PERE 2026 Interview with Robin Langdon

6:51 · English · Acquisitions

Watch this video

Enter your details to access the library. We'll email you a code to verify your address.

See how we use your data in our Privacy Policy.

About this video

"The model is not the product"

Full description

Olivia from PERE interviews Robin Langdon from 5D regarding the benefits of using a vertical platform rather than relying on the big lab models. The theme is "the model is not the product" and companies assessing AI technologies should be benchmarking the the overall enterprise platform approach.

Transcript

Robin, thanks for joining us at Tokyo Forum. It's great to have you with us. Yeah, thanks Olivia. Nice to be here. So, I'll get straight into it. Two weeks ago, Greg Brockman, the president of OpenAI, said the model alone is no longer the product. He called it model harness product symbiosis. What does that mean for institutional real estate? There's been a common mis misconception really from industry and and the whole of real estate that maybe AI the big model labs like OpenAI like Anthropic are all going to just create some kind of magical solution that will solve everything and we just have to wait you know tomorrow or next week for that to happen and and really the quote from Brockman it was really saying that they are just part of the solution and really for solving some of the institutional enterprise problems like you know commercial real estate In an investment management side, you really have to combine the underlying model with the the the the boundaries and the security, the guard rails, the reasoning logic of a harness around the top.

And what he means by harness is that that that control layer, the intelligence layer that controls what the model does. So essentially what he's saying is it's if you compare it to commercial real estate, it's similar to saying yes, the model is like having a junior analyst that's working for you. and and the AS is very good at doing some of the underlying grunt work, but what you really need is that kind of senior analyst coming in who's able to do the the reviews to check to approve.

And that's where this harness he's talking about comes in to do that. And so he's talking about really a couple of different areas there. is saying well there can be errors coming out of of our model which can be there can be different you know inconsistencies coming through so you need this harness layer to be able to have checks and balances to be able to say okay I I need to go and test this and evaluate whether that was the right answer coming through he's also saying things like the the consistency the reasoning the audit trail all of these are super critical for especially for in the investment industry it's highly regulated and so you need to be able to show the how did AI come to a certain answer. so show the reasoning logic and to be able to show a complete audit trail citing you know right back to a a source document.

Yeah. so if there's ever a question from a regulator for example saying hey you know where did you get this particular answer from saying yes that came from clause 16.2 of this particular contract and that's what the audit trail would actually look like. It would demonstrate exactly where each thing has been cited from. Exactly. Yes. And I think a lot of people today are playing around with some of the the standard, you know, models out there. Yeah.

And they're they're attaching a, you know, a contract to a to a ChatGPT session or a Copilot. and they're getting a great answer from just, you know, from one document thinking, "Oh, this is pretty cool." But actually, if they then, you know, subsequently want to come back and go, "Well, actually, how did I get that answer?" And they try it again, maybe they might get a different answer. you do find a lot. So the consistency is not there. and so you need to have that that vertical solution in place where you have the harness around the model which is really controlling what it's doing. and providing that that consistency in there.

And how is Ellie different from using ChatGPT? Yeah. So Ellie is is really providing that that kind of senior analyst type analogy. So Elias, you know, can sit on top of lots of different models and Ellie also has some proprietary technology as well for being able to do highly accurate extraction of insights from unstructured data. So there might be a you know a 400page you know contract where you want to extract out data and you know a Copilot or a ChatGPT might be 60 70% accurate in in doing that is benchmarked being well over 98 and a half% accurate for that.

So that can mean millions in terms in terms of you know mistakes and loss revenue when you're trying to do a a big deal or manage a portfolio a set of assets going through and Ellie does really provide that all of the guard rails so all of the data is completely sandboxed you know Ellie's not sending out data to the big models for training purposes it remains institutional data and and that data is the moat you know for a for a lot of you know these CRE investment firms as well.

So, do you think that horizontal models are going to become good enough that the vertical AI will become irrelevant? Yeah, it's a good question. I think and that's what a lot of people are sort of almost waiting for. They're going ah no problem. Yeah, this is going to be fine tomorrow. They're going to invent something. And I think that's why Brockman's quote was so striking when he said actually guys don't wait for the horizontal model to to solve everything.

It's not there needs to be the vertical model which provides the harness. and that's really the buzzword this year. It's about the harness which is really just the control layer across those horizontal models about how they work. providing the the the kind of the the review mechanism, providing the working papers, providing the check sums and and balances, providing some of the mathematical models of doing things like you know variance analysis and complex you know five-year discounted cash flow models that you know some of these models like a Claude might struggle with and that's where that vertical piece comes in to handle it and and Ellie is just completely trained on commercial real estate so you don't have to explain a huge prompt to Ellie to say, "I need you to do this model and I need this, this, this, and this, and this." Ellie knows it.

Ellie straight away say, "Hey, I need a DCF model." Ellie goes, "Great. I know that, and I know where your data is. I've got the context. I know what your what your assumptions are for previous models." And Ellie just comes out straight away with the answer. That's amazing. And where do you think the gap will widen over the next 24 months? the companies that start to adopt AI now properly, they're going to have a huge advantage moving forward. And it's really about a lot of a lot of companies we've been talking to are saying, "Oh, we're benchmarking the various models and look and playing with different models and see where they go." And they're really approaching the problem in the wrong way. it's not about benchmarking the model itself.

It's about benchmarking the the overall vertical solution and the organization because the organization themselves have to transform. and so those companies that start to do that now and transform themselves as a business are really going to be in the top 15% of, you know, investment firms down the line. And and you know, really that means they're going to be the first to get deals. They're going to have the greatest insights in coming through and they're just going to be leaving behind a lot of the competition who just haven't been focusing and they're just purely waiting for the horizontal models to take off. Well, it's been really great to talk to you today. Really appreciate your time. Thanks, David. Thanks so much.

Fifth Dimension for deal teams

From data room to cited IC memo

Video details
Roles
Owner, Asset Manager, Portfolio Manager, Investor / LP, Lender, Broker, Compliance / Risk
Processes
Acquisitions, Asset Management, Due Diligence, GP Due Diligence, Lease Abstraction, Loan Abstraction, Portfolio Analytics, Risk & Compliance, Valuation
Geographies
APAC