.