CRE Debt Underwriters - How concerned are you about AI?
After using Notebook LM by Google, I'm not entirely sure this thing isn't too far away from taking an appraisal, leases, CoStar data, and spitting out a draft memo that would only require you to review and edit. Is anyone else concerned about this not necessarily taking their job, but certainly decreasing the need for the current number of underwriters in banking?
AI's impact on CRE debt underwriting is a hot topic, and here's the scoop based on the most helpful WSO content:
AI and automation are already making waves in finance, and CRE debt underwriting is no exception. Tools like Google's Notebook LM and other machine learning applications can indeed process large datasets—like appraisals, leases, and CoStar data—and generate draft memos or analyses. This capability could significantly reduce the time spent on routine tasks, making underwriters more efficient.
However, while AI can handle data crunching and generate initial outputs, it struggles with the nuanced judgment calls and relationship-driven aspects of underwriting. For example, interpreting unique deal structures, assessing qualitative factors, and negotiating terms are areas where human expertise remains critical.
That said, the concern about a reduced need for underwriters isn't unfounded. As AI becomes more advanced, the demand for junior-level roles focused on repetitive tasks may decline. This shift could lead to a leaner workforce, with a greater emphasis on senior professionals who can oversee AI outputs and handle complex decision-making.
In summary, while AI might not replace underwriters entirely, it could reshape the role, emphasizing strategic oversight and reducing the need for entry-level positions. Staying ahead in this evolving landscape means focusing on skills that AI can't replicate, like critical thinking, relationship management, and creative problem-solving.
Sources: Google partners with Goldman Sachs in automating Investment Banking, Calling out Goofy @CRE and the rest of Wall Street Oasis. Real Estate is the worst asset class going forward., https://www.wallstreetoasis.com/forum/hedge-fund/machine-learning-taking-over-hf-research-analyst-roles-in-near-future?customgpt=1, HYPSM ex-Quant Fund Intern Senior career advice and opinion on the future of finance and the world, GPT-4 & Microsoft Co-Pilot
PGIM is actively testing AI right now that will replace 95% of their Core/Core+/Value Add originations teams headcount. Last I saw/heard, the team structure will be one originator who just manages relationships with one associate who just double checks the numbers and edits the pricing memo they put out to their portfolio managers. They believe AI can now handle deal screening, underwriting, memo drafting, and market comp analysis—basically everything but relationship maintenance.
Debt AM and Surveillance will be 100% AI automated, not surprised because that's pretty brain dead work anyway. It's kind of messed up because the AI company that is developing the software for PGIM to replace all these people is using their input to bring it on faster. 2027/2028 is estimated rollout.
What do you mean by surveillance?
Back office loan monitoring. Brain dead work
That's nuts. What's the name of the AI company developing this software?
Its proprietary with the help of outside consultants. The capital P
This seems far fetched - what’s your source?
Source: Trust me bro
Curious if you've seen or heard of any updates on this since?
Short answer, if I were in Debt UW still, no. I'm probably more bullish than most, but I actually think AI creates more efficiency and allows for junior-level people to churn out more output. A lot of the stuff, like deal screening, modeling, drafting memos, is already pretty automated, at least in my experience. Most MF credit teams are already lean as-is. Banking went a little crazy with hiring during COVID, so there may be multiple factors contributing to a reduction in headcount besides just AI. I may be naïve, but we tend to overestimate the short-term and underestimate the long-term impacts of new technologies.
Become more efficient in what exactly? Checking the output? Why would an employer pay multiple people 150K+ to do that?
so there is a logical loop here. How can you check the output if you don't know the deal? You would have to read it anyway to check it.
I used to be in debt. Maybe some shops will replace uw with ai, as noted above w/pgim, but I don’t know how that could work well for high touch deals. It is beyond my comprehension how AI could manage an underwriting process on a deal where budgets are shifting, things are changing, judgement calls need to be made, structure added, etc. Pair that with difficult borrower who doesn’t have all the requested docs and it seems like the thing will turn into a disaster. I can understand how it could add efficiency to things like memos and checklists, but those are just parts of a larger process to ensure the deal checks out (or can be restructured in a satisfactory manner).
imagine ai on both the sellers, brokers, and buyers side all communicating with each other. i saw this video the other day about 2 ai callers talking to each other, and they recognize they're both ai, so they decide to speak in like some more efficient mode of english to convey thoughts faster
I’ve imagined and what I see is dumb.
If I was an LP I would never invest with a group that utilizes an originator+ai underwriter setup instead of a traditional originator+human uw setup. Why outsource dd to a black box computer?
Beyond the reasons I noted above - what benefits would come from replacing a human uw with ai? Who would those benefits accrue to? Are the investors going to see higher returns? lol no
FWIW that video is fake - ai hasn't created its own robot language to talk to each other out loud (yet)
Everyone on the debt side knows this so it's the worst kept secret- There are very few people on the debt side more focused on self preservation than underwriters. Reason is simple, the job is not rocket science and you get paid $250K+ for working 40 hours a week and sometimes it's 30 hours a week with the ebb and flow. So they are focused on one thing- nobody should know how good they have it and they will actively be against portions of their work be outsourced to India because then the fear is at some point their entire role could be outsourced or with fewer underwriters needed, there could be displacements. AI is just the 2.0 version of their work being outsourced to India. At every large company there will be a "strategy" person who will propose looking into AI or some new shiny toy every few months (largely this person is also focusing on self preservation and justifying their $250K+ pay and existence) and underwriters will do everything they can to resist and shoot these down. Reason is simple, if enough of them resist and if they have support from the senior management, how will "AI" replace them? Particularly the big banks have more money than they know what to do with them (hence the strategy person and underwriters earning $250K+ and tons of lifers) so I dont really see people voluntarily choosing to self sabotage and replace themselves with AI. It's not AI that will one day decide that there will be fewer underwriters going forward because of AI, it is people after all (who have their own goals and it's often just self preservation) that will be the gatekeepers and decide if AI is going to result in fewer underwriters or not.
With all that said if there is assistance with AI for memo creations and analysis of the numbers for instance, I personally think it will only be beneficial as the job will then transition to a more project manager role which is kinda is with managing logistics of 3rd party reports (AI could assist with reviewing and summarizing them versus the manual reviews that are being done today) site inspections, closing timelines, etc.
"Reason is simple, the job is not rocket science" You could say the majority of people working in finance. And who is getting paid $250k for 40 hours a week on the underwriting side? We top out at $150k at best assuming you are in a HCL market.
a lot of jobs are not rocket science but once you get to underwriting stabilized assets especially in asset classes like multi's (and that alone is an industry that employs tons of people given $100B+ is financed very year by just Agencies), you sometimes have to wonder if this is even real. I mean while almost everyone who underwrite stabilized multis will have a college degree, do you even really need it though? It's that simple. Even other assets classes certainly have a rinse and repeat nature to them once you have done it for 5+ years. You certainly get to use more of your braincells in some other roles in finance or if you dont get to use your braincells and are just doing mindless work but get paid well, those roles typically require you work 60 or 80+ hour weeks. Debt underwriting particularly at the big banks is often one of those have your cake and eat it too type jobs, or at least that is the goal. $150K is base for junior underwriters at my place. There is a small bonus on top of that. Senior underwriters will make $200K+ on just base and with bonuses make $250K+.
That is actually a way more applicable mentally framing of AI for me. “If something can be outsourced to a call center, it can probably be done by AI” makes more sense to me than most of the AI boosterism that I have seen.
Voluptatem nemo corporis voluptatem. Repudiandae similique iure magni voluptatem hic laborum. Non mollitia ut non voluptas nam sint. Saepe eius qui ipsa molestiae. Minima maxime maiores quisquam. Doloribus labore odio odit voluptatem voluptates et sit.
Corporis labore quia accusantium vel excepturi. Corporis vero libero modi. Ut ea repellendus repudiandae animi quo dolorem ea. Nostrum labore distinctio assumenda nam atque necessitatibus quod. Voluptatem in est assumenda esse ea qui distinctio at. Adipisci non quasi consequuntur tenetur recusandae.
Porro corporis iusto perspiciatis et adipisci nam eaque facere. Voluptates et aspernatur quo tempora magnam deserunt. Dolor et repudiandae recusandae error debitis ut ab. Ratione esse quod id ea consequatur. Laborum nihil tempore quia et.
See All Comments - 100% Free
WSO depends on everyone being able to pitch in when they know something. Unlock with your email and get bonus: 6 financial modeling lessons free ($199 value)
or Unlock with your social account...
Commodi hic cupiditate beatae temporibus sit fuga itaque. Enim et quia nostrum illo aut qui quod. Nesciunt eligendi assumenda beatae deserunt.
Ipsam voluptatibus voluptatem hic et qui. Sint earum excepturi libero aut dolorem quia id. Explicabo sed delectus expedita ipsum ut qui. Accusantium sed id doloribus ut quia nemo quis beatae.