Fireside Chat — The Agent’s GPS: Causal AI in a Post-Pattern World

1385
04:00:24.360 –> 04:00:27.850
Julia Nimchinski: Welcome back, Tuba! Long time no see!

1386
04:00:29.910 –> 04:00:35.559
Julia Nimchinski: So this is gonna be a battle of the mind. Are we debating? Are we chatting? What are we doing?

1387
04:00:36.360 –> 04:00:42.469
Mark Stouse, Proof Analytics: I can tell you that my earlier conversation with Tuba was pure magic.

1388
04:00:43.420 –> 04:00:43.930
Mark Stouse, Proof Analytics: I have…

1389
04:00:43.930 –> 04:00:49.460
Julia Nimchinski: I have the auto recording, so… My agent was there.

1390
04:00:49.590 –> 04:00:51.200
Julia Nimchinski: Let’s dive into it.

1391
04:00:55.180 –> 04:00:57.929
Julia Nimchinski: So what, take us away. Yeah, let’s dive into it.

1392
04:00:58.470 –> 04:00:59.260
Mark Stouse, Proof Analytics: Awesome.

1393
04:00:59.810 –> 04:01:03.999
Tooba Durraze: Awesome. Maybe we can start with a quick round of introductions. Mark, if you want to go first.

1394
04:01:04.520 –> 04:01:19.189
Mark Stouse, Proof Analytics: Sure, I’m Mark Stuce. I am the chairman and CEO of Proof Causal Advisory. We do a… well, we focus entirely, pretty much, on, causal AI and what it does for Cohens.

1395
04:01:20.160 –> 04:01:31.490
Tooba Durraze: Awesome, and then for everyone in the room, Tuba, founder and CEO of Amoeba AI, and we purely work on a neuro-symbolic AI architecture, so I’m very excited for this conversation, because

1396
04:01:31.620 –> 04:01:42.830
Tooba Durraze: I know we’ve been talking a ton about agents, et cetera, and a lot of that concentrated on a transformer-based architecture, so now we have two people in the room who are working on an architecture that’s a little bit different.

1397
04:01:42.840 –> 04:01:51.709
Tooba Durraze: So let me throw it to you first, Mark. Maybe you can describe, in plain language, what causal really means to folks in the room.

1398
04:01:53.390 –> 04:01:57.270
Mark Stouse, Proof Analytics: So, I think that it really helps to say this, right? That…

1399
04:01:57.890 –> 04:02:03.290
Mark Stouse, Proof Analytics: Most of what we call AI in business today is just correlation.

1400
04:02:03.520 –> 04:02:06.110
Mark Stouse, Proof Analytics: And I think, you know, we all…

1401
04:02:06.790 –> 04:02:21.489
Mark Stouse, Proof Analytics: remember Statistics 101, maybe it didn’t make you feel all that great or smart or whatever, but you probably remember, anyone listening remembers, correlation does not imply causality. And…

1402
04:02:21.970 –> 04:02:24.210
Mark Stouse, Proof Analytics: We all kind of wondered…

1403
04:02:24.450 –> 04:02:37.100
Mark Stouse, Proof Analytics: whenever that was, like, okay, that’s kind of interesting, but how is it really relevant? Well, we are now entering a time when it becomes not just highly relevant, but

1404
04:02:37.740 –> 04:02:48.210
Mark Stouse, Proof Analytics: a significant part of almost everything that we do, particularly in business, right? So, correlation shows us patterns.

1405
04:02:48.560 –> 04:02:51.110
Mark Stouse, Proof Analytics: In the rearview mirror, right?

1406
04:02:51.570 –> 04:02:57.420
Mark Stouse, Proof Analytics: And… That’s kind of a problem today, because…

1407
04:02:57.550 –> 04:03:04.889
Mark Stouse, Proof Analytics: Everyone can feel and see the marketplace volatility, the speed with which things are changing.

1408
04:03:05.440 –> 04:03:07.600
Mark Stouse, Proof Analytics: what does that mean? It means that…

1409
04:03:07.700 –> 04:03:14.519
Mark Stouse, Proof Analytics: past isn’t prologue anymore. To the extent that it ever was, it’s definitely not now.

1410
04:03:15.270 –> 04:03:22.370
Mark Stouse, Proof Analytics: And so, we’re seeing… the patterns… that were…

1411
04:03:23.100 –> 04:03:29.970
Mark Stouse, Proof Analytics: identifiable, even recently, becoming separated.

1412
04:03:30.230 –> 04:03:43.480
Mark Stouse, Proof Analytics: faster and faster and faster from today’s reality and tomorrow’s probabilities, and correlation just can’t keep up with that, but causation or causality can.

1413
04:03:44.290 –> 04:04:03.240
Tooba Durraze: Yeah, I completely agree with you. I think… I think we’re in a world where the future looks very different from what the past was, so a lot of our historical signals, etc, which we’re used to seeing in traditional go-to-market functions, like forecasting, etc, maybe don’t apply in the same way. So, let me ask you, where do you see

1414
04:04:03.330 –> 04:04:15.219
Tooba Durraze: If we talk about Gen AI, and specifically, like, how folks are using LLMs now, where do you see that struggling the most in go-to-market motions, and how does causal AI maybe fill in that gap?

1415
04:04:16.620 –> 04:04:19.950
Mark Stouse, Proof Analytics: Well, I mean, an LLM, an SLM…

1416
04:04:20.230 –> 04:04:29.210
Mark Stouse, Proof Analytics: even traditional BI, but in a different way, right, is built on correlation, on pattern match.

1417
04:04:29.310 –> 04:04:31.330
Tooba Durraze: Not causation, so…

1418
04:04:31.820 –> 04:04:37.369
Mark Stouse, Proof Analytics: It can tell you what usually happens, but it can’t tell you why it’s happening or what to do about it.

1419
04:04:39.230 –> 04:04:45.509
Tooba Durraze: You can pretend to know why it’s happening, and that’s something, actually, to be very aware of for the folks in the room.

1420
04:04:45.830 –> 04:04:51.310
Tooba Durraze: Just because it doesn’t know, doesn’t mean it doesn’t say it like it does know, essentially.

1421
04:04:52.320 –> 04:05:02.449
Mark Stouse, Proof Analytics: Yeah, I mean, one of the… one of the big challenges is getting just a little bit technical, with Gen AI, right, is what’s called regression to the mean.

1422
04:05:03.010 –> 04:05:05.830
Mark Stouse, Proof Analytics: Right? And so, the mean is the average.

1423
04:05:06.330 –> 04:05:14.150
Mark Stouse, Proof Analytics: And so everything tends in a correlation-based, machine learning-based kind of system.

1424
04:05:14.250 –> 04:05:17.779
Mark Stouse, Proof Analytics: to regress to the mean. And so…

1425
04:05:18.420 –> 04:05:30.720
Mark Stouse, Proof Analytics: Why that… why is that bad? Okay, well, it averages out everything, that… that’s kind of somewhat of a problem for a lot of people, but also, if we look at…

1426
04:05:31.090 –> 04:05:38.200
Mark Stouse, Proof Analytics: How much of the time The average, or the mean, represents reality.

1427
04:05:38.400 –> 04:05:43.200
Mark Stouse, Proof Analytics: It’s… it can… in an open system, which we’re all very much operating in.

1428
04:05:43.330 –> 04:05:45.260
Mark Stouse, Proof Analytics: It could be, like, 2%.

1429
04:05:46.220 –> 04:05:49.259
Mark Stouse, Proof Analytics: Right? So it’s not really reliable.

1430
04:05:49.450 –> 04:05:50.930
Mark Stouse, Proof Analytics: Even as guidance.

1431
04:05:51.500 –> 04:05:52.090
Tooba Durraze: Yeah.

1432
04:05:52.590 –> 04:06:03.640
Tooba Durraze: No, absolutely. And again, the intention here is not to say that LLMs aren’t great at certain things, but I think understanding what the purpose is is really important, because

1433
04:06:03.640 –> 04:06:15.609
Tooba Durraze: you can apply, when we talk about AI these days, everyone kind of conflates that with us talking about GenAI and, like, a subset of Gen AI, which are LLMs. So I think it’s important to understand that these tools are great.

1434
04:06:15.610 –> 04:06:32.070
Tooba Durraze: But you have to know where to apply them, where they’re good, instead of kind of using them as a multi-purpose tool, essentially. So there’s a concept that you talk about a lot, which is that… the GPS, essentially. Can you dive into that a little bit? Because…

1435
04:06:32.370 –> 04:06:51.339
Tooba Durraze: for me as well, I’ve always looked at, like, neurosymbolic AI as, like, a compass, meaning it’s directional, and folks are used to treating, whether it’s business intelligence, whether it’s numbers, analytics, etc, as, like, a very precise science, which I don’t think is the case. So, curious to hear from you on your concept around the GPS.

1436
04:06:51.950 –> 04:07:05.010
Mark Stouse, Proof Analytics: Yeah, it’s definitely not the case. That said, it’s also not… Just completely laissez-faire and, and…

1437
04:07:05.700 –> 04:07:11.689
Mark Stouse, Proof Analytics: you know, only generally representative either, right? There’s… it’s kind of a balance, and so…

1438
04:07:11.810 –> 04:07:16.400
Mark Stouse, Proof Analytics: We talk a lot about fit-for-purpose levels of accuracy.

1439
04:07:16.450 –> 04:07:18.769
Tooba Durraze: You know, if it’s life and death.

1440
04:07:18.770 –> 04:07:22.279
Mark Stouse, Proof Analytics: Then you’re gonna want a very high confidence score.

1441
04:07:22.470 –> 04:07:37.829
Mark Stouse, Proof Analytics: If you are doing something else, marketing, go-to-market, comes to mind. You know, not only are you not going to get a high confidence score, because it’s almost all human behavior related.

1442
04:07:38.040 –> 04:07:41.150
Mark Stouse, Proof Analytics: But also, you know, you’re just gonna…

1443
04:07:42.120 –> 04:07:49.900
Mark Stouse, Proof Analytics: You’re far more interested in how it’s iterating than it… what it is at a particular moment.

1444
04:07:50.980 –> 04:07:58.029
Mark Stouse, Proof Analytics: And I think that that’s what a lot of people are really getting at when they talk about the idea of being directional.

1445
04:07:58.440 –> 04:07:59.309
Mark Stouse, Proof Analytics: Yeah. You know?

1446
04:07:59.430 –> 04:08:05.070
Mark Stouse, Proof Analytics: it’s… It’s a GPS in the sense that…

1447
04:08:05.310 –> 04:08:11.499
Mark Stouse, Proof Analytics: Right? Actually, it’s literally a GPS, the math is extremely similar.

1448
04:08:11.500 –> 04:08:13.979
Tooba Durraze: Yes, I was going to laugh and say that, actually.

1449
04:08:13.980 –> 04:08:23.439
Mark Stouse, Proof Analytics: Yeah, no, it is. It’s why it’s the analogy that never breaks, right? Yeah. And so,

1450
04:08:24.020 –> 04:08:37.659
Mark Stouse, Proof Analytics: you know, you’re… you’re… you think about a… the GPS on your phone, it’s… it’s going… it knows where you are, you tell it where you want it to go, or where you want to go, and it’s going to show you

1451
04:08:38.100 –> 04:08:41.380
Mark Stouse, Proof Analytics: two or three options. You’re gonna pick one.

1452
04:08:41.510 –> 04:08:50.890
Mark Stouse, Proof Analytics: And critically, and this is the direct compare, it’s going to track what you do, what you control, namely the car.

1453
04:08:51.340 –> 04:08:56.530
Mark Stouse, Proof Analytics: And it’s gonna track a whole bunch of factors you do not control.

1454
04:08:56.700 –> 04:09:02.569
Mark Stouse, Proof Analytics: but that either help or hurt your journey, right? Speed it up, slow it down.

1455
04:09:02.910 –> 04:09:03.680
Tooba Durraze: Yeah.

1456
04:09:03.680 –> 04:09:09.519
Mark Stouse, Proof Analytics: Could easily, you know, and this is where you get to the whole idea that just because

1457
04:09:09.660 –> 04:09:16.719
Mark Stouse, Proof Analytics: The root to value, in this case, was… correct and optimized.

1458
04:09:16.720 –> 04:09:17.120
Tooba Durraze: Hmm.

1459
04:09:17.120 –> 04:09:22.610
Mark Stouse, Proof Analytics: in January does not mean that that is still it in September.

1460
04:09:24.260 –> 04:09:42.159
Mark Stouse, Proof Analytics: And so you’re gonna, you know, a lot of times you have to reroute. I was, I was with a CMO of… it’s a customer of ours, not, you know, a couple years ago, and he was up on stage, and he said, oh, the thing I love about proof

1461
04:09:42.680 –> 04:09:46.600
Mark Stouse, Proof Analytics: is that… it means I’m never wrong.

1462
04:09:47.480 –> 04:09:48.799
Mark Stouse, Proof Analytics: And of course, I just…

1463
04:09:48.930 –> 04:09:57.560
Mark Stouse, Proof Analytics: pretty much had a heart attack right there, right? And so, you know, when he came off the stage, I’m like, dude, that was really cool, but…

1464
04:09:57.700 –> 04:09:59.900
Mark Stouse, Proof Analytics: You know, what did you mean?

1465
04:09:59.900 –> 04:10:00.560
Tooba Durraze: Yeah.

1466
04:10:00.740 –> 04:10:03.899
Mark Stouse, Proof Analytics: You know, and he’s like, well, think about it for a second.

1467
04:10:04.010 –> 04:10:10.890
Mark Stouse, Proof Analytics: He goes, If, because of the readouts improved, and a GPS,

1468
04:10:11.370 –> 04:10:16.769
Mark Stouse, Proof Analytics: I’m never wrong, I’m just endlessly corrected.

1469
04:10:18.100 –> 04:10:20.460
Mark Stouse, Proof Analytics: and I get to my destination.

1470
04:10:20.800 –> 04:10:33.679
Mark Stouse, Proof Analytics: with an explainable variance, if I’m late, explainable variance, guys, means you’re… you’re not worried… you didn’t nail it, but you have a good reason that’s documented for the fact that you’re, in this case, late.

1471
04:10:33.940 –> 04:10:37.609
Mark Stouse, Proof Analytics: Right? And he’s like, so, how is…

1472
04:10:37.730 –> 04:10:40.490
Mark Stouse, Proof Analytics: How am I wrong in saying I’m never wrong?

1473
04:10:41.210 –> 04:10:53.570
Mark Stouse, Proof Analytics: And I’m kinda like, Well, factually, that’s probably pretty much true, pretty much defensible, right? It just doesn’t sound…

1474
04:10:53.970 –> 04:10:57.150
Mark Stouse, Proof Analytics: Right. You know, so we had a good laugh about it.

1475
04:10:57.510 –> 04:11:08.250
Tooba Durraze: Yeah, like, I would say, you can get from A to B, like, you can walk from A to B, as well as drive from A to B. You’re still getting the result where you’re reaching your destination, but

1476
04:11:08.250 –> 04:11:21.970
Tooba Durraze: reaching it in the right way, in the most precise way, in the most optimized way, rather, is, like, the thing. Yeah, so just because you took a path and you got to your end destination doesn’t mean that was, like, the optimal path for you to take, essentially.

1477
04:11:22.350 –> 04:11:23.190
Mark Stouse, Proof Analytics: That’s right.

1478
04:11:23.350 –> 04:11:33.760
Mark Stouse, Proof Analytics: That’s right. And that is actually, quickly, very, very relevant right now. For the last two and a half years since Delaware’s Chancery Court

1479
04:11:33.940 –> 04:11:39.030
Mark Stouse, Proof Analytics: Made a major new ruling on fiduciary duty, spreading it out.

1480
04:11:39.390 –> 04:11:47.770
Mark Stouse, Proof Analytics: to all corporate officers and lowering the threshold for breach considerably, right?

1481
04:11:47.900 –> 04:11:56.110
Mark Stouse, Proof Analytics: the ability to say, I am managing my function iteratively.

1482
04:11:56.310 –> 04:12:06.200
Mark Stouse, Proof Analytics: And always with an eye towards… Minimizing risk, And maximizing shareholder value.

1483
04:12:06.630 –> 04:12:16.160
Mark Stouse, Proof Analytics: That’s exactly where not only That the… the calculation of causality matters.

1484
04:12:16.650 –> 04:12:22.560
Mark Stouse, Proof Analytics: The ability to iterate it at the clock speed of the business is super important.

1485
04:12:23.050 –> 04:12:36.659
Tooba Durraze: Yeah, it’s not a nice-to-have anymore. And, like, this is… again, the intention is not to scare the people in the room who are listening to this, but I think, like, as a founder, like, I’m constantly thinking about my fiduciary responsibility in that sense.

1486
04:12:36.660 –> 04:12:44.119
Tooba Durraze: And I’m constantly optimizing, but sometimes when executive teams get a little bit bigger, a little bit bloat gets introduced into the company.

1487
04:12:44.570 –> 04:12:59.559
Tooba Durraze: you think about, like, okay, it’s fine if I miss my target by X percent in quarter one, I’m gonna make up for it in quarter two, and then that can have a compounding effect, because you can’t presumably predict very well on what’s gonna happen in quarter two.

1488
04:12:59.580 –> 04:13:07.319
Tooba Durraze: So the idea of, like, having a system that optimizes consistently against your business goals is…

1489
04:13:07.410 –> 04:13:11.650
Tooba Durraze: Very important, not just, from the business perspective, but also from the

1490
04:13:11.780 –> 04:13:23.689
Tooba Durraze: fiduciary responsibility perspective. So, given that, I’m curious on, your thoughts on agents. Now, we’ll define agents right now as

1491
04:13:24.750 –> 04:13:43.299
Tooba Durraze: conducting some sort of, like, autonomous flow using, like, LLMs, instructions, etc. So what are your thoughts on how folks are thinking about agents these days, especially because there’s a lack of observability on, like, the paths the agents would take, because they’re LLM-based, non-deterministic?

1492
04:13:45.430 –> 04:13:50.110
Mark Stouse, Proof Analytics: Yeah, I mean, I think that the problem here is, is that everyone…

1493
04:13:50.840 –> 04:13:54.540
Mark Stouse, Proof Analytics: By the way, this also goes for Gen AI in general, right?

1494
04:13:54.540 –> 04:13:55.040
Tooba Durraze: Yep.

1495
04:13:55.490 –> 04:14:00.700
Mark Stouse, Proof Analytics: Everyone’s really focused On efficiency and productivity.

1496
04:14:01.510 –> 04:14:03.989
Mark Stouse, Proof Analytics: Not being effective.

1497
04:14:04.570 –> 04:14:05.460
Mark Stouse, Proof Analytics: Right?

1498
04:14:05.500 –> 04:14:09.239
Tooba Durraze: And so, when you program an agent right now.

1499
04:14:09.360 –> 04:14:15.560
Mark Stouse, Proof Analytics: It’s totally correlation-based. It’s essentially… this is a little harsh, but not by much.

1500
04:14:15.720 –> 04:14:21.150
Mark Stouse, Proof Analytics: It is, it’s very sophisticated automation.

1501
04:14:21.920 –> 04:14:22.520
Tooba Durraze: Yeah.

1502
04:14:22.870 –> 04:14:24.330
Mark Stouse, Proof Analytics: So…

1503
04:14:24.490 –> 04:14:34.400
Mark Stouse, Proof Analytics: The problem is, is that in an ever-changing, highly volatile, open system, like the marketplace, and like your response to the marketplace.

1504
04:14:34.750 –> 04:14:42.880
Mark Stouse, Proof Analytics: The instructions that you are giving your agents are almost immediately dated.

1505
04:14:45.290 –> 04:14:52.239
Mark Stouse, Proof Analytics: And if you can’t have a GPS on board effectively with every agent.

1506
04:14:52.500 –> 04:14:55.980
Tooba Durraze: Yeah. Will not know how to…

1507
04:14:56.050 –> 04:15:05.070
Mark Stouse, Proof Analytics: compensate, right? So you’ll start getting… by the way, this is one of the major symptoms of…

1508
04:15:05.260 –> 04:15:13.789
Mark Stouse, Proof Analytics: the failures that are being reported on all this is that the agents are kind of shutting down and running home to mama.

1509
04:15:13.790 –> 04:15:15.420
Tooba Durraze: for additional… Yes.

1510
04:15:15.420 –> 04:15:17.910
Mark Stouse, Proof Analytics: instructions, right? Yeah.

1511
04:15:18.060 –> 04:15:20.279
Mark Stouse, Proof Analytics: And that… and that’s a… and that’s a problem.

1512
04:15:21.190 –> 04:15:40.409
Tooba Durraze: Yeah, I think one of the things that could be really interesting is how you layer on, kind of, causal AI to Gen AI. You layer on neurosymbolic AI to gen AI. Again, this technology exists in a way that you can kind of marry the two pieces together, and kind of get the best of both worlds, but

1513
04:15:40.480 –> 04:16:00.409
Tooba Durraze: like, agents in the way that we think about them today, will only get us part of the way there if they’re only GenAI-based. So, given that, and the folks in the room are mostly on the go-to-market side, on the marketing side, how does… how does the role of a go-to-market leader evolve in, like, a causal world, let’s just say?

1514
04:16:01.230 –> 04:16:10.270
Mark Stouse, Proof Analytics: Well, there’s some things… I mean, number one, it’s gonna really cause out… causal AI, causal inference, really helps you understand

1515
04:16:10.740 –> 04:16:26.470
Mark Stouse, Proof Analytics: exactly what’s going on, not just with your machine, I’m putting quotes around that, but in the way that it interacts with the marketplace ecosystem, I’m not putting quotes around that.

1516
04:16:26.470 –> 04:16:27.280
Tooba Durraze: Yeah. Right?

1517
04:16:27.530 –> 04:16:35.110
Mark Stouse, Proof Analytics: And so, it is… For example, most B2B marketers do not

1518
04:16:35.530 –> 04:16:39.339
Mark Stouse, Proof Analytics: No, it has probably never crossed their mind.

1519
04:16:39.840 –> 04:16:50.880
Mark Stouse, Proof Analytics: that… The value that they contribute to the go-to-market enterprise is they are a non-linear multiplier.

1520
04:16:51.400 –> 04:16:55.010
Mark Stouse, Proof Analytics: Of sales productivity, sales performance, right?

1521
04:16:55.380 –> 04:16:56.400
Mark Stouse, Proof Analytics: And…

1522
04:16:57.070 –> 04:17:05.620
Mark Stouse, Proof Analytics: When you don’t know that, and you also don’t know that the stuff out there in the marketplace is killing you right now because of…

1523
04:17:05.960 –> 04:17:20.150
Mark Stouse, Proof Analytics: volatility, etc, right? You’re not compensating for it, you don’t know how to compensate for it, you’re just executing the same old roadmap, the same old framework, right?

1524
04:17:20.380 –> 04:17:22.439
Tooba Durraze: And not realizing that.

1525
04:17:22.650 –> 04:17:30.410
Mark Stouse, Proof Analytics: The efficacy and impact of your investments have been falling for years.

1526
04:17:31.100 –> 04:17:36.850
Mark Stouse, Proof Analytics: We’re getting ready to release a report From 2018 to 2025.

1527
04:17:37.450 –> 04:17:47.809
Mark Stouse, Proof Analytics: Right? That shows that… the effectiveness of B2B marketing across a very large number of companies has fallen

1528
04:17:48.730 –> 04:17:52.780
Mark Stouse, Proof Analytics: Well, very recently, it’s just crossed the 50% mark, right?

1529
04:17:52.780 –> 04:17:54.080
Tooba Durraze: Wow, okay.

1530
04:17:54.080 –> 04:18:00.300
Mark Stouse, Proof Analytics: And in 2017, 2018, it was, like, 20-25% in effect, right?

1531
04:18:01.560 –> 04:18:04.329
Tooba Durraze: Yeah, and, like, I… all that is to say is, like.

1532
04:18:04.420 –> 04:18:20.699
Tooba Durraze: I think we think of the idea of, like, being data-driven as, like, we’re applying any and every kind of data, like, into our strategy. So, there are folks that… who are applying this data, but then the inefficiency is still, like, rising. From 25% to 50% is a huge leap, so think of how much opportunity

1533
04:18:20.700 –> 04:18:28.450
Tooba Durraze: They’re leaving on the table that they could optimize for if the right systems were connected in the right way.

1534
04:18:28.640 –> 04:18:30.809
Tooba Durraze: Right? So, now…

1535
04:18:30.810 –> 04:18:33.660
Mark Stouse, Proof Analytics: It also helps you know what to cut.

1536
04:18:33.830 –> 04:18:34.420
Tooba Durraze: Yeah.

1537
04:18:34.420 –> 04:18:36.460
Mark Stouse, Proof Analytics: without hurting anything.

1538
04:18:36.750 –> 04:18:37.470
Tooba Durraze: Yeah.

1539
04:18:37.470 –> 04:18:41.029
Mark Stouse, Proof Analytics: By the way, this also really applies to the CFO.

1540
04:18:41.740 –> 04:18:49.140
Mark Stouse, Proof Analytics: CFO is usually cutting pretty gratuitously, just to meet an EPS target.

1541
04:18:49.140 –> 04:18:49.760
Tooba Durraze: Yep.

1542
04:18:50.460 –> 04:18:58.800
Mark Stouse, Proof Analytics: And that usually entails cutting things that they… if they could understand the impact, particularly time-lagged impact.

1543
04:18:59.080 –> 04:19:01.159
Mark Stouse, Proof Analytics: Right? They wouldn’t make those cuts.

1544
04:19:01.750 –> 04:19:20.350
Tooba Durraze: Yeah, agreed. I would agree with that. So, like, going back to marketers, and the age-old question, like, marketers talk a ton about measuring their value, you talked about, like, the nonlinear relationship between the impact that they have on the business, and one of the things that comes up a ton is ways to solve for

1545
04:19:20.450 –> 04:19:26.679
Tooba Durraze: things like attribution. To me, attribution in general is the wrong problem to solve, because…

1546
04:19:27.210 –> 04:19:41.989
Tooba Durraze: Ideally, you’re not looking for, like, which activity is worth how much to convert, but you’re looking at what sets of activities are, like, actually driving things, and that… that… the sets of activities, to your point, it’s an open system, change fairly regularly.

1547
04:19:41.990 –> 04:19:48.779
Tooba Durraze: So I’m curious on your thoughts on just attribution and attribution tools in general.

1548
04:19:50.540 –> 04:19:57.729
Mark Stouse, Proof Analytics: Yeah, I mean, it’s kind of sad in the sense… okay, so number one… you know, MTA, multi-touch attribution.

1549
04:19:57.730 –> 04:19:58.290
Tooba Durraze: Yeah.

1550
04:19:58.610 –> 04:20:00.250
Mark Stouse, Proof Analytics: is baseless.

1551
04:20:01.210 –> 04:20:03.849
Mark Stouse, Proof Analytics: Right? So I completely agree with you there.

1552
04:20:04.340 –> 04:20:11.360
Mark Stouse, Proof Analytics: I think that one of the… one of the bummers is, is that… The word attribution

1553
04:20:12.080 –> 04:20:16.380
Mark Stouse, Proof Analytics: Technically, is indivisible from causality.

1554
04:20:16.890 –> 04:20:17.330
Tooba Durraze: Yeah.

1555
04:20:17.330 –> 04:20:23.640
Mark Stouse, Proof Analytics: But because of the way it’s been used for the last 10 years, It’s sort of poisoned

1556
04:20:23.920 –> 04:20:27.810
Mark Stouse, Proof Analytics: At least in the marketing or go-to-market space.

1557
04:20:27.810 –> 04:20:28.510
Tooba Durraze: Yeah. Right?

1558
04:20:29.390 –> 04:20:33.669
Mark Stouse, Proof Analytics: So, it also kind of implies that…

1559
04:20:34.650 –> 04:20:38.510
Mark Stouse, Proof Analytics: Marketing’s role is sort of additive.

1560
04:20:40.270 –> 04:20:44.410
Mark Stouse, Proof Analytics: And thus, you can somehow share revenue credit?

1561
04:20:45.310 –> 04:20:48.600
Mark Stouse, Proof Analytics: And that’s just not the way it works, like, at all.

1562
04:20:49.030 –> 04:20:57.069
Mark Stouse, Proof Analytics: Right? It’s, you… you’re not sharing revenue credit, you are showing, or should be showing.

1563
04:20:57.340 –> 04:21:03.190
Mark Stouse, Proof Analytics: How much more efficient and effective sales is because you exist.

1564
04:21:04.590 –> 04:21:04.990
Tooba Durraze: Yeah.

1565
04:21:04.990 –> 04:21:09.710
Mark Stouse, Proof Analytics: than they would if you didn’t exist, right? And…

1566
04:21:09.890 –> 04:21:18.379
Mark Stouse, Proof Analytics: that’s… that’s a classic calculation in causal AI. I mean, that’s… that’s just kind of a normal output, right, on anything.

1567
04:21:18.380 –> 04:21:18.860
Tooba Durraze: Captain.

1568
04:21:18.860 –> 04:21:20.040
Mark Stouse, Proof Analytics: Marketing and sales.

1569
04:21:20.340 –> 04:21:32.989
Tooba Durraze: Yeah, I feel like… I feel like the thoughts around… and you’re right, people, like, in general, I even use the term attribution, but I think that it gets conflated with multi-touch attribution, that’s what we’re talking about right now. Does…

1570
04:21:33.210 –> 04:21:49.560
Tooba Durraze: Sort of put marketing sometimes as a second-degree citizen to, like, the business, as in they have to prove that the things they did contributed to the revenue generated, instead of first-class citizens to say, you were an accelerant to, like, certain things, causing your revenue to rise.

1571
04:21:49.630 –> 04:21:58.320
Tooba Durraze: So if I’m a marketer now, and, you know, I have a ton of agents, Gen AI tools, at my disposal.

1572
04:21:58.320 –> 04:22:13.729
Tooba Durraze: And I’m listening to this, and I’m saying, like, okay, I need to think about things like causal AI, neurosymbolic AI, etc. Where would I start? What would be, like, the first thing I could do, essentially, to try to understand this a bit better and try to apply this?

1573
04:22:14.170 –> 04:22:23.549
Mark Stouse, Proof Analytics: Well, this, this, I guess my answer would probably not surprise anybody, right? I think you have to really understand causally.

1574
04:22:23.920 –> 04:22:26.159
Mark Stouse, Proof Analytics: What’s been going on to this point?

1575
04:22:26.290 –> 04:22:31.760
Mark Stouse, Proof Analytics: Right? There’s an internal benchmark, a causal audit.

1576
04:22:31.930 –> 04:22:33.590
Mark Stouse, Proof Analytics: That has to be done.

1577
04:22:33.750 –> 04:22:34.230
Tooba Durraze: Yep.

1578
04:22:35.460 –> 04:22:44.619
Mark Stouse, Proof Analytics: One of the coolest things in the last several years is that even if your data, your natural data, your operational data.

1579
04:22:44.830 –> 04:22:46.490
Mark Stouse, Proof Analytics: It’s kind of marginal.

1580
04:22:46.630 –> 04:22:52.550
Mark Stouse, Proof Analytics: Or maybe even doesn’t exist in the way that maybe it should.

1581
04:22:52.660 –> 04:22:59.370
Mark Stouse, Proof Analytics: There’s a lot that we can do today with very high-fidelity synthetic data.

1582
04:22:59.980 –> 04:23:06.400
Mark Stouse, Proof Analytics: That can help fill in the gaps and simulate, and it’s represented clearly as a… as…

1583
04:23:06.800 –> 04:23:09.709
Mark Stouse, Proof Analytics: What it is, which is a, you know, a mix…

1584
04:23:09.940 –> 04:23:15.959
Mark Stouse, Proof Analytics: Of synthetic and natural data, or maybe a pure synthetic model, so that no one’s confused.

1585
04:23:16.160 –> 04:23:16.550
Tooba Durraze: Yeah.

1586
04:23:16.550 –> 04:23:31.399
Mark Stouse, Proof Analytics: But you… but you get a real… you can… you can get to a level of understanding that used to be just kind of, like, outside of what was possible unless you were a nation-state or something, right?

1587
04:23:31.400 –> 04:23:39.180
Tooba Durraze: Yeah, yeah, that was… that’s a fallacy around AI that has existed, right? It’s like, I don’t have enough data to get started, or…

1588
04:23:39.230 –> 04:23:52.320
Tooba Durraze: in go-to-market case, folks will often say, like, my data is, like, too messy to even be useful in some of these AI systems, and I feel like synthetic data is a really, really good option to kind of bridge that gap. Now, again.

1589
04:23:52.950 –> 04:24:00.399
Tooba Durraze: It’s a GPS, it’s directional. I think we’re used to consuming data sometimes in a way where we feel like it has to be, like, precise.

1590
04:24:00.550 –> 04:24:08.310
Tooba Durraze: Essentially. Do you think that, for marketers and leaders, there should be more trust around things like synthetic data?

1591
04:24:09.140 –> 04:24:09.699
Tooba Durraze: Or do you think…

1592
04:24:09.700 –> 04:24:17.610
Mark Stouse, Proof Analytics: I think that there’s gonna have to be… there’s gonna have to be. Yeah. There’s just… we’re kind of getting to a place where…

1593
04:24:18.070 –> 04:24:22.259
Mark Stouse, Proof Analytics: If you don’t do… if you… if you don’t trust…

1594
04:24:22.610 –> 04:24:23.250
Tooba Durraze: Yeah.

1595
04:24:23.250 –> 04:24:26.659
Mark Stouse, Proof Analytics: Certain things, you’re just gonna stop.

1596
04:24:27.270 –> 04:24:31.769
Mark Stouse, Proof Analytics: you’re gonna stop. You won’t be able to go any further

1597
04:24:32.110 –> 04:24:38.550
Mark Stouse, Proof Analytics: Or you’ll be able to progress, but much more slowly. You will totally lose a competitive edge.

1598
04:24:38.900 –> 04:24:43.090
Mark Stouse, Proof Analytics: Not only organizationally, corporately, but personally.

1599
04:24:43.430 –> 04:24:44.200
Tooba Durraze: Yep.

1600
04:24:44.420 –> 04:24:48.830
Mark Stouse, Proof Analytics: And I think that that’s… That’s just kind of the way it is.

1601
04:24:49.580 –> 04:24:56.510
Mark Stouse, Proof Analytics: We can… we can debate that. I mean, is that good? Is that bad? You know, that kind of thing, but it’s not going to change

1602
04:24:57.180 –> 04:24:59.029
Mark Stouse, Proof Analytics: The reality of it.

1603
04:24:59.290 –> 04:25:00.950
Mark Stouse, Proof Analytics: At all.

1604
04:25:01.130 –> 04:25:01.720
Tooba Durraze: Yeah.

1605
04:25:02.210 –> 04:25:12.810
Mark Stouse, Proof Analytics: So, I mean, it’s, and you’re not, you know, I think everybody really needs to stop and consider

1606
04:25:14.160 –> 04:25:25.220
Mark Stouse, Proof Analytics: That if you can’t show what your company is getting in return for the money that they’re spending on fill-in-the-blank, right?

1607
04:25:25.330 –> 04:25:29.419
Mark Stouse, Proof Analytics: Increasingly, you’re gonna be very unlikely to get it.

1608
04:25:31.130 –> 04:25:34.239
Mark Stouse, Proof Analytics: Yeah. And that’s a problem.

1609
04:25:34.520 –> 04:25:41.599
Mark Stouse, Proof Analytics: In a lot of ways, but it’s a huge problem in the sense that go-to-market is the revenue engine of the company.

1610
04:25:42.580 –> 04:25:44.040
Tooba Durraze: No, absolutely, yeah.

1611
04:25:44.040 –> 04:25:44.860
Mark Stouse, Proof Analytics: So…

1612
04:25:44.860 –> 04:25:56.590
Tooba Durraze: it sometimes happens slowly over time as well. I think there’s drift sometimes that people don’t actually notice until it’s a little bit too late. So that’s another thing to watch out for, because I see a lot of times

1613
04:25:56.730 –> 04:26:12.560
Tooba Durraze: books, kind of repeating playbooks that have worked, and it might continue to work within a certain degree, but you might see depreciating returns that you don’t pick up on, essentially. So understanding the root cause of, like, why something is working, and applying that root cause to other things.

1614
04:26:12.590 –> 04:26:23.870
Tooba Durraze: that you could do that would make it work. I think it’s more important than, you know, yesterday, everyone thumbs up my post, and today, if I write a similar post, everyone’s gonna thumbs up my post as well.

1615
04:26:23.900 –> 04:26:36.489
Tooba Durraze: I bring that up because there’s a lot of debate around, how personalization at scale kind of impacts your go-to-market or your marketing motion, and the argument I always bring to the table is, like, you don’t know what

1616
04:26:36.650 –> 04:26:43.019
Tooba Durraze: What is the affinity that you’re personalizing on that is appealing to the person on the other side?

1617
04:26:43.140 –> 04:26:52.850
Tooba Durraze: So is it their title? Like, everyone has, like, so many attributes and affinities, right? So you don’t know, and unless you figure that out, it’s very hard to kind of correlate that and say.

1618
04:26:53.430 –> 04:27:00.230
Tooba Durraze: Yeah, like, they may already have opened an email, but the reason why they opened an email is very important to know, to kind of repeat that success.

1619
04:27:00.430 –> 04:27:01.450
Tooba Durraze: Okay.

1620
04:27:01.450 –> 04:27:07.530
Mark Stouse, Proof Analytics: I’m gonna… I’m just gonna… I’ll just say this real fast. Most people… find…

1621
04:27:08.110 –> 04:27:13.569
Mark Stouse, Proof Analytics: the fact that causal AI is a Pretty definitive readout.

1622
04:27:13.830 –> 04:27:14.370
Tooba Durraze: Yeah.

1623
04:27:14.370 –> 04:27:15.829
Mark Stouse, Proof Analytics: Kind of scary.

1624
04:27:16.420 –> 04:27:21.910
Tooba Durraze: Right? Like, it’s gonna… it’s gonna tell me I’m wrong, right? I’ve been doing it all wrong. Yeah.

1625
04:27:21.910 –> 04:27:26.610
Mark Stouse, Proof Analytics: Let me just say that that’s not usually how it goes in go-to-market.

1626
04:27:26.930 –> 04:27:32.980
Mark Stouse, Proof Analytics: You’re gonna be more… you’re gonna score higher on the test, so to speak, than you probably think.

1627
04:27:33.050 –> 04:27:49.890
Mark Stouse, Proof Analytics: Right? And the most important thing is that while you may have to learn a few things up front that is… that don’t make you feel really good, it’s gonna limit your downside risk going forward.

1628
04:27:50.070 –> 04:27:52.180
Mark Stouse, Proof Analytics: In a huge way.

1629
04:27:52.600 –> 04:27:53.180
Tooba Durraze: Yeah.

1630
04:27:53.730 –> 04:27:54.160
Mark Stouse, Proof Analytics: So…

1631
04:27:54.160 –> 04:28:05.749
Tooba Durraze: I agree. It’s a more stable and dependable system than the current systems. I think that we have to change our consumption behavior from these systems as well, because we’re so used to getting quick wins.

1632
04:28:05.890 –> 04:28:21.049
Tooba Durraze: out of these systems, but those quick wins, the opportunity cost is, like, it’s actually causing you to be more inefficient. So taking the time to understand a system, implementing a system, letting it kind of do its thing, and then applying it makes a huge difference.

1633
04:28:21.210 –> 04:28:32.830
Tooba Durraze: I know we have other panelists joining us, but I… my last question is, what is your prediction over the next 3 to 5 years? Does everyone adopt causal AI? Do people adopt neurosymbolic AI, or how does that evolve?

1634
04:28:34.360 –> 04:28:41.250
Mark Stouse, Proof Analytics: You know, if the situation in Delaware hadn’t happened, and if similar situations

1635
04:28:41.350 –> 04:28:46.960
Mark Stouse, Proof Analytics: in the EU, the UK, the PRC, Hadn’t happened.

1636
04:28:46.960 –> 04:28:47.900
Tooba Durraze: Yeah.

1637
04:28:47.900 –> 04:28:52.620
Mark Stouse, Proof Analytics: I would have said that it would take a lot… it would take 10 years, okay?

1638
04:28:52.880 –> 04:28:53.540
Tooba Durraze: Okay.

1639
04:28:53.540 –> 04:28:56.590
Mark Stouse, Proof Analytics: to see causal AI be pervasive.

1640
04:28:56.800 –> 04:28:57.530
Tooba Durraze: Yeah.

1641
04:28:57.530 –> 04:28:59.620
Mark Stouse, Proof Analytics: the regulatory pressure

1642
04:29:00.290 –> 04:29:08.900
Mark Stouse, Proof Analytics: Which, you know, it takes about 3 years for all that to kind of filter through a system. That was the way it was for SOX as well, right?

1643
04:29:09.290 –> 04:29:14.300
Mark Stouse, Proof Analytics: So, we’re almost to that point here, and

1644
04:29:14.990 –> 04:29:24.379
Mark Stouse, Proof Analytics: It’s definitely a tailwind for our business, and for the business of our competitors… It… you just won’t…

1645
04:29:24.780 –> 04:29:31.540
Mark Stouse, Proof Analytics: You literally won’t be able to operate anymore without it, because nobody is willing to put up

1646
04:29:31.940 –> 04:29:33.589
Mark Stouse, Proof Analytics: with the guesses.

1647
04:29:33.820 –> 04:29:34.410
Mark Stouse, Proof Analytics: Right?

1648
04:29:34.410 –> 04:29:35.030
Tooba Durraze: Yeah.

1649
04:29:35.240 –> 04:29:36.820
Mark Stouse, Proof Analytics: And… and that’s the…

1650
04:29:36.990 –> 04:29:46.550
Mark Stouse, Proof Analytics: And the fact that now a CMO or a CRO or somebody can be sued directly by shareholders.

1651
04:29:46.550 –> 04:29:47.450
Tooba Durraze: Scary.

1652
04:29:47.450 –> 04:29:51.000
Mark Stouse, Proof Analytics: That’s it, yeah, no, it’s… I mean, it’s bringing a whole new…

1653
04:29:51.270 –> 04:29:59.090
Mark Stouse, Proof Analytics: Real fast, right? I was just part of a case, I was consulting for a case,

1654
04:29:59.250 –> 04:30:12.949
Mark Stouse, Proof Analytics: in California, large Delaware, domiciled company, sued by, and, you know, shareholders, naming the CRO, the CIO, and the CDO, the Chief Data Offic

1655
04:30:14.150 –> 04:30:19.619
Mark Stouse, Proof Analytics: Kind of the people, process, technology representatives, on this project?

1656
04:30:19.790 –> 04:30:26.570
Mark Stouse, Proof Analytics: And it was all about CRM data quality. It was about nothing else, okay? Just that.

1657
04:30:27.620 –> 04:30:37.090
Mark Stouse, Proof Analytics: They ran analysis, and mainly, mainly pattern match, because it was essentially a fraud detection case.

1658
04:30:37.190 –> 04:30:41.630
Mark Stouse, Proof Analytics: And 14 years worth of data, two-thirds of it came back.

1659
04:30:42.090 –> 04:30:47.890
Mark Stouse, Proof Analytics: just… completely engineered by God only knows how many sales guys over the years.

1660
04:30:47.990 –> 04:30:48.570
Tooba Durraze: Yeah.

1661
04:30:48.750 –> 04:30:56.260
Mark Stouse, Proof Analytics: And they went immediately to settlement, and I… you know, it’s… it was a lot of money. I mean, a lot of money.

1662
04:30:56.880 –> 04:31:03.300
Mark Stouse, Proof Analytics: And so, now that company is dealing with 5 more suits…

1663
04:31:03.810 –> 04:31:08.790
Tooba Durraze: Not about CRM data quality, but about data quality in other parts of the business.

1664
04:31:09.840 –> 04:31:21.140
Tooba Durraze: Yeah, and again, I will close this out. Julius, I’m sorry, we could talk forever about these topics, clearly, as you can tell. But essentially, again, the intention here is, like, not to…

1665
04:31:21.260 –> 04:31:38.160
Tooba Durraze: scare the folks in the room, but to more inform them that it is our responsibility to kind of get to learn about these things, and sometimes not to get caught up in the hype of, like, every new piece of technology and what that offers. Like, there is a lot of foundational tech here that I think

1666
04:31:38.160 –> 04:31:48.440
Tooba Durraze: serves us well in terms of businesses, so the best way you can do that is just educate yourself, is read. And spending time tinkering and reading, I think it’s the only way forward.

1667
04:31:49.690 –> 04:31:51.550
Tooba Durraze: Let me pass it back to you, Julia.

1668
04:31:51.550 –> 04:31:59.060
Julia Nimchinski: Phenomenal parasite chat, thank you so much. Super, super thought-provoking. What’s the best way to support you, Tuba and Mark?

1669
04:32:00.060 –> 04:32:00.930
Tooba Durraze: Go ahead, Mark.

1670
04:32:02.320 –> 04:32:05.740
Mark Stouse, Proof Analytics: You know, I would say learn, right?

1671
04:32:06.420 –> 04:32:08.409
Mark Stouse, Proof Analytics: Just be open to learning.

1672
04:32:08.900 –> 04:32:12.850
Mark Stouse, Proof Analytics: You know, I see a lot of…

1673
04:32:13.640 –> 04:32:17.300
Mark Stouse, Proof Analytics: Fireside chats, or a lot of panels, or whatever, that…

1674
04:32:17.610 –> 04:32:22.270
Mark Stouse, Proof Analytics: Are all the same kind of people talking to…

1675
04:32:22.500 –> 04:32:25.610
Mark Stouse, Proof Analytics: Just to people in their space that are…

1676
04:32:25.780 –> 04:32:30.370
Mark Stouse, Proof Analytics: just younger than they are in their career, right? And…

1677
04:32:30.990 –> 04:32:34.070
Mark Stouse, Proof Analytics: That is… that… that is being inbred.

1678
04:32:34.190 –> 04:32:36.209
Mark Stouse, Proof Analytics: In your thinking, right?

1679
04:32:36.550 –> 04:32:37.130
Tooba Durraze: Yes.

1680
04:32:37.130 –> 04:32:47.629
Mark Stouse, Proof Analytics: I mean, go outside of marketing and say to other, you know, functional leaders, how are you dealing with this, this, and this?

1681
04:32:48.070 –> 04:32:58.350
Mark Stouse, Proof Analytics: Right? I mean, I got started along this whole line of thought by listening 20 years ago, or more, listening to CIOs in briefing centers.

1682
04:32:58.620 –> 04:33:01.929
Mark Stouse, Proof Analytics: Talk about exactly the same problem.

1683
04:33:02.170 –> 04:33:13.510
Mark Stouse, Proof Analytics: And I was sitting there going… as a CMO, I was sitting there going, shit, I can just, like, do a search and replace on what they’re saying, and it applies the same to me.

1684
04:33:13.869 –> 04:33:16.849
Mark Stouse, Proof Analytics: And so I started to learn how they were dealing with it.

1685
04:33:18.119 –> 04:33:25.119
Mark Stouse, Proof Analytics: So I think that’s the best way to support this, is… Realize that learning means…

1686
04:33:25.700 –> 04:33:28.420
Mark Stouse, Proof Analytics: That you don’t know. And that’s okay.

1687
04:33:29.560 –> 04:33:39.609
Tooba Durraze: Same for me. I think we’re both evangelists of, like, the more informed people are, the better off we’ll all be. So, yes, I’ll take Mark’s advice on that. Learn

1688
04:33:39.610 –> 04:33:54.770
Tooba Durraze: read… reach out to both of us on LinkedIn or otherwise. We’re happy to share, like, papers, documents, and things, et cetera, to help inform you. But, yeah, keep your eyes open on everything that exists, and keep your eyes… keep your head forward on what’s coming next.

1689
04:33:55.520 –> 04:33:57.179
Julia Nimchinski: Thank you so much again. Thank you.

1690
04:33:57.330 –> 04:33:58.160
Mark Stouse, Proof Analytics: Super.

1691
04:33:58.340 –> 04:33:59.009
Tooba Durraze: Thanks.