AI Agents and Appraiser Oversupply

[J Grant]

With an agent, you will basically have a wizard trainee that can process and input information at instantaneous speeds. My agent will be different than yours, and yours will be different than everyone else's. Appraisers will be able to imbue it with their individual market knowledge and their approach to valuation. In Year 1 it might just be looking over your shoulder watching your moves on the computer and listening to your reasoning. The better you are at training it to focus on the correct things, the more you correct its mistakes, the more valuable it will be to you. You can train it to refine data and select/exclude sales based on how you do things. You can train it to develop adjustment ranges based on all the various techniques available. Just like the supervisor trainee model, in Year 2, efficiency will really kick in because you won't be teaching nearly as much and you'll just be supervising.

Now, it's possible that appraisers won't adopt it en masse, so those who do take advantage will have the efficiency without the fee deflation. But that won't last long because eventually the market for GSE appraisals will get efficient too, and that will be deflationary.

It is possible that as efficiency grows, and as reviewers implement agents in their processes, the GSEs and lenders will add more reporting requirements to raise the quality bar. It is possible your software provider will not only steal your appraisal data, but they will now much more information to steal.

I don't think any of this is certain, just the direction I could forsee things going. Maybe it will get regulated, or there will be privacy roadblocks. Maybe we'll be going back to typewriters and snail mailing reports. Maybe tghe quality won't every be as good as I'm making it out to be. Then again, still seeing a lot of $35-40/sf GLA adjustments when they should be closer to $75-$125/sf.

The wizard assistant might save 10 minutes. Not for nothing but big freaking deal. Effiench is a ridiculous goal for this kind of work. Thought, time and human analysis still out think any machine that looks at the numbers still has no context for them.

Lousy appraisers will do lousy appraisals faster and good appraisers might do them only a little faster because a good appraiser wants to understand the market and verify/test their conclusions.

AI might empower more lousy appraisers to produce more because the don;t care. We see it all the time here, people looking for someone else to write it for them, or dump data in.
If speed becomes what is rewarded over competence and results, for what purpose, one might ask - none of the efficiency will increase the number of loans, which is finite given how expensive RE is to buy or refinance - meaning in a slower market which frankly looks set to last for years, lenders will be starving while the few loans they make close a day faster due to efficiency.

 

[Tom D]

Lousy appraisers will do lousy appraisals faster and good appraisers might do them only a little faster because a good appraiser wants to understand the market and verify/test their conclusions.

AI might empower more lousy appraisers to produce more because the don;t care. We see it all the time here, people looking for someone else to write it for them, or dump data in.

I haven't seen bad appraisers get better from 8 page reports to 30 page USPAP reports. But those doing fast without do care will AI them faster at even a lower fee.
You just answered, where will appraiser be in a couple of years. Well the good ones, gone.

 

[J Grant]

I haven't seen bad appraisers get better from 8 page reports to 30 page USPAP reports. But those doing fast without do care will AI them faster at even a lower fee.
You just answered, where will appraiser be in a couple of years. Well the good ones, gone.

Fees already stink, they can not get lower and still attract good people or keep them in the field. the minor time saving from AI should not open the door to lower fees - if it does, the bad appraisers will prevail because they do not care about results, understand the data, or anything - they data dump and sign, not caring is almost a reqwuriment at that point to get it done fast enough.

The needs for speed makes no sense, these are 15-30 year loans and each one is hundreds of thousands of dollars.

 

[Non Sequitur]

Heard it for 40 years.. and we are still here.

Not all of us. Appraisers are dropping like flies, a recent Fannie slide show illustrated a +/-25% drop in unique licenses submitting appraisals to their portal when comparing the time periods before and after the pandemic. IMO that's stunning.

 

[alebrewer]

The housing market seems to generate less and less income for lenders and agents ( and appraisers ) the more "efficient" it becomes.

Isn't that the entire purpose of efficiency gains? Implementing AI (or any process change for that matter) is only enacted if the change results in reducing costs - either by reducing the time to produce a gadget, or improving the technology/cost to make the gadget. The gadget you're talking about is, of course, housing. Lending institutions will replace loan officers with bots if, and only if, the change results in improved revenues (or lowered costs, depending on which side of the coin you're observing). RE offices will employ AI in a manner that does the same - whether that's building a program to sift through potential buyers, using bots to direct different buyers to different agents, etc.

It's no different than using technology to increase the size of chickens. We can now harvest more chicken meat per pound of feed - which reduces costs (and increases revenues).

Is any of this immoral? That's a question for philosophers - the market is amoral.

 

[J Grant]

Isn't that the entire purpose of efficiency gains? Implementing AI (or any process change for that matter) is only enacted if the change results in reducing costs - either by reducing the time to produce a gadget, or improving the technology/cost to make the gadget. The gadget you're talking about is, of course, housing. Lending institutions will replace loan officers with bots if, and only if, the change results in improved revenues (or lowered costs, depending on which side of the coin you're observing). RE offices will employ AI in a manner that does the same - whether that's building a program to sift through potential buyers, using bots to direct different buyers to different agents, etc.

It's no different than using technology to increase the size of chickens. We can now harvest more chicken meat per pound of feed - which reduces costs (and increases revenues).

Is any of this immoral? That's a question for philosophers - the market is amoral.

Reducing costs to what gain? RE lending is not a consumer product with mass production where they lower the costs and increase sales. (such as chicken size ) There is no increased revenue in proportion to cost in the lending business, and that includes valuation,- unless one is on the AMC side, perhaps.

The consumer pays for the valuation, not the lender, so how does reducing the cost of a valuation help them?
From a consumer viewpoint, if valuation costs $100 instead of $500, it won;t make more people run and take out more loans. The costs of an appraisal to a consumer was always worth it to get the loan. And if the valuion ccst is reduced, the consumer is still paying thousands to tens of thousands in loan fees and points.

More people made a ton of money in every facet of the RE lending and sales made a ton of money when it was less "efficient". less now -

A market does not have to be amoral - a market is what people want it to be. That is why regulations exist as safeguards - wasn't it fun after Bush scrapped regulations and the market soared then tanked?

 

[alebrewer]

There is no increased revenue in proportion to cost in the lending business, and that includes valuation,- unless one is on the AMC side, perhaps.

Huh? I literally have no words to respond... Of course lenders engage in cost reducing activities - EVERY...SINGLE...DAY. If a lender does 1k units a month and can cut $500/loan from their costs, then they've generated an additional $500k in revenue in a month.

Do you even think about what you're saying, J?

The consumer pays for the valuation, not the lender, so how does reducing the cost of a valuation help them?

Wow - just wow. The valuation is such a small part of the overall cost of the loan - yet in J's world, the valuation is the ONLY meaningful part of the loan transaction.

More people made a ton of money in every facet of the RE lending and sales made a ton of money when it was less "efficient". less now -

Money can be made two ways - increasing volume or reducing costs. You're being duplicitous if you assert that the money taking that was done in 20-21 was anything other than volume.

A market does not have to be amoral - a market is what people want it to be.

That is the exact opposite of what a market is. Markets are organized and driven strictly by the behavior of buyers and sellers. Regulation is the antithesis of market forces. Regulation is government intervention into market activities. That said - it is my personal belief that some level of regulation is necessary to preserve the 'free' nature of markets (contract enforcement for example), but even that is anti-market behavior.

 

[George Hatch]

Volumes being equal, saving $100 is considerably more profitable than charging $100.

 

[CGinMN]

The most likely scenario is that AI will force appraisers to get better. It will be very easy to run your report through an LLM to see if it meets the USPAP, Fannie/Freddie guides, and lender overlays. To the appraisers who think it's bad that the GSEs are now requiring you to summarize your support for market conditions, things could get a lot worse for you in all directions.

Not all of us. Appraisers are dropping like flies, a recent Fannie slide show illustrated a +/-25% drop in unique licenses submitting appraisals to their portal when comparing the time periods before and after the pandemic. IMO that's stunning.

Yep, it's right here. We went from 40,000 down to 29,000. That is a massive drop. I could see dropping another 50% in 3-5 years.

 

[George Hatch]

Some of that decline is attributable to the big reduction in transactions, much of which was driven by the increases in mortgage interest rates. I'm seeing datasets where the number of sales in 2023-2024 have been down by 40% when compared to previous periods.