Ninety percent of the room raised their hand. The question was who here is building something, and the room was a Claude meetup in Toronto — three hundred people, with four or five hundred more on a waitlist. Lawyers. Doctors. Everyone building something on a weekend.

Chris Grimes and Reuven Gorsht start there and end up somewhere less comfortable: if anyone can build the product, the product is not the business.

They work through what that means for AI startups, for the app layer sitting between the frontier labs and the customer, and for mortgage lending specifically. Chris rebuilt a top-three fitness app in ninety minutes on a Sunday. A tool set that needed Claude Code four months ago is now a feature of the model provider. Half a trillion dollars went into startups in the first half of 2026, and roughly 43 cents of every dollar went to two companies.

If the product can be built over a weekend, the product was never the business. So what is? That is the question the rest of the episode is trying to answer, and in regulated lending it turns out to have a different answer than it does anywhere else.

What you'll take away

• What the frontier labs absorbing the app layer does to the companies funded to sit there
• Why lenders do not shop a vendor list, and what actually gets a vendor into a budget
• Where the margin turns out to be, and why it is not in the AI business
• What makes auditability, rather than capability, the thing that decides who wins in lending AI

Chapters

(00:00) A maxed-out AI week

(01:16) Three hundred people at a Claude meetup

(04:06) A great time and a terrible time to build

(06:08) Why lenders don't shop vendors

(09:29) A product mid-flight, and testing distribution first

(10:46) Half a trillion in funding, and where it went

(12:22) The app layer, and the fitness app built on a Sunday

(16:36) Long Lake, Amex, and margin as the real prize

(20:12) Razor-thin lending margins and the tax agent

(27:00) Fannie, OSFI E-23, and the coming washout

(30:23) The cleaning company paying four thousand a month

(33:17) Start with the problem, not the AI

Mentioned in this episode

Anthropic, whose economic index comes up in the discussion of what people are actually building with these tools

Lovable, the platform used to rebuild a top-three fitness app in an hour and a half

Long Lake, the HOA management company founded in 2023, and the episode's central example of where margin really sits

Amex Global Business Travel, the low-margin operation Long Lake acquired, and the case study Reuven builds the argument on

OSFI E-23, Canada's model risk management guidance, cited as the deadline Canadian lenders are working toward

Fannie Mae, whose AI transparency and traceability requirements took effect recently

The Mortgage Bankers Association, credited on tape for the cost-to-originate figures

Your hosts

Chris Grimes is CEO of FundMore.

Reuven Gorsht is CEO of Deeded and The Variable.

One builds the tool. One absorbs the friction.

Related episode

AI vs. The Human Element— Balancing Trust and Technology — two AI conferences in one week, fifteen vendor demos, and the three questions a lender has to answer that nobody could

Listen and subscribe

Captivate — the-signal.captivate.fm

Apple Podcasts — The Signal

Spotify — The Signal

YouTube — The Signal Podcast

New episodes every Tuesday.

The hardest business to build right now is a business

2

whose product is intelligence.

3

All right.

4

Welcome, Chris, how's it going?

5

What, how's your week?

6

Chris: A lot been going on, but,

I, I maxed out, my Max account,

7

my Enterprise account, and, almost

maxed out my OpenAI account all in

8

the same week, so that's a record.

9

Never…

10

I haven't really hit that token maxing

thing before, but, uh, yeah, it was

11

e- it was either a very expensive

week, depending on how you look

12

at it, or a very productive week

13

Reuven: So you got no,

there's no working AI.

14

how do you deal with that?

15

Chris: I didn't know what to

do with myself at about 4:00

16

on, was it Wednesday afternoon?

17

I just sat there and said,

"Well, what do I do now?"

18

Reuven: Crazy.

19

Well, busy, busy week, I guess.

20

Yeah.

21

Um, on, on my end, um, and, and this

is kinda what we wanted to maybe

22

dive into in, in, in the, in this

episode is, So this was, this wasn't

23

this week, it was actually last week.

24

Time's flying here in the summer.

25

but I went to this, Claude meetup,

it was, it was here in Toronto.

26

About, I think about 300

some odd people showed up.

27

the organizer said they had a wait

list of about 4 or 500 other people

28

that they couldn't let in 'cause just

the venue couldn't accommodate it.

29

wanted to share just a, a couple insights.

30

First of all, it's nice to see so many

people out there, getting interested

31

in AI or, this was specific to Claude,

although I don't think it was sponsored

32

by Anthropic or anything like that.

33

I did a little bit of networking

before, before the event started, and

34

met some very interesting people, and

just very quick observation, it was,

35

uh, you know, you had young people, you

had older people, you had, new grads,

36

you had experienced entrepreneurs.

37

but the common thread is everybody…

38

And I think, when they started

the event, they asked, to raise

39

their hand if they have a company

or they're building something.

40

I think a good, 90% of the room

just, like, raised their hand, and

41

they, they're building something.

42

So I met, a, a young gentleman who

was a CPA, and he's, discovered

43

just through some of his work,

all the manual work that, that's

44

required to close the books, and he's

selling an AI that, that does that.

45

just really all, all sorts of

bu- all sorts of businesses.

46

But, what really got me, Chris, was the…

47

When they started the panel, so they had,

basically three experienced entrepreneurs,

48

or actually two entrepreneurs, one VC.

49

And, the topic was really all about,

what's the state of the market?

50

What's the state a- of AI?

51

And this, question came up, I think

it might have been from the audience.

52

It says, "Is now time

to build an AI company?"

53

And, one, one of the entrepreneurs said,

there's never been a better time to build

54

a company or become an entrepreneur."

55

And, the other one kind of said

something that was really interesting.

56

He said, now is probably the worst

time to build an AI company."

57

and then he went on to describe just,

the saturation and the noise and the, the

58

fact that, even if you look at Anthropic,

they released that economic index

59

analysis, and you see a peak activity.

60

You talk about token maxing during the

week, Chris, like the weekends is when,

61

consumption goes through the roof.

62

So people are doing work.

63

They're, lawyers are vi-

vibe coding, doctors are.

64

Everyone's building something.

65

Chris: It actually, it reminds me of

66

Reuven: that session, it, it was, it was

just it was overwhelming to see how much

67

noise and how much interest there was.

68

Chris: I was gonna say it, it

reminded me of, another episode

69

of another podcast I listened to.

70

And, Andreessen Horowitz was on there,

and he was talking about exactly this.

71

This is not too long ago,

four or six weeks ago.

72

And he said, you know, it's, it's a great

time and a terrible time because what,

73

what the different- differentiator was

for, you know, someone without engineering

74

capacity is now being completely disrupted

because you can use AI to go to market.

75

And then on the flip side, if

you, you know, you have zero

76

coding experience, well, building

software is, is democratized.

77

There's n- there's no one…

78

Like, anyone can do it now, which I,

you know, is, is what you're talking

79

about with, you know, everyone vibe

coding something on the weekend.

80

You know, whether that's building

something as simple as a website or

81

building the next mobile app, it's,

it's at everyone's fingertips now.

82

Reuven: Yeah.

83

I think the panel, you know, they

were asked to provide some advice

84

for the room, and, most of the advice

was about, how do you differentiate?

85

And, y- we hear the co- the constant

themes and for, for our listeners,

86

there are common themes like,

getting proprietary data or the

87

data flywheel or, understanding

the domain, understanding judgment,

88

going to the next frontier, right?

89

and, um, a lot of the, a

lot of those things are…

90

they're great.

91

They're the right,

they're the right advice.

92

But, what really got me is, is really

that, uh, tone shift that, y- any- anybody

93

can, like you said, Chris, anybody can

vibe code something over the weekend.

94

The other thing is, for folks that

are running these software companies,

95

whether if it's an AI native company

or a traditional company, their

96

customers are now vibe coding parts

of their software over the weekend.

97

what have you seen, just curious,

uh, Chris, out there in terms of,

98

companies, starting to commoditize

kind of what you do in your business?

99

Have you seen some of that?

100

Have you seen attempts to do that?

101

Chris: Should probably sort

of declare my stance, I guess.

102

I mean, I'm basically the

guy on the panel, right?

103

I mean, I've been…

104

I'm shipping software, and,

105

you know, I've been doing

so for the last six years.

106

So, you know, if, I guess if your

thesis is right or theirs is,

107

then, you know, I'm gonna become a

cautionary tale in this whole story.

108

M- and, and, and maybe not.

109

But, you know, I think, I think what's,

what, what's been interesting is…

110

And then I'll take the FundMore lens

for a second because it's, it's what

111

we, you know, as, as we do this.

112

But, you know, when…

113

With the lenders we approach, the lenders

we speak to, you know, one thing's clear.

114

Lenders aren't picking a, a vendor,

you know, from, from a list, right?

115

That today, you know, if they're

buying, and I think that's a question,

116

question mark, because I think there is

a little bit of, you know, hesitation

117

or trepidation in the market right

now because there's so much noise.

118

There's so much coming at them.

119

and, and, you know, if they did put a list

together, you know, I would even go back

120

three years, and you would, you basically

have someone in a procurement team.

121

You know, if you're, if you're

sitting there as an executive

122

and said, "I need a…"

123

I use erasers.

124

we need new erasers for the office."

125

They

126

would, would give the, list to

the, to the procurement team.

127

The procurement team would go out and find

all the vendors in the world that make

128

erasers, and they would have their budget,

and they would find the right vendor.

129

And all of a sudden on, you

know, a week later you'd have

130

your erasers in your office.

131

And I know I, I use that in, a little

bit generally, but in, in coming…

132

Because I wanna sort of give an

example of, you know, how these lenders

133

should be thinking about software now.

134

Because, you know,

they're coming at the…

135

They're, th- every day there's a

new AI solution coming into lending.

136

and, you know, the scarcity of it is, is

no longer, is no longer the challenge.

137

And, and building the code

is no longer the challenge.

138

but, you know, I think the one thing

that maybe that didn't come out

139

in that, in that panel, and, and

maybe it did, but, I wasn't there.

140

But it's this, it's the permission

to get access to the money and, you

141

know, to get access to that dollar.

142

And that only comes with, you know, 20

years or 30 years or 15 years of building

143

these relationships, knowing that if you

had to pick up, you know, make a call,

144

someone on the other end's gonna, other

end will actually answer your call.

145

Where if you just built an app to, extract

data from documents, well, frankly, drop

146

it in Gemini, or drop it in Claude, or

drop it in Copilot, and every one of

147

those solutions will give you exactly

what's happening in that document today.

148

It's, that's, it- it's, it's not, there's

no secret sauce there anymore, right?

149

And so, because it's too easy

to access that now, you know, to

150

your point, well, is it secure?

151

You know, is, is the- You know,

what is this, what is this

152

actually being trained on?

153

Because if you think about it for, in,

you know, in, in this mortgage industry

154

that we both operate in, you know,

it's, it's very specific, type of,

155

type of outputs that they need to run

through their, their lending journey.

156

And so, you know, just because you know

the name of the street doesn't necessarily

157

mean it feeds into the workflow later on.

158

And so understanding that at a really

r- rudimentary level and then really

159

deep within that vertical helps as

well when I think you're, you're,

160

you're out in, in market trying

to, to, to promote these things.

161

a-and actually I, I want one more

example and then I'll pass it back

162

to you for a second because it

just, it just popped in my head.

163

And, and you know, I think, you know,

when, when you think about those, those

164

three, those three types of things, and

we actually have a product that's sort

165

of mid-flight right now at FundMore.

166

And w- you know, we're, we're w- before,

before it hits the market, we had already

167

written down sort of, you know, in plain

English what was, what was gonna make

168

this, a defensible asset and how are we

gonna distribute it and where, who, who…

169

What were the relationships in market that

could help us actually access the market?

170

And it wasn't about, you know,

product marketing or the motion

171

or the sales motion behind it.

172

Frankly, it wasn't even

about the product working.

173

It was do we have the relationships

to distribute this at scale?

174

And if we don't and it, and, and

they decide to potentially not

175

back that idea, then we were out.

176

We weren't gonna build it and we

weren't gonna go forward with it.

177

So that's…

178

And, you know, we're sort of

mid-flight in that right now.

179

We've got the prototype because like you

said, we can vibe code it on a weekend.

180

and now we're out in market pitching it

to see if the distribution is there and

181

if it's there, then we're gonna spend

the next eight weeks productizing it.

182

But that's the cycle, right?

183

The, the, the life cycle of that, of,

of, of how we're looking at it now.

184

Reuven: Yeah.

185

No, that's brilliant, and, couldn't

agree with you more, Chris.

186

just the amount of noise, if I look at,

my Facebook feed or Instagram, I start

187

getting all those ads for the, industry

solutions, and I don't think there's been

188

a week where I haven't seen something new.

189

But they're all fairly generic, right?

190

They all have that

generic value proposition.

191

And, maybe just to put some numbers

on just how much we're seeing happen

192

out there is, is, you know, so

global startup funding first half

193

of 2026 hit an all-time record.

194

The number's $510 billion,

so half a trillion, right?

195

when you look under the covers,

it sounds like half a trillion

196

dollars is pouring into startups.

197

Wow, the world is changing.

198

Underneath the covers, about 43 cents out

of every dollar went to two companies.

199

That's OpenAI and

Anthropic, got 217 billion.

200

so it, obviously it's a large amount

of money going out to the rest of

201

the market, probably some, larger and

companies that have a lot more traction.

202

And like you said, Chris, that, that

level of distribution and relationships.

203

but at the end of the day, that,

that noise is only going to

204

accelerate, and it does become a

bit of survival of the fittest.

205

So if I'm, if I'm putting myself

in a buyer's chair, I really gotta

206

ask the question of how many of

these companies are really gonna be

207

around in the next, couple years?

208

Any thoughts there?

209

Chris: I think that's, I

mean, that's the big question.

210

And, you know, I think if you look

at y-your, your point that half that

211

money base has gone to two companies.

212

And, and you can throw Google

in there too, they've just, you

213

know, a lot of that's been in

self-investment, and they don't

214

really count themselves as a startup.

215

But you have these three predominant

players, and I think, you know, f- I

216

was looking at a chart the other day,

and you look at the usage, OpenAI

217

users are still exponentially greater.

218

Then you have Gemini Google

users, then you have Claude users.

219

And these, you know, make up majority

of the people playing with AI today.

220

But I think one thing that's,

that's becoming super clear

221

around this is that they can't…

222

I-if they're gonna take all that money in,

it can't just be about building the next

223

frontier model because the intelligence

stack on that is gonna be limited.

224

And so the next step is building

the app layers on top of it.

225

And you, you know, we're starting,

you started to see that with,

226

you know, the capacity of what

Cowork can do for you today.

227

you know, you can build these

workflows and these agents and these

228

artifacts and these HTML outputs

that, you know, before you, you could

229

never have done that without coding.

230

And even really four months ago,

like we talked about this in the

231

last episode, I built that, team of

agents to support my decision-making.

232

I had to use Claude Code to execute that.

233

Today, I could build it on, with Cowork.

234

And so the, you know, the

h-how easy it is to be able to

235

bring these things to market.

236

And, and the reason why I reference that

is because all those other companies out

237

there that are being heavily funded and,

you know, the level, the Lovables and

238

things like this, that, you know, when

they got released were f- game changing.

239

It was incredible.

240

I could go in there, I could build an app.

241

You know, I, I, I built a…

242

And I did it for fun because I wanted

to learn how it worked, and I built,

243

basically a fitness tracking app,

and it had the whole, the nutrition

244

pieces, it had all the, the, you

know, the, the core metrics you'd

245

wanna track and all these things.

246

And then I looked at it and said,

"Well, this is basically if you

247

go onto your, your Apple or Google

store, and you're gonna g- and you

248

look in like fitness apps and you

just go find the top three," you…

249

I basically replicated that in, in, in

an hour and a half on a Sunday, right?

250

It was…

251

And, but now, and that was on Lovable.

252

Well, today, I could build

that whole thing within Claude

253

or probably, or Anthropic.

254

And, so-

255

Reuven: part of your plan essentially

without paying for another tool,

256

Chris: Exactly.

257

Exactly.

258

And so, you know, all that money

that's gone into these other companies,

259

the question is, where's that,

where's that money going and gone?

260

Because these, these three, leaders

in the space, in order to continue to

261

be at the forefront and driving the

revenue value they're, they're cr- or

262

the revenue they're creating, they have

to continue to build that app layer.

263

And if they're gonna be the ones leading

the app layer, this won't be like any

264

transformation in the past, right?

265

Where, and, and I, I mean, I- you wrote

a really good, post about this on your

266

Substack, I think it was last week, right?

267

Or earlier this week.

268

And, you know, I, I commented on it

because, you know, my, my theory, and this

269

is how much this is all changing, I, you

know, I, I gave some examples about, about

270

history and, you know, I've always enjoyed

history and I talked about a little

271

bit of railway on my comment to you.

272

But it was talking about, you know,

if you look at the top five companies

273

in, in Canada and in the US that

are responsible for, in Canada,

274

connecting coast to coast, and in the

US c- connecting all four corners,

275

at one point they all went bankrupt.

276

And they built the rail, and, it was all

the train companies and, that moved the

277

people that actually made all the money.

278

Now, most, a couple of those companies

still exist, but that's because the

279

government stepped in and huge bailouts

back then to keep, to actually make these

280

into logistics companies effectively.

281

And, uh, but that's, you know, I make

that comment because I think, you know,

282

a week ago I would've said, "Holy jeez,

this is heading in the same direction."

283

And you're seeing these, all this

money being poured into these three

284

or four companies that are building

the, the infrastructure for the

285

pipeline for this next app layer.

286

But, I think what we're seeing, and

I'll, and, and, and I think, you

287

know, you mentioned this to me the

other day, but, you know, OpenAI's

288

finding a way to cut their, their token

cost in half, and, and, or by 20%.

289

And, you know, and, and

like, what's your…

290

Like, h- how do you see this now?

291

Because if they're able to start

giving this away for virtually

292

free, that changes everything.

293

Reuven: Yeah, no, I, I think, I think,

Chris, I'm really changing my lens, right?

294

As I run a services company.

295

Deeded, at the end of the day,

we use AI, but we're pushing real

296

estate files, we're dealing with

brokers, we're dealing with lawyers.

297

Like there's a lot of back and forth.

298

But I- I'm way less excited about all

this technology and the benchmarking

299

and the frontiers, for the same

reason we just talked about.

300

It's a crowded and noisy space.

301

The hardest business to build right now is

a business whose product is intelligence.

302

And I think, power to the people,

they've got tons of money behind them.

303

And I'm sure, there will be winners

and trillion dollar companies

304

that are already emerging.

305

It's, it's fascinating to watch.

306

It's a great, business, business case.

307

It's gonna be one of those lessons in

history, Chris, if we, if we're old

308

enough to quote that at some point.

309

but the easiest leverage right now is,

belongs to those folks that already

310

own a customer, they already have a

workflow, and most importantly, they

311

have a reason for someone to pay, right?

312

So in other words, you have an

existing business, and it could be…

313

And I've got a couple examples,

that I've written in my recent blog

314

that, that are, really traditional

businesses, and you wouldn't think

315

that they're sexy or they're exciting.

316

one company, for example,

Chris, is called Long Lake.

317

started in 2023.

318

Their business is managing

HOAs, homeowner associations.

319

They're…

320

So they're collecting dues, they're

running their board meetings,

321

they're, arranging for landscaping,

all that sort of stuff, right?

322

doesn't sound exciting, but I know it's

a business, that someone's gotta do.

323

and this company, out of nowhere

ac- acquired, something called

324

Amex, Global Business Travel.

325

So Amex, we know Amex for being in, in

the credit card business, but Amex has a

326

huge, global business travel, division.

327

They did about 2.7

328

billion in revenue in 2025.

329

Uh, but their profits have been

very low, so they've had about

330

130 million operating profits.

331

So if you do the math, it's

about, you know, it's a 4.8

332

you know, sub five margin business, right?

333

so Chris, so this company at, Long

Lake that manages HOAs goes and

334

buys this like, massive, giant,

and I think they paid about 6.3

335

billion for essentially a 4.8%

336

margin business.

337

and at, on the surface,

it doesn't make any sense.

338

I think if you kinda do the math,

that's a lot of money for, essentially

339

a very low-yielding business.

340

what was under the covers is that,

Long Lake has really figured out,

341

how to use AI for their operations.

342

So their cost of delivery on that A- HOA

management and all these acquisitions

343

that they've done, basically, i- has 80%

common infrastructure, which now they're

344

rolling into that, Amex acquisition.

345

So they're, in fact, running the business

from, that traditionally runs 0 to

346

5% margin, same thing on the property

management HOA, now they're able to

347

deliver that at over 20% margin, right?

348

I think there's, and there's

several other examples.

349

I think, Chris, you might have a couple

as well that just have figured it out.

350

They've figured it out h- either

by transforming their business or

351

building from scratch, but that's

where the big needle is moving on

352

billions of dollars, obviously.

353

You move the needle even on

one point of margin and you're

354

doing really well, right?

355

Any thoughts there?

356

Chris: Yeah, it doesn't…

357

I mean, it's, it's interesting, right?

358

It, it's, it's not gonna take a lot to

drive real margin in a lot of businesses.

359

I mean, I'm … Maybe I'll pick on lending

again, or mortgage lending in particular.

360

But, you know, margins in this is, a- as

you know, are, like, razor thin, right?

361

Primarily, it's a very

people-intensive process.

362

you know, and, and, and it's not

people working on high-value work.

363

It's not people working on driving

new revenue channels or bringing

364

new money into the business, or

even figuring out how, you know, of

365

the, of the 100 loans that walk in

the door, how do we actually fund a

366

hun- fund 100 of those loans, right?

367

And they're always in the

weeds, so they're funding 20,

368

30%, 35, 40% on a good day.

369

… At the end of it, they're sitting

back and saying, "Geez, our margins

370

are three, four," you know, on a, on

a good day five maybe, perc- percent.

371

And y- you're sitting there

and saying, "How do you really

372

run a business on this?"

373

you know, kind of just echoing what you

said, but if you can get to the point

374

where, you know, you're not spending

the 100 hours on that file, but you're

375

spending six hours on that file, then all

of a sudden, you know, margins change.

376

You know, I, I think, you know, I

think there was an, an example of this

377

with OpenAI in, is it Crete, I think?

378

Something like that.

379

you know, and they, they built a,

effectively a tax agent that, you

380

know, drafts returns and improves

on itself and, you know, you know,

381

across that firm, you know, at least

in the pilot stage, like seven…

382

It was like 7,000 returns

they ran in, in, in session.

383

And, you know, it was north of 97%

accuracy and, you know, the, it's

384

equivalent of, like, one senior accountant

who would spend 180 hours on a tax,

385

files down to something like 15 hours.

386

So, you know, all of a sudden it

doesn't take a team of, a, a, an army

387

of, of accountants to be able to, to

make it through tax season, right?

388

You can, you can run it on one or

two, and you're just as profitable.

389

Or you can be … You, you can decide that

you don't kill yourself during tax season

390

and, you know, you keep your team and

everyone's working a normal working hour.

391

But I think the same thing, you know,

is, is prime for, for the businesses

392

we're in today too is, you know,

how, how do you make that, that one

393

underwriter, how do you make that one

processor, how do you make that one

394

funding specialist or risk officer,

whatever it is, become, you know, way

395

more efficient by re- by reducing that,

that burden of, of, diligence they have

396

to do to be able to make those decisions?

397

Reuven: Yeah.

398

And I think, like part of the concern

that I'm hearing is that, companies

399

have been burned before, right?

400

Let's put it this way.

401

It's, we're not AI isn't anything new.

402

It's another, you know,

CapEx, sometimes OPEX, right?

403

for companies.

404

And there's been a good,

25, 30 years of technology.

405

and you talk about, the moving the

needle on, let's say, cost to originate

406

a loan, and I don't think we've seen

it come down, unless I'm, unless

407

there's something I'm missing, right?

408

I think it's only, it's only gone up, um,

409

Chris: They gone

410

Reuven: the thought was that,

businesses have seen this movie before.

411

We've had different generations

of technology previous to AI,

412

different, loan origination systems.

413

There was, mobile.

414

There was a whole bunch of stuff before

made some pretty bold promises, but

415

I'm not sure that, many businesses

actually saw the results, because the

416

numbers on the cost of originating a

lo- a loan, really haven't changed much

417

Chris: No, they haven't.

418

I mean, they've only gone up, right?

419

Like it's something I, I've, you

know, I've been saying since I really

420

moved into technology around 2018.

421

I think back then it was around 9,000

some odd dollars to take a, a lead

422

to, to fund, you know, through that

process and with all the people that

423

have to touch it and everything else.

424

Today, I think, you know, I

think MBA just released a stat,

425

somewhere north of 12,000.

426

Yet, if you look at all the money that's

been put into supposedly automating this

427

process, AI or not, that somewhere in

there we should have seen that gain.

428

We haven't.

429

It's still, you know, it's,

it's o- it's only gone up.

430

And, you know, maybe with

inflation it's still $9,000.

431

So maybe the reality is, you

know, with cost of money, it's

432

the same cost as it was in 2016.

433

But I s- I'm sure if you looked at tech

budgets, vendor budgets, internal, build

434

teams, you know, and everyone else,

those, those have only, have only grown.

435

It's not like they've shrunk.

436

And, and so we, you know, I'm not

sure where we've been able to get

437

the lift we're-- through the last

eight years that everyone promised.

438

And so I think, I do think this is

where, you know, we sh- we should be

439

able to see the leverage in-- with

AI, and, and the c- and where how

440

agents can come in and actually do

the work, opposed to the past where

441

you had to have a machine and a human

kind of work together to do the work.

442

And so I'm, I'm optimistic or positively

optimistic about this next kind of decade

443

here in lending and, and some of the

opportunities that I think we're gonna see

444

Reuven: so Chris, just, maybe

without being cynical, like

445

what, what has to change, right?

446

So if I'm looking at, Long Lake as an

example, or, there's a company called

447

Current in the US, and, their business is,

they're buying and they're consolidating

448

these accounting practices, right?

449

So they've taken, about 50

CPA practices and consolidated

450

them with one, AI platform.

451

And basically, they have a tax

processing agent that literally

452

refreshes itself every 48-hour cycle.

453

This year they processed

about 7,000 returns.

454

Their average prep time is down 31%.

455

Data accuracy is at 98%.

456

and one accountant's workload

went from 180 hours to 15.

457

we're talking about, meaningfully

moving the needle on doing more,

458

doing the same with far less, and I

think that's their thesis that they

459

continue to, acquire these CPA firms.

460

And, you know, a- again,

I, I wanna, be realistic.

461

We're, we're, we're talking

apples and oranges, tax return

462

to mortgages to lending.

463

But what do you think, what do

you think needs to happen for the

464

industry to have the same model?

465

Like the current the industry, having

someone that really figured it out, that

466

can show tangible numbers and say, "Look,

one loan officer just, was able to, to

467

go from doing 50 files a year to 500."

468

Chris: I mean, I think that question

still needs to be answered, and I think

469

part of that is because the regulators

haven't decided yet how they're fully

470

going to, regulate, I guess is the right

word, um, or oversee these, this sort of

471

next generation of technology with AI.

472

You have, you have Fannie in the US

that made a statement, I think it was

473

about six months ago, and actually

the date was yesterday or Thursday.

474

they had every lender who was

a Fannie lender had to comply

475

to a bunch of, standards.

476

And most of it was around visibility

and transparency and accountability

477

and traceability of that, of what

an- what your AI is doing for you.

478

Which actually comes back to probably

how we started this conversation

479

because it is a super crowded space.

480

Every time you turn around, there's

another company that's evolved because

481

they vibe quoted something in their

basement, and they're out pitching lenders

482

saying, "Hey, this is your solution."

483

The problem is, is that they've

tr- they're using these large

484

language models as the train…

485

as, as what they were trained on

to be able to execute that task.

486

Chris-AI: without having access to

the real data and the, and the real

487

structure then of, of how these loans

actually work through a process, it's

488

very difficult, I, I think, to actually,

um, adhere towards what Fannie asked.

489

And I think in Canada, we saw…

490

we're seeing the same thing with OSFI

E-23 just coming up in a, in a few weeks

491

where, you know, version one of that's

landing as well, and it's giving, you

492

know, lenders some guidance, again,

around visibility and transparency

493

around what their AIs have to be a…

494

Chris: they have to be auditable.

495

and again, if you're relying on a large

language model as your source of, of, Of,

496

of, of work, well, it's not that open.

497

I, I don't know how I, I, I can…

498

I mean, I'm like everyone else.

499

You, you can go read about what, you know,

Anthropic or OpenAI have done to go and

500

train their models, their frontier models,

and but that's not giving me the…

501

And, and I can ask it how it sort of

came to it, but it's not really open.

502

So, you know, I, I'm, I'm not really

answering the question because I think

503

there's a big question mark right now.

504

I think three weeks ago or six months ago,

it would've been easier to answer that.

505

today, I think it, I think there

has to be a little more caution put

506

into this, specifically for mortgage

lending, because of, of how, you

507

know, they're, the, the, the lenders

are gonna be governed, around that.

508

so you know, again, I will sort of

toot the Fundmore horn here a little

509

bit or, or others in this space, but,

you know, having the relationships,

510

having the clients, like you said

earlier, having, the technologies

511

is, is, is easier to come by.

512

But then figuring out how you

actually take all that and make

513

this truly transparent in an

AI world is gonna be the next,

514

quote-unquote, "frontier of lending."

515

And I think the companies that figure

that out first, like how you do that,

516

whether that's, a small model, a,

or a, or, or a large language model

517

that's purely trained, you know, on, on

whatever, on, on lending as an example.

518

Those are gonna be the ones that

actually find a path forward, I think.

519

And I really believe there's gonna be a

washout in this, in this market because,

520

because the regulators are going to

want governance that most of these small

521

companies aren't gonna be able to provide

522

Reuven: Yeah, no, for, for sure, Chris,

and it sounds like, again, full circle

523

back to all the noise and the tug of war.

524

There's always that tug of war

between, what's right, what's

525

regulated, what's gonna be around

in years, in 10 years, right?

526

So we've always gotta think that or have

that in very high consideration because

527

build cost's super low, next to nothing.

528

Token cost, dropping like a fly.

529

that intelligence is, not, let- let's,

let's call it a non-obstacle, although

530

as you pointed out, there is, there's

the training and, and, and judgment and

531

taste and all the, all, all that stuff

that needs, interrelationship matters

532

that need to be, inserted in there.

533

but nonetheless, that, that

moat has really shifted.

534

So whether you're looking at it from

a, software vendor perspective, or if

535

I'm looking at it from a services lens,

what gets me excited is, now you've

536

got these traditional businesses that

have traditionally had margins erode,

537

accounting, parts of legal, homeowner

association management, um, um, to

538

a guy this week that was running a

cleaning company, 17 cleaners, and

539

was just about to quit the company

because just couldn't take it anymore.

540

He's working, 18-hour days trying to

manage the scheduling and all that stuff.

541

he's paying $4,000 a month, for an AI

agent that's now essentially managing

542

all that appointment booking, all the

back office, all the rescheduling, all

543

that stuff, and he's actually enjoying

his business and he's growing it, right?

544

it- it's, there's a lot of magic

there, but the magic, you know, I

545

think we would agree, Chris, happens

in the application of the technology.

546

you, again, you won't see me

excited and jumping for joy

547

about the next frontier model.

548

Those companies are great.

549

love the innovation.

550

I get surprised every time.

551

what matters is can I get my loan quicker?

552

Can I get my office clean?

553

Can I remove friction versus adding

more friction, just because I wanna,

554

I wanna add technology, I wanna be,

or I wanna take the risk of being

555

not compliant with some of these

regulations that are coming out, right?

556

Chris: Yeah, no, I think that's,

I, I, I think that's well said, and

557

I think that's how I see it too.

558

I mean, may-maybe just before, you know,

we wrap up, 'cause you did ask me the

559

question, maybe I'll just put it back

to you because I think, you know, maybe

560

you you have a, a different

lens on it, a little bit.

561

Like, like if you were sitting in

front of a sales team at, at a, a large

562

institution today, and you're like,

"Okay, we have all these loan officers.

563

They're producing on average,

you know, 10 loans a month.

564

I have another guy

doing 50 loans a month."

565

You know, like, what, what would you

suggest to even bridge the gap there?

566

Because, you know, like if

you had everyone running at

567

50, you'd be in great shape

568

Reuven: it's fu- funny enough, I had the

conversation this week, so it's fresh

569

on my mind and i- what I'm finding is,

it all started with an AI conversation.

570

Let's talk about AI.

571

I said, the first question was,

like, "Okay, what do you do

572

today that's actually effective

and working, and what do you do

573

that's a complete waste of time?"

574

Because you're in ef- if you

wanna get efficiency, it's, yes,

575

technology's gonna be the tool and

enabler, a lot of teams are still,

576

they're still doing legacy practices

for either, managing relationships

577

or touching base with clients.

578

So my first question to this person

that I was chatting with, I said,

579

what are you really good at?"

580

And he said, I'm good

at talking to people."

581

your answer is not AI.

582

Your answer is, just go

talk to more people, right?

583

At the…

584

And, and but obviously, that's a kind of

a simple, on the surface answer, Chris.

585

But you start with that problem in mind

to say, "Okay, so my, my, my superpower

586

is going to talk to more people.

587

How do I free up enough time?

588

Because I know I've got all this time

in the back office, whether if it's

589

paperwork or getting stuck between

portals and systems and complexity

590

and regulation and compliance."

591

and that's where we started to erode at

some of these bottlenecks versus, touching

592

on the, the, the, the, what he thought was

the problem was, like, I'm not using AI.

593

What we really came, came up with is,

you gotta find ways to free up more

594

time so you can talk to more people.

595

And when you talk to more people,

if you're, if you're good at it,

596

you're gonna get more business

and you're gonna do more loans.

597

Chris: It's actually, um, you

know, a, a, a good point, and

598

it's something we actually use

in our AI lending labs, right?

599

When we bring, you know, these executives

together in a room, when w- the, you

600

know, a- after, you know, we give the,

the sort of the market update on where

601

AI is and we group them, group them,

the first ses- the first 45 minutes

602

is let's just dig into that problem.

603

Before we ask AI to solve it, let's

dig into the problem and then, you

604

know, collectively amongst peers, we

get an incredible sort of challenge

605

that we put into, you know, in, in

AI, you know, Claude Cowork or, or

606

OpenAI and, and really try to build

out the, the workflows from that.

607

But you're right.

608

It starts with that problem and,

and if you can identify it, and you

609

can figure out then, you know…

610

Then, then you can start figuring

out what the solutions are and how

611

you actually get to, to the outcome.

612

Reuven: Yeah, and what I'm finding

is just, just maybe as some closing

613

thought, Chris, is that the initial

problem that is articulated when you

614

start digging and peeling the onion and

getting, a couple layers deep, probably

615

about 80% of the time that is not the

real problem, or it's just a symptom

616

that they're seeing in the business.

617

I'm not closing enough business.

618

I'm not I'm behind my, my,

uh, my target for the year."

619

And when you start looking

under the covers, there's a lot

620

more that sometimes leads to a

completely different issue, right?

621

So it's important to, for anyone

listening, not to peg that and jump into,

622

jump to conclusion to say, my problem

is I'm not growing because I'm not using

623

AI, because Chris is using AI in his

company, and I'm, I'm now the laggard."

624

The solution really is to start doing

some, working on the business and

625

thinking about, what is the real problem?

626

What is that root cause?

627

And sometimes it's gonna be an AI fix,

which, you know, we all talk about with

628

the popular thing to do, or sometimes it

just may be a process that's broken, and

629

it's been sitting there for 15 years, and

you just haven't felt the pain until now.

630

Chris: Said just like every

executive coach I've ever had.

631

Reuven: Perfect.

632

Chris: Yeah

633

Reuven: with that, I think, we wrap

up this episode and, you and I will

634

be on the road next week, and, uh,

we'll look forward to bringing you

635

some more insights from the industry.

636

Always open to feedback, so s- drop

us a line, let us know what you

637

think, and, uh, talk to you next week.

638

Chris: And maybe just before we

go, since I did mention it, we

639

do run these AI lending labs.

640

If you're a lending executive and you know

you're new to AI or you just want to know

641

what's next and actually put your hands

on keyboards to actually build out agents,

642

you know, happy to have you join us.

643

We're running two of them.

644

We have one coming up in Calgary

later in October, and we have one

645

in Orlando in the US, in November.

646

So look forward to

seeing you there as well