How to Build a Fraud Team That Earns its Investment

WEBVTT

1
00:00:02.700 –> 00:00:18.220
Jerry Hoff: Welcome, everybody, to the Blueprint series, How to Build a Fraud Team That Earns Its Investment. This is going to be a very fun and very timely discussion, kind of Halloween edition. We don’t have any Halloween, we don’t have any Halloween

2
00:00:18.220 –> 00:00:26.830
Jerry Hoff: Affectations, but nevertheless, should be fun. I’m here with SIFT Field CTO Kevin Lee.

3
00:00:26.830 –> 00:00:37.879
Jerry Hoff: and trust and safety leader, and… I didn’t know this until recently, alumni. We both went to Washington University in St. Louis, Barrington Lloyd-Lovet. How are you guys doing today?

4
00:00:38.340 –> 00:00:40.120
Kevin Lee: Excellent! Happy to be here.

5
00:00:40.560 –> 00:00:42.570
Barrington Lloyd-Lovett: Yeah, I’m doing very well. Glad to be here.

6
00:00:42.930 –> 00:00:43.480
Jerry Hoff: Yeah.

7
00:00:43.620 –> 00:00:50.770
Jerry Hoff: Very good. Well, I want to make sure everybody knows that we’ve got a bunch of people watching as we stream this.

8
00:00:50.820 –> 00:01:08.620
Jerry Hoff: please ask questions. This is not us just speaking out into the internet. We want to make sure this is a two-way conversation, so please feel free to enter your questions into the chat, and we will be answering them either as we get them, or we will have a Q&A section at the end. This will be a 45-minute

9
00:01:08.780 –> 00:01:15.630
Jerry Hoff: webinar, and it goes by very, very quickly, so, strap yourselves in. So, this first question

10
00:01:15.640 –> 00:01:23.379
Jerry Hoff: again, we’re talking about, really, how to, you know, get the most out of your… out of your program. AI is obviously the place to start.

11
00:01:23.380 –> 00:01:43.309
Jerry Hoff: every leader on the planet right now is looking at AI and how it can meaningfully increase some value in its organization. And of course, that goes doubly for fraud. So, my first question, Barrington, what do you think about this? Where can AI meaningfully increase a fraud team’s capacity? What comes to mind on this particular question?

12
00:01:44.070 –> 00:01:52.680
Barrington Lloyd-Lovett: The first thing that comes to mind for me when I think about AI complementing what humans can do, and what a fraud team can do, is the…

13
00:01:53.030 –> 00:01:57.640
Barrington Lloyd-Lovett: the speed at which it can gather information doing an investigation. So, you know.

14
00:01:57.680 –> 00:02:16.669
Barrington Lloyd-Lovett: AI is effectively a tool that we use to do things that we can do, but faster and more broadly. And so, if you have, you know, whether it’s a particular incident, or you’re trying to investigate, kind of, more broad trends, you can send out an agent, or maybe a team of agents, to gather a bunch of information quickly, and do the first, maybe even the second order

15
00:02:16.680 –> 00:02:26.399
Barrington Lloyd-Lovett: Of, kind of processing of that. And then from there, have your fraud team look at what’s come out, validate that information, and then decide what to do with it.

16
00:02:27.770 –> 00:02:33.620
Jerry Hoff: That’s interesting. You know, everybody uses AI, I use it, you use it, everybody uses it. Are there some.

17
00:02:33.620 –> 00:02:35.409
Kevin Lee: Gotta, make sure you have enough tokens.

18
00:02:35.410 –> 00:02:51.920
Jerry Hoff: Yeah, you’ve got to make sure you have enough tokens, right? We were just talking about that, where you can run out of tokens and be kind of stuck, but are there certain types of tasks that are more amenable to AI and certain tasks that you would not want to give to AI? What does that breakdown look like?

19
00:02:53.280 –> 00:03:07.959
Kevin Lee: I can take on that one a little bit, and then curious to hear what Barrington Barrington takes, too. Really, so the lowest common denominator here, to me, is around repetitive, high-volume tasks, whether it’s data aggregation, even decisions.

20
00:03:07.960 –> 00:03:16.140
Kevin Lee: When I do look at investigations and the teams that I work with, still, unfortunately, so much time is spent

21
00:03:16.150 –> 00:03:34.649
Kevin Lee: compiling data, not even thinking about the data itself and contextualizing the story, and whether it’s a transaction or account that might be at play. So much time is still spent on just collecting the data, where… and AI still has some room to grow here with regards to how connected that data is.

22
00:03:34.650 –> 00:03:46.659
Kevin Lee: But that’s really where I’ll call it step one when it comes to leveraging AI. It’s just data aggregation, and then, you know, humans can provide a bit more context on top.

23
00:03:47.900 –> 00:03:49.449
Barrington Lloyd-Lovett: Yeah, I am.

24
00:03:49.450 –> 00:03:49.910
Jerry Hoff: Go ahead, Ben.

25
00:03:49.910 –> 00:03:53.310
Barrington Lloyd-Lovett: Sorry, just to add to the… I think…

26
00:03:53.320 –> 00:04:12.199
Barrington Lloyd-Lovett: Or maybe to go on the things that we wouldn’t want AI to touch, I think things that need to be, you know, reported to authorities, particularly, you know, a lot of the more sensitive things when it comes to child safety, those things still need to, in many cases, be validated by an account permanently, or your, your,

27
00:04:12.630 –> 00:04:25.150
Barrington Lloyd-Lovett: making decisions that will affect someone’s ability to use the product long-term. There should at least be someone making it, you know, having a good look on, is this the right decision? Is this…

28
00:04:25.190 –> 00:04:39.189
Barrington Lloyd-Lovett: something that we know is going to be highly disruptive, not just for a person, but maybe for a team or a company, that should have human validation. Unless you know the signals are incredibly clear that this is always bad behavior.

29
00:04:40.690 –> 00:05:00.359
Jerry Hoff: I like that. I think… I think this is where the rubber hits the road. I mean, we’ve all seen, you know, organizations like Gartner and so forth have reported that the majority of these AI projects fail. So, you know, figuring out exactly what it is, because right now, we seem to be where, you know, there’s quite a bit of… of,

30
00:05:00.630 –> 00:05:17.500
Jerry Hoff: maybe a bit too much positives, you know, positivity as to what AI can do. And a lot of people are like, oh, it’s going to take all the jobs, it’s gonna do everything, but we all know that’s not the case. So I think figuring out exactly what AI can do reliably and what it can’t, and I like that idea of

31
00:05:17.560 –> 00:05:32.370
Jerry Hoff: If it’s something that is going to interface you with any kind of regulatory or legal environment, that might be something where the humans need to be more involved, at least for the time being. So, very interesting. Interesting.

32
00:05:32.640 –> 00:05:44.160
Jerry Hoff: So, moving forward, and Kevin, I want to start with you, if you don’t mind. What or how might a fraud team’s structure change as AI agents take on specialized tasks?

33
00:05:44.280 –> 00:05:46.290
Jerry Hoff: And this is where we’re getting into, like.

34
00:05:46.490 –> 00:06:03.739
Jerry Hoff: you know, like, when I was a kid, and I was watching Star Wars, you know, I never thought that, like, you know, as an adult, we would be leading, essentially, bots, right? Like, intelligent, you know, intelligent programs, and leading them. But, you know, we talked a little bit about this before, the webinar, Kevin, where

35
00:06:03.930 –> 00:06:17.020
Jerry Hoff: Everybody’s going to have to become a leader, right? You’re going to have to have various agents, you’re going to be delegating agents, you’re going to be having, you know, other organizations’ agents agents, and you’re going to have agent interactions, and probably dispute resolution. So, what’s your.

36
00:06:17.020 –> 00:06:17.470
Kevin Lee: take…

37
00:06:17.910 –> 00:06:23.180
Jerry Hoff: How all of this is going to change as we have these agents taking on specialized tasks.

38
00:06:24.170 –> 00:06:41.340
Kevin Lee: Yeah, I was actually at an industry meetup a couple days ago, and I was chatting with, of course, the topic of AI came up, and we’re talking about managing humans and managing agents, and some folks in the room were like, yeah, I feel like I might be a better

39
00:06:41.340 –> 00:06:44.290
Kevin Lee: Agent manager than a human manager, which…

40
00:06:44.290 –> 00:07:00.660
Kevin Lee: comes with some, maybe, baggage there as well. But really, what we settled on in kind of this Agentic fraud world, or fraud org of the future here, I found this on LinkedIn from Brian Davis, and I thought it actually did resonate with me where there is

41
00:07:00.720 –> 00:07:21.769
Kevin Lee: a separation happening where, in the previous slide, we talked about how, okay, AI can do a lot more of the grunt work, a lot of the data consolidation, maybe some summary. Of course, we want a human in the loop, but really what we’re looking at here, and there’s different variations to this, but essentially what AI agents have started, or introduced.

42
00:07:21.770 –> 00:07:26.640
Kevin Lee: Is, whether it’s a set of skills, or a set of tasks, that…

43
00:07:26.680 –> 00:07:38.889
Kevin Lee: Humans and agents will continue to work together, but let’s say the bottom of the pyramid, you have, yes, your senior investigators in the mix, you have Fraud Ops as well, but you have much more

44
00:07:39.390 –> 00:07:50.230
Kevin Lee: playbooks and runbooks that AI agents are doing most of the work on, and as you work your way up the chain here, whether it’s execution to orchestration.

45
00:07:50.360 –> 00:08:09.190
Kevin Lee: humans, from a strategy perspective, from a content perspective, continue to play a heavier hand in that. But this is just one real illustration of what different orgs are doing. So I’ve seen, actually, variations of this play out in real life now, where there is

46
00:08:09.290 –> 00:08:20.890
Kevin Lee: Still a human in the loop, but the skill set is changing, where, let’s say, and new roles are being created, where, yeah, you have to…

47
00:08:21.520 –> 00:08:29.570
Kevin Lee: create, or you have to build, not only the agents that you create and the skills, but how do you manage that, and how do you orchestrate that? And these are things that

48
00:08:29.590 –> 00:08:45.660
Kevin Lee: 6 months ago or a year ago, they were still more, I’d say, speculative, and now there’s some real experimentation going on. And, Jerry, to your earlier point, there’s still a lot more failures than successes, in my opinion. Like, there’s just a ton of experimentation going on, which…

49
00:08:45.840 –> 00:08:59.000
Kevin Lee: involves token usage and more, kind of, maybe not using tokens the most efficiently as we could or should, but this is part of the process of, kind of, leading that frontier side and figuring out what works, what doesn’t. Once we figure out something that does work.

50
00:08:59.000 –> 00:09:07.789
Kevin Lee: how we automate it a little bit more effectively, but this is, I think, a good roadmap or a visualization of what teams could look like.

51
00:09:09.740 –> 00:09:27.949
Jerry Hoff: So, if we go to the bottom here, context intelligence. So, kind of getting to what you both said in terms of where we would use AI the most, this ability to gather information, is that, like, mainly Agentic? That very bottom green layer, would you say? That’s, what, like, 90% Agentic? Something along those lines?

52
00:09:28.200 –> 00:09:41.819
Kevin Lee: The majority should be. I don’t know if it’s gonna get to 90, and actually, yeah, I could exceed it, but certainly, when I look at the maturity curve of a lot of different fraud teams at the moment, it’s not at 90%, not even close. But it is getting… it is…

53
00:09:42.380 –> 00:09:50.399
Kevin Lee: growing more rapidly, depending on… I’ll call it the skill set of the humans out there, where I think about…

54
00:09:50.570 –> 00:10:08.390
Kevin Lee: hiring different fraud teams and building them in the past, it was more traditional, like, transaction-by-transaction queue type of work, and that’s really faded off to the side now. Then it shifted towards, oh, we need to hire someone with more SQL skills or data analyst skills. Now it’s…

55
00:10:08.620 –> 00:10:24.689
Kevin Lee: moving even more in that direction when it comes to… they don’t need a CS background, they don’t need a data science background per se, but they really need to understand, essentially, what prompts to put into the system, and put it in such a way that it is

56
00:10:25.090 –> 00:10:30.980
Kevin Lee: The more clear, the better, so the system can kind of deliver the output that is desired.

57
00:10:31.860 –> 00:10:35.080
Jerry Hoff: Gotcha. Yeah, I’m sorry, go ahead, Barrington.

58
00:10:35.080 –> 00:10:38.820
Barrington Lloyd-Lovett: Sorry, I just wanted to jump in here. I think this is a really interesting question, because…

59
00:10:39.010 –> 00:10:48.560
Barrington Lloyd-Lovett: On some level, If people don’t understand the core work, then they won’t be able to build, use.

60
00:10:48.750 –> 00:10:56.020
Barrington Lloyd-Lovett: manage the agents. And so, I don’t know what the percentage looks like, but I think…

61
00:10:58.030 –> 00:11:12.059
Barrington Lloyd-Lovett: A failure mode is new people coming in to the industry, not having the skills, not having built the understanding of, you know, what an investigation looks like, what fraud might look like, etc, and having to rely on agents.

62
00:11:12.060 –> 00:11:21.509
Barrington Lloyd-Lovett: That are, you know, they’re limited in the capacity based on when they were built, who they were built by, what they were looking for. And fraud and fraud signals change, you know, continually.

63
00:11:22.010 –> 00:11:38.670
Barrington Lloyd-Lovett: as we figure out what attackers are doing, they’re going to adjust, so then we have to adjust. Agents can only adjust so far without human input, particularly because humans are the ones that actually decide whether we want to, like, you know, we basically decide what fraud is. And so, I don’t know…

64
00:11:39.650 –> 00:11:53.919
Barrington Lloyd-Lovett: I don’t know that 90% is sustainable long-term, unless we want to kind of write humans out of the industry entirely, which is a separate thing. And I don’t think… I don’t know how well that works with, like, adaptive models.

65
00:11:54.330 –> 00:11:58.300
Barrington Lloyd-Lovett: Of fraud prevention and fraud team building.

66
00:11:58.920 –> 00:12:08.189
Jerry Hoff: So, there’s a lot to unpack there, but I mean, yeah, one giant thing that… one great point you just made, out of many, is,

67
00:12:08.830 –> 00:12:14.649
Jerry Hoff: organizations that are just starting their fraud program should not jump right to AI. If you can’t do it with people.

68
00:12:14.780 –> 00:12:17.339
Jerry Hoff: Jumping right to AI is going to…

69
00:12:17.820 –> 00:12:20.669
Jerry Hoff: it’s gotta get you into a world of hurt.

70
00:12:21.070 –> 00:12:23.439
Jerry Hoff: What’s gonna happen in that case?

71
00:12:24.000 –> 00:12:40.950
Kevin Lee: I mean, I was… I mean, to make it a little bit more personal, I was literally going through this type of discussion, actually with my son. He’s younger, still in grade school, and he’s going through, like, multiplication and division and all that kind of stuff. And, of course, you could use a calculator, you can use Gemini or ChatGPT to answer this stuff.

72
00:12:40.950 –> 00:12:55.979
Kevin Lee: But knowing the fundamental is actually more important than ever. It doesn’t mean you have to do the long form all the time and forever in perpetuity. Like, yes, eventually you’re going to graduate and you’re going to use calculators and different LLMs and things like that to help you and aid you, but…

73
00:12:55.980 –> 00:13:04.319
Kevin Lee: you need to understand the fundamentals, and especially in the fraud space, and the reason being is from… I’ll look at it from a bad actor standpoint.

74
00:13:04.430 –> 00:13:18.090
Kevin Lee: their objective is to create the chaos to break the game, to not follow the rules in order to monetize whatever they’re looking to monetize. And so, from a human-in-the-loop standpoint, and understanding the fundamentals of, I’ll call it the business, the fraud business.

75
00:13:18.480 –> 00:13:35.529
Kevin Lee: you have to have that core skill set, and you still need to learn it, and one of the best ways to do it is by honing your craft, understanding the quote-unquote long form, and then you can build yourself up into this kind of more efficient standpoint. But if you were to start here.

76
00:13:35.930 –> 00:13:42.580
Kevin Lee: It’ll be too easy from a fraud perspective to undercut it, and then it’ll just… the system will degrade too quickly.

77
00:13:44.640 –> 00:13:58.469
Jerry Hoff: So, when I’ve led pretty large organizations in the past, and if one of my, you know, somebody in my org does something wrong, I’ve got an answer for it, right? So, as the boss, that’s where the buck stops, you know, with me.

78
00:13:58.880 –> 00:14:03.889
Jerry Hoff: what do you do in the agent world? So if an agent misbehaves, who,

79
00:14:04.230 –> 00:14:21.440
Jerry Hoff: I guess maybe it’s too early to know, but what does that look like in terms of you’ve got an agent that’s done something really bad, right? Maybe, you know, it’s caused a big financial loss. Is it the manager? Is it the programmer? Is it the model? Is it the vendor? Where does the…

80
00:14:21.530 –> 00:14:24.350
Jerry Hoff: Where do we cast the blame, or is it all of the above?

81
00:14:25.230 –> 00:14:26.829
Barrington Lloyd-Lovett: It’s a good question, I, I mean…

82
00:14:26.830 –> 00:14:27.540
Kevin Lee: Barrington.

83
00:14:27.540 –> 00:14:36.440
Barrington Lloyd-Lovett: I think it’s something… Ideally, the organization comes up with the answers to that before they implement.

84
00:14:36.920 –> 00:14:38.230
Barrington Lloyd-Lovett: Agentic work.

85
00:14:38.350 –> 00:14:51.730
Barrington Lloyd-Lovett: Right? So, you have to know as an organization, if this goes wrong, how are we going to respond to it? If something happens, who’s responsible for it? And being able to answer to your leadership team before you put that in place, I think, is important.

86
00:14:52.100 –> 00:14:55.680
Barrington Lloyd-Lovett: Ultimately, you know, if an agent is built

87
00:14:55.880 –> 00:15:03.689
Barrington Lloyd-Lovett: To do a certain thing, and you don’t have guardrails for it to not do other things, or even if you do have guardrails and those break or whatever,

88
00:15:05.550 –> 00:15:18.369
Barrington Lloyd-Lovett: the person implementing it, I think, is at least partially responsible, but I think responsibility will ultimately be shared amongst multiple people. It’s the person who wrote the skills, the person who wrote the prompt.

89
00:15:18.450 –> 00:15:33.230
Barrington Lloyd-Lovett: the, you know… And then, you know, if it affects customers, they don’t really care whose fault it is, they’re just gonna hold the whole company responsible. And I think that’s something that people should really keep in mind as they implement, these flows.

90
00:15:34.630 –> 00:15:39.339
Kevin Lee: I mean, from my perspective, I’ll keep it even more simple, like, ultimate accountability.

91
00:15:39.480 –> 00:15:56.920
Kevin Lee: will continue to lead to a human, whether it’s the engineer, the operations person, the program manager, whatever it might be. Yeah, just like any retro post-mortem we do today, like, we focus on the problem, and kind of what teams and people need to change, and all those things, but

92
00:15:57.550 –> 00:16:10.440
Kevin Lee: it is going to lead up to a human, at least for the foreseeable future, in my opinion. And this, to me, this is why QC is so important, where we can deploy different types of automation, ML,

93
00:16:11.000 –> 00:16:18.559
Kevin Lee: Like, agents to do these things, but when it comes to accountability, the buck stops for me is still out of human.

94
00:16:19.960 –> 00:16:36.879
Jerry Hoff: I love that, and what’s interesting about that, from both of your comments, is organizations need to put… I mean, they have to do this work up front, and I don’t know, like, when I talk to various organizations that are doing AI implementation, I don’t know that they’ve had these in-depth thoughts as to

95
00:16:36.880 –> 00:16:45.820
Jerry Hoff: you know, AI policy, AI, I guess, hierarchy to some degree. There’s just a lot of groundwork that needs to be done that

96
00:16:45.880 –> 00:16:49.660
Jerry Hoff: I think organizations need to think about before they even start implementing this.

97
00:16:49.880 –> 00:16:51.810
Jerry Hoff: Very good.

98
00:16:51.950 –> 00:17:08.509
Jerry Hoff: Interesting. I mean, it seems obvious when we say it out loud, but I mean, I’m just saying, like, from what I see, the truth of the matter is, I think most organizations, and maybe this is why Gartner is saying that a lot of these are going to fail, they jump right to the tech, whereas there needs to be kind of a little bit more on the,

99
00:17:08.510 –> 00:17:14.379
Jerry Hoff: Delegation of roles and responsibilities, policy, and all that other kind of,

100
00:17:14.480 –> 00:17:20.020
Jerry Hoff: you know, guard… human and… human and Agentic guardrails, so that we can work together, so…

101
00:17:20.150 –> 00:17:24.219
Jerry Hoff: So, folks watching, make sure you have those things properly thought out.

102
00:17:24.890 –> 00:17:43.200
Jerry Hoff: Real quickly, then, on the orchestration, the only… the only layer that says what percentage, which is why it keeps, you know, attracting my eye, is that execution. Kevin Barrington, what… what is… and it looks like human only on the leadership. For orchestration and context, what is it, like, a 50-50, or does it totally depend on the organization, or just…

103
00:17:43.440 –> 00:17:45.530
Jerry Hoff: Give me a sense in terms of…

104
00:17:46.060 –> 00:18:01.590
Jerry Hoff: roles and responsibilities, the orchestration intelligence context, who’s doing what? Like you said, with agents, it’s not 80%, it’s not 90%, maybe 50-50 for the bottom layer, and then orchestration? Is that… is that mainly human, or what’s that breakdown, Kevin?

105
00:18:02.300 –> 00:18:06.029
Kevin Lee: For the bottom layer, it will eventually get to…

106
00:18:06.090 –> 00:18:20.580
Kevin Lee: mainly agent kind of impact here, and I think as we kind of move forward into the conversation, like, there’s ways to different… set different expectations with the C-level around that as well, but it… as we kind of move up the chain here, it’ll become more…

107
00:18:20.580 –> 00:18:31.399
Kevin Lee: human in the loop, and as we kind of near the base of it, it is definitely more automated, which is, I mean, honestly, what we see today across the business, when we look at

108
00:18:31.470 –> 00:18:41.949
Kevin Lee: you might be an e-commerce company or an iGaming company. The vast majority of accounts or transactions being processed through the system, whether it’s by rules or ML or something else.

109
00:18:41.950 –> 00:18:52.070
Kevin Lee: it is highly automated, and a human is not touching the bulk of that stuff. This is… I mean, what we’re talking about now is the next iteration of that, and abilities to make

110
00:18:52.300 –> 00:19:03.310
Kevin Lee: higher impact decisions without a human in the loop. But as you go up the value chain, and again, rules will change, we’ll get more human, kind of involvement, up the chain.

111
00:19:04.000 –> 00:19:05.590
Jerry Hoff: Makes sense, makes sense.

112
00:19:06.380 –> 00:19:25.629
Jerry Hoff: regarding industry, so, you know, I’ve done… obviously, I come from an information security background, and so I’ve noticed that certain organizations are better suited from an, like, almost like a temperament and infrastructure point of view. As an example, financial organizations are very used to

113
00:19:25.710 –> 00:19:34.209
Jerry Hoff: quantifying risk. So having risk quantification questions with a financial is much different than having, like, risk quantification

114
00:19:34.360 –> 00:19:35.820
Jerry Hoff: You know,

115
00:19:36.050 –> 00:19:46.939
Jerry Hoff: questions with, with other, with other industries. Is there anything like here where, from your point of view, there are certain organizations that are better suited to, to adopt

116
00:19:47.210 –> 00:19:54.290
Jerry Hoff: you know, AI in their organization, or is it just completely dependent on the individual organization as opposed to the vertical?

117
00:19:56.560 –> 00:20:03.399
Kevin Lee: Oof. I mean, from my perspective, It’s less about the industry.

118
00:20:03.970 –> 00:20:08.599
Kevin Lee: It is more about the company, because, I mean, I work with a lot of different clients.

119
00:20:08.600 –> 00:20:27.579
Kevin Lee: there are industries that are more tech-forward, let’s say neobanks and fintech companies, yes. But there are… can be more, I’ll call it, traditional brick-and-mortar companies that, from a culture perspective, from a technology perspective, they are leaning hard on this stuff, and so there… more change can be created on that front.

120
00:20:27.670 –> 00:20:31.630
Kevin Lee: But really, at the end of the day, the…

121
00:20:31.930 –> 00:20:38.039
Kevin Lee: pace at which companies are adopting this, the ones that tend to see better success earlier on.

122
00:20:38.700 –> 00:20:41.070
Kevin Lee: I mean, I think about the fundamentals, it’s around

123
00:20:41.270 –> 00:20:46.739
Kevin Lee: what data is available? How digestible is it for these models to consume?

124
00:20:47.240 –> 00:20:53.169
Kevin Lee: you can have… I’ve worked with some legacy companies that have… they’re, like, 50-plus years old, but…

125
00:20:53.170 –> 00:21:09.449
Kevin Lee: from a documentation standpoint, they did their job in terms of making this stuff consumable. It doesn’t have to be via MCP or something else yet, but they’ve structured their data in a way that these models can sit on top of it relatively easily and start consuming and making

126
00:21:09.470 –> 00:21:24.809
Kevin Lee: tangible decisions on it. And then you can work with some more startups that their data is everywhere, and they might use AWS or GCP or other things, and the systems don’t talk to each other, and so their maturity curve is slower, even though they might be a newer company.

127
00:21:26.440 –> 00:21:34.709
Jerry Hoff: Interesting, yeah. So it’s always building on top of that existing framework and organization that plays into your ability to adopt AI. That makes sense.

128
00:21:35.600 –> 00:21:40.909
Jerry Hoff: Alright, well, moving on… This is a great ques… I mean, I love this question, because…

129
00:21:41.090 –> 00:21:48.199
Jerry Hoff: you know, how can we align C-suite expectations with what we can deliver today? The reason why this, from my point of view, is so important is

130
00:21:48.240 –> 00:22:05.709
Jerry Hoff: when I speak with C-suite folks, or really anybody, on AI, there is wild, you know, speculation and expectations as to what AI can do and what it can’t do, and you’ll be some people that say, well, it can do nothing, and other people that say, it can do everything, it can run everything, and so…

131
00:22:05.910 –> 00:22:18.259
Jerry Hoff: depending on the, the expectations of your C-suite, it might be a totally different, challenge. So, fraud leaders have to manage upwards and really make sure those expectations are set

132
00:22:18.260 –> 00:22:27.710
Jerry Hoff: So, Barrington, what are your thoughts on this? How can we align C-suite expectations with what we can deliver today? What are some challenges you’ve seen in communicating that up?

133
00:22:28.330 –> 00:22:42.239
Barrington Lloyd-Lovett: I mean, I think the biggest challenge is, communicating people, like, humans’ ability to direct what agents can do, to use the AI tools that exist, to even really understand the tools themselves. Like.

134
00:22:42.390 –> 00:22:59.829
Barrington Lloyd-Lovett: what people say it can do, if you can actually demonstrate it, then, you know, that’s, I think, an expectation-setting technique. Okay, it says that it can do this, why don’t we try that? Show me how that works. Or, let’s look up even, you know, a YouTube video. Show me how these things,

135
00:23:00.110 –> 00:23:10.299
Barrington Lloyd-Lovett: Show me these things in action. So I think really going in and saying, here’s what we can do as a team right now. We just, the humans that are on it, these are the skills that we have, these are the tools that we have, this is what we can do.

136
00:23:10.530 –> 00:23:26.750
Barrington Lloyd-Lovett: Anything that an AI tool is going to be able to do is what we can do, but faster and on a broader scale. So if you wanted to do this other thing, well, can we do that with our existing toolset? Can we do that with our existing data set? Do we know people who are able to put those things together, build, you know.

137
00:23:26.750 –> 00:23:33.800
Barrington Lloyd-Lovett: even just with our existing APIs, could we put something together to be able to do that? If we can’t do that, the ITIL won’t be able to do that either.

138
00:23:34.880 –> 00:23:36.800
Jerry Hoff: Makes sense. Kevin, what are your thoughts?

139
00:23:37.570 –> 00:23:46.609
Kevin Lee: Yeah, like, to me, like, 3 main points come to mind. First is, whenever you’re working with the executive staff or the C-suite, I think

140
00:23:46.610 –> 00:24:01.360
Kevin Lee: the term AI transformation is, by now, a bastardized term. Like, it just means so many things that it can mean nothing as well. And where I’ve seen the most success is where fraud leaders really anchor on specific, measurable.

141
00:24:01.510 –> 00:24:17.079
Kevin Lee: workflows and improvements, where literally it’s a side-by-side demo of, like, look, in the current world, this is how we do things, it takes X amount of time, and in this new world, should we get the budget, because these tools can cost quite a bit of money.

142
00:24:17.080 –> 00:24:26.469
Kevin Lee: is if we invest here, these are the desired business outcomes that we can achieve. And you have that iteration and that experimentation process, that’s one piece.

143
00:24:26.720 –> 00:24:39.140
Kevin Lee: The other one, in terms of setting expectations, as you mentioned a little bit earlier, is really being explicit about what this ramp is going to be. You mentioned the Gartner study, where many, many, or the vast majority of

144
00:24:39.350 –> 00:24:47.510
Kevin Lee: experiments with AI fail. And that’s okay, because we’re still in the early innings here, and so it’s not meant to be

145
00:24:48.150 –> 00:24:53.990
Kevin Lee: a panacea of, like, yes, it’s… this is gonna be fantastic on all levels. Like, there will be…

146
00:24:54.040 –> 00:25:10.119
Kevin Lee: F-ups along the way. We may not set all the right guardrails up, initially, and so, yeah, we’re going to need, again, some more humans in loop to make sure the system is behaving as intended. But we have to be realistic about what that percentage lift is…

147
00:25:10.120 –> 00:25:17.289
Kevin Lee: expected. It’s not gonna be 100X or, you know, 200, it’s gonna be 20 or 30%, like, that’s actually a great outcome to start.

148
00:25:17.370 –> 00:25:22.130
Kevin Lee: And the final thing that comes to mind is, framing AI in…

149
00:25:22.700 –> 00:25:41.750
Kevin Lee: essentially a capacity and speed scenario where it’s not about instant headcount savings. I know a lot of different teams are kind of put into a tough spot around, cost reductions and OPEX and stuff like that at the moment, but really, again, to make it more tangible, it’s about looking at what’s the current

150
00:25:42.000 –> 00:25:54.880
Kevin Lee: Fraud volumes out there, and if we deploy particular tools, agents, software, etc, what is the expected speed advantage or capacity advantage that we might gain? It’s not…

151
00:25:54.880 –> 00:26:02.280
Kevin Lee: it doesn’t necessarily have to be, oh, if we do this, we can reduce the team by X percent. It’s more so about, hey, as a company.

152
00:26:02.350 –> 00:26:21.320
Kevin Lee: if we’re aligning with company-level goals here, in the year 2027, we want to grow by 40%, but we know that we can’t grow our ops team or our Fraud team linearly to that, and so what can we do? Where can we invest in order to scale appropriately? And it just so happens there are some agents available that can do a lot of the, kind of.

153
00:26:21.340 –> 00:26:24.979
Kevin Lee: Basic work out there, to make this

154
00:26:25.150 –> 00:26:28.320
Kevin Lee: Kind of realization happen from a company perspective?

155
00:26:30.390 –> 00:26:33.629
Jerry Hoff: AI is evolving so quickly, and again, you know, kind of

156
00:26:33.830 –> 00:26:40.200
Jerry Hoff: you know, in the spirit of aligning with C-suite expectations, you know, AI is evolving so quickly, is it…

157
00:26:41.230 –> 00:26:42.120
Jerry Hoff: How…

158
00:26:42.300 –> 00:26:55.430
Jerry Hoff: do you have any advice on where, you know, where the puck is going? We’re always having to aim where the puck is going, right? So it’s almost like we start a pilot, but as we’re doing the pilot, you know, the puck is moving, and the costs are variable.

159
00:26:55.430 –> 00:27:03.120
Jerry Hoff: So, you know, we were just talking about, you know, running out of tokens, and like, what happens when that happens? Or, you know, does it… does it,

160
00:27:03.890 –> 00:27:04.790
Jerry Hoff: Does it…

161
00:27:05.070 –> 00:27:15.469
Jerry Hoff: it has become problematic with the ROI calculation in terms of what we give to AI and what we don’t. Like, when the variable costs of tokens are factored in.

162
00:27:15.560 –> 00:27:26.900
Jerry Hoff: I know these are just kind of more, like, vague questions, but do you see any, like, any advice or any thoughts on how leaders can, can kind of juggle those, those variables?

163
00:27:29.350 –> 00:27:31.729
Kevin Lee: From my perspective, it comes down to…

164
00:27:33.290 –> 00:27:53.150
Kevin Lee: I’ll call the infrastructure that is in place to experiment, and again, kind of setting expectations around what we can and cannot do. It is vague of, like, what we cannot do, or, like, people are discovering new things all the time around, oh, I developed this agent, or someone launched a new type of model, that can do, BEC and

165
00:27:53.150 –> 00:27:54.460
Kevin Lee: D things.

166
00:27:54.520 –> 00:28:01.569
Kevin Lee: I think for my end, what has helped me and the teams that I work with is around setting

167
00:28:01.650 –> 00:28:15.639
Kevin Lee: the infrastructure in place so the data is consumable, and we can run more than just one test or two. Like, we can iterate quite quickly, and once we see improvement, and we check it, or double check it, or triple check it in some cases, then we can kind of

168
00:28:15.770 –> 00:28:24.860
Kevin Lee: move more quickly forward from there. But it is… it does involve setting aside that budget and time and resources to fail, and…

169
00:28:24.940 –> 00:28:37.610
Kevin Lee: like I said, I think we’re in the early innings of this stuff, and so we’re going to fail more than we’re successful, and that’s part of being, you know, generating different frontier models and being on, kind of, the cutting edge of things. You have to kind of

170
00:28:37.720 –> 00:28:41.960
Kevin Lee: Get some dirt on your face, kind of fighting through the forest here.

171
00:28:42.860 –> 00:28:44.240
Jerry Hoff: Love it, love it.

172
00:28:44.240 –> 00:28:49.499
Barrington Lloyd-Lovett: One way that I think can be useful to just cut down the cost of experimentation

173
00:28:49.810 –> 00:29:06.819
Barrington Lloyd-Lovett: is running things locally. You know, you don’t have to have the best model, or the most recent, or something in the cloud. If you can run a model locally, you can experiment quite quickly, and you don’t have to worry about tokens, because it’s just, you know, the cost of electricity of running your server or computer that you have it on.

174
00:29:06.960 –> 00:29:14.350
Barrington Lloyd-Lovett: I think particularly when you are very early, like, very early stage company, if you have the capacity to do that, that’ll save you a lot.

175
00:29:14.520 –> 00:29:15.839
Barrington Lloyd-Lovett: Like, a lot of money.

176
00:29:17.120 –> 00:29:28.519
Jerry Hoff: Well said. I’m a big fan of local models, and I think that’s probably worthy of its own discussion in general in the future, as local models versus frontier or, you know, third-party models. Interesting.

177
00:29:28.880 –> 00:29:34.110
Jerry Hoff: Okay, well, moving forward, What’s the most common mistake

178
00:29:34.180 –> 00:29:43.749
Jerry Hoff: Fraud leaders make when asking for more resources. I forgot who we’re on. Is it Barrington or Kevin? Barrington, do you want to take a shot at it? What’s the most common mistake?

179
00:29:43.750 –> 00:29:59.539
Jerry Hoff: Fraud leaders make when asking for more resources, and this is probably general, right? In general, especially when we’re doing back office work, whether it be security or fraud or whatever, getting that budget right and making that argument is always, it’s always a challenge. So, what are your thoughts here?

180
00:29:59.970 –> 00:30:12.530
Barrington Lloyd-Lovett: Yeah, I think, I think people, you know, when you ask for more resources, you need to be able to speak to what the return is, what the return, or what the expected return is and should be. I think particularly if you…

181
00:30:13.020 –> 00:30:14.870
Barrington Lloyd-Lovett: Are asking for resources.

182
00:30:15.080 –> 00:30:25.089
Barrington Lloyd-Lovett: you should also be able to speak to what the alternatives that you… what the alternatives are that you’ve thought of, that you’ve considered, and why you want to go with the one that you’re presenting.

183
00:30:25.200 –> 00:30:33.849
Barrington Lloyd-Lovett: I had a manager at Notion, Emma Osher, a few years back, she, she said that you should always have 3.

184
00:30:33.850 –> 00:30:49.730
Barrington Lloyd-Lovett: options when you come up. You have a strong argument for the one that you want. Humans like threes, we don’t really know why, but we do. But you should have a strong argument for the one that you want, but also, like, reasonable and strong arguments for why you should or shouldn’t go with the others, and what the trade-offs would be for each one.

185
00:30:49.730 –> 00:31:03.560
Barrington Lloyd-Lovett: And so I think demonstrating that you’ve not just thought of, oh, I need this thing because we have more tickets coming in, we need more people. Okay, well, what are some other things that you could do with that? Maybe more people will actually make it harder because you have to spend time onboarding them, and…

186
00:31:03.560 –> 00:31:10.030
Barrington Lloyd-Lovett: You know, then the tickets… the ticket spike was short, now you have extra headcount, so, like, actually having a real…

187
00:31:10.090 –> 00:31:24.660
Barrington Lloyd-Lovett: understanding of what the outcomes are, whether you need more people, whether you need more, agents, whether you need more documentation, or, you know, a BPO, actually understanding what you’re getting out of it, I think is really important.

188
00:31:26.060 –> 00:31:30.270
Jerry Hoff: Nice, nice. I have your RY and 3. Sorry, go ahead, Kevin.

189
00:31:30.790 –> 00:31:34.559
Kevin Lee: No, it’s, the only thing I’ll kind of add to that one is.

190
00:31:35.400 –> 00:31:49.090
Kevin Lee: Barrington mentioned kind of, like, the three options. Oftentimes, the fourth option, or the… maybe one of the three options, might be the do nothing, and kind of keep the status quo. What I think is also important is to call out what’s

191
00:31:49.420 –> 00:31:56.720
Kevin Lee: already been optimized, or what we’ve tried. So, have we reached the ceiling? Oftentimes, I’ve seen pushback from…

192
00:31:56.720 –> 00:32:15.169
Kevin Lee: executives around, okay, whether it’s people or agents, whatever you want to invest in, but are we making the most of what we have already? I’ve definitely been in conversations with… it’s almost like a Tulapalooza. It’s like, oh, we just need this tool and that tool, and pretty soon, you have, like, 50 different tools, and you’re using

193
00:32:15.300 –> 00:32:29.150
Kevin Lee: 10% of it each one. It’s just incredibly difficult to manage, especially from a cybersecurity standpoint. And what I would recommend also, when you do make the ask, is we’ve essentially coming up with the argument, like, hey.

194
00:32:29.540 –> 00:32:48.080
Kevin Lee: executive staff, we’ve maxed out how far we can go with the current toolset. It’s not that we’re, like, not using the tools that we’ve been given, but we are using it to the max, and now, in order to get to the next set of desired distance outcomes, we’re gonna need, you know, D, E, and F items.

195
00:32:49.800 –> 00:33:03.220
Jerry Hoff: And so those can be put in pretty quantified terms, I’m assuming, right? Because you can say, hey, if I have these additional tools, I can reduce fraud, and we can realize a greater percentage of profit, whatever it happens to be, pretty easily. It’s, you know, it’s…

196
00:33:03.430 –> 00:33:23.040
Jerry Hoff: it should be able… do you… I guess you would also have to give a timeline for these things. It would take… there’s some lag time for these tools to take effect, I would assume. And also, some of those tools would have to be in place even to adopt an AI strategy, like what we talked about before. You have to have your house in order before you can start automating. Are those all things that,

197
00:33:23.340 –> 00:33:26.300
Jerry Hoff: You know, kind of, where do you draw the line in,

198
00:33:26.700 –> 00:33:39.769
Jerry Hoff: when you’re explaining, I guess, to the CFO or to the C-level on budgets, where do you draw that line in terms of technical, like, planning? Do you explain all of that, or do you try to give a high-level

199
00:33:39.880 –> 00:33:42.019
Jerry Hoff: Perspective, or what are your thoughts there?

200
00:33:42.640 –> 00:33:51.999
Jerry Hoff: I speak as somebody who’s done quite a number of budget requests, and it’s never easy, it’s never a smooth thing, so look, just getting… want to get your thoughts.

201
00:33:53.340 –> 00:33:59.020
Kevin Lee: In my experience, let’s say we’re chatting with the CFO, Really, it comes down to…

202
00:33:59.570 –> 00:34:13.470
Kevin Lee: and I’ll keep it… I tend to have greater success keeping it simple, because they’re not going to understand all the ins and outs of the tools. Really, they’re… they’re concerned about the investment, and kind of what ROI are we going to get back on that thing.

203
00:34:13.469 –> 00:34:21.730
Kevin Lee: And so, I typically frame it into, let’s say, what are the initial investments? It’s kind of like the table stakes that we need. Oftentimes, that’s just around

204
00:34:22.090 –> 00:34:32.810
Kevin Lee: making the data sortable, digestible for whatever sits on top. It might be humans, it might be LLMs, it might be something else. Understanding what that

205
00:34:32.949 –> 00:34:46.460
Kevin Lee: is what’s needed, and maybe, again, there’s some shortcomings there, and what that can potentially unlock. And this is where it comes down to setting expectations as well. Doesn’t mean, like, oh, we get this additional piece of infrastructure, it unlocks, like, kind of…

206
00:34:46.600 –> 00:35:05.369
Kevin Lee: the Emerald City here, like, it’s not gonna get us all the way, all the way there, but, it gives the CFO, and typically CFOs are more conservative when it’s session when it comes to budgeting, so you want to build in some buffer there. Having that kind of frank discussion, and then, if there are

207
00:35:06.330 –> 00:35:17.889
Kevin Lee: the tool that’ll sit on top of that, what additional unlock that’ll… what will we get, reasonably? Because some of this is a swag, to be honest, right? Like, how much, kind of, impact is it going to make?

208
00:35:17.890 –> 00:35:29.319
Kevin Lee: But where I found the most success is tying that back into the company objectives. So, many companies right now are planning for 2027, and figuring out headcount budgets or other OpEx budget things.

209
00:35:29.460 –> 00:35:36.899
Kevin Lee: It’s what are the biggest tools or peoples, sometimes there’s a reorg involved, that will help the company get to

210
00:35:37.490 –> 00:35:48.239
Kevin Lee: whatever pillar that, is, set for 2027, and I found that to be the most… one of the more compelling things that’ll get more sign-off from a CFO.

211
00:35:48.960 –> 00:35:56.349
Jerry Hoff: Gotcha, gotcha. Great. Well, we only have a few more minutes, so there is a few other questions that we wanted to go through.

212
00:35:56.490 –> 00:35:59.129
Jerry Hoff: So, next question.

213
00:35:59.580 –> 00:36:12.960
Jerry Hoff: this is very timely, right? We’re hearing about, unfortunately, this in the news quite a bit. So if a team has to reduce capacity, how can leaders identify where cuts will do the least harm? And I really like the way that, you know, we phrased that here, because

214
00:36:13.130 –> 00:36:31.090
Jerry Hoff: I’ve seen slash and burn type organizations where you’ve got all sorts of tasks and responsibilities that, you know, are kind of dangling in the wind there. So, Kevin, what are your thoughts in terms of reducing capacity? How do we make sure that we identify where the cuts do the least harm in the fraud world?

215
00:36:31.920 –> 00:36:36.310
Kevin Lee: Yeah, so, I mean, a couple things to mine… come to mind, and these are…

216
00:36:36.530 –> 00:36:42.220
Kevin Lee: really difficult discussions to be had. It varies by… I mean.

217
00:36:43.000 –> 00:36:46.310
Kevin Lee: People are the solutions in all these things, but…

218
00:36:46.750 –> 00:36:58.950
Kevin Lee: in our industry, in the economy, etc, like, difficult decisions need to be made, and the more information that one has, like, the better educated decision one can make. It may not be…

219
00:36:59.560 –> 00:37:04.100
Kevin Lee: A fun thing or anything like that, but it is something that,

220
00:37:05.070 –> 00:37:21.010
Kevin Lee: you might realize, after doing the research is like, yes, I can, quote-unquote, afford to lose X number of people, or trim the team here or there. To me, what’s… in just chatting with peers in this space, and other executives.

221
00:37:21.010 –> 00:37:38.679
Kevin Lee: really, it comes down to mapping the work by risk and value, and not by volume, per se. So, whether it’s the cues, the reviews, like, what actually prevents the most loss? So, if we’re talking really about fraud, where… what drives the needle when it comes to straight

222
00:37:38.690 –> 00:37:54.939
Kevin Lee: money. And it might be a few accounts that are high dollar, it might be a lot of lower, like, accounts with lower dollar values, but really what drives the dollar figure here? And again, if we’re talking to the CFO, and usually it’s about budgeting here, what is going to cause the least harm?

223
00:37:55.100 –> 00:38:09.169
Kevin Lee: In that realm, and also setting expectations with eStaff, or the executive staff here, where whatever cuts are made, again, this is assuming that we’ve gone through the optimization stuff, like, we’re trying to get the most out of the existing toolset.

224
00:38:09.570 –> 00:38:13.449
Kevin Lee: It really comes down to… Setting the expectation of.

225
00:38:14.780 –> 00:38:19.929
Kevin Lee: Should we make these tough decisions and, you know, lose some folks off the team?

226
00:38:20.130 –> 00:38:30.930
Kevin Lee: we are going to have to stop doing something, and that stop doing something may lead to more financial loss. And that sometimes can turn the tide of the discussion of.

227
00:38:31.520 –> 00:38:33.889
Kevin Lee: Oftentimes, it can be a zero-sum game, where…

228
00:38:34.030 –> 00:38:39.760
Kevin Lee: I’ve had discussions personally on the fraud side, where it’s like, hey, we might need

229
00:38:39.850 –> 00:38:49.459
Kevin Lee: to retain the people directly on Team A, and Team B, we actually can be a bit more aggressive on the cut, so it’s not like

230
00:38:49.460 –> 00:39:00.329
Kevin Lee: all teams must reduce by 20%. There are some teams that are more impacted, and some that are definitely less impacted, in that, in that, kind of in those discussions.

231
00:39:01.720 –> 00:39:18.020
Jerry Hoff: Agreed. And, Barrington, I remember we had spoken a little bit before the webinar, and you were talking about retraining, and, you know, we were talking about this, you know, kind of tendency to over-hire, and then over, you know, over-reduce, and we have this kind of

232
00:39:18.020 –> 00:39:24.650
Jerry Hoff: wave. What are your thoughts in terms of, like, reinvestment or leveling up the team and so forth?

233
00:39:25.010 –> 00:39:32.639
Barrington Lloyd-Lovett: Yeah, you know, one of the things I think is a good hiring strategy overall is you hire for…

234
00:39:32.950 –> 00:39:36.479
Barrington Lloyd-Lovett: People who have the interest and capacity to do good work.

235
00:39:36.640 –> 00:39:53.779
Barrington Lloyd-Lovett: And maybe they have experience doing the work that you need them to do. But it’s less about, do you have this experience, and more, are you the kind of person that can do this job and can learn? So then if it comes to a case where, okay, for whatever reason, this team needs to reduce capacity, or we’re, you know, we’re not seeing the need to have

236
00:39:53.780 –> 00:40:00.070
Barrington Lloyd-Lovett: this number of people over here. Okay, are there other areas in the company? Are there other areas…

237
00:40:00.070 –> 00:40:05.480
Barrington Lloyd-Lovett: that we haven’t been touching? Are there other, problems that we haven’t solved? Can we…

238
00:40:05.900 –> 00:40:20.270
Barrington Lloyd-Lovett: redirect those people in that… in that way. So you don’t necessarily… and sometimes you do just ultimately have to make a decision, we have too many people, this person’s not performing, whatever the case is, but if there’s an opportunity for someone that you want to retain.

239
00:40:20.710 –> 00:40:40.230
Barrington Lloyd-Lovett: to have them do something else, or build a new function, or whatever the case is, there’s an opportunity then to train somebody in a new skill set. And that both gives them the ability to, you know, do something new, but also you give the signal to them, and to the team, and to the company that we are here to be successful as a group.

240
00:40:40.310 –> 00:40:43.640
Barrington Lloyd-Lovett: Even if what we’re doing day-to-day changes.

241
00:40:45.230 –> 00:40:57.609
Jerry Hoff: I mean, organizations have short-term, medium-term, and long-term financial realities, right? So we can’t take that away, but at the same time, I’ve seen this, and just like both of you have seen, I’m sure, as well, this, you know, this forever curve of

242
00:40:57.670 –> 00:41:08.830
Jerry Hoff: lots of hiring, and then, you know, letting go, and the thing is, every time you get somebody new into an organization, there is a lag time. It takes them a couple months to understand how the company works, how the…

243
00:41:08.830 –> 00:41:18.420
Jerry Hoff: you know, what are the internal processes, and so forth, so letting people go, it just keeps setting the clock back. So I… I think… I like the idea of

244
00:41:19.110 –> 00:41:36.450
Jerry Hoff: finding ways to keep the staff that you have. Maybe not hiring new, but I wouldn’t preemptively or rationally let a lot of folks go, because I think that just the demand for people is going to be growing, just as AI is almost going to whet our appetite for more and more in every company.

245
00:41:36.720 –> 00:41:37.840
Jerry Hoff: That’s how I see it.

246
00:41:38.790 –> 00:41:52.009
Kevin Lee: Yeah, that’s actually… literally, I was chatting with someone last night about this, where there is the old adage, good people are hard to find, and the… to Barrington’s point, the capacity to learn

247
00:41:52.200 –> 00:41:59.329
Kevin Lee: is… is critical, and they’re always, like, 2 months or more in terms of ramping up a new person, and where…

248
00:41:59.520 –> 00:42:14.729
Kevin Lee: like, certainly other companies that, like, this peer that I was talking to last night, they had a riff and go through the process, and, like, 6 months later, they rehired some people and, like, things… and they rehired some new people as well, but they weren’t that different from.

249
00:42:14.740 –> 00:42:24.770
Kevin Lee: The folks that were let go, and literally the discussion we had was, like, if we were to retain this person, and do we believe that

250
00:42:25.100 –> 00:42:29.400
Kevin Lee: Joe, or Bob, or Jill, If they can learn that new skill.

251
00:42:29.650 –> 00:42:36.140
Kevin Lee: Then it… it changes the game, and identifying those good people,

252
00:42:36.140 –> 00:42:54.220
Kevin Lee: is, like, still remains a key focus area, and granted, not every company can afford to make that decision and kind of hold on X number of months, or whatever it might be, but that’s something to keep in mind, where that’s always from… I see it as, from a leadership perspective, as a people manager perspective.

253
00:42:55.220 –> 00:43:10.319
Kevin Lee: always be identifying those folks, and giving them the runway to learn new skills, and hopefully apply those skills to challenges or problems that customers are facing, or the business is facing, and give them the runway to experiment on that. It’s not just about

254
00:43:10.970 –> 00:43:16.560
Kevin Lee: mapping out their day-to-day to the nth degree, like, they should have the… autonomy and the…

255
00:43:16.730 –> 00:43:20.019
Kevin Lee: Kind of authority and expectation to…

256
00:43:20.180 –> 00:43:26.279
Kevin Lee: Explore and learn new things so they can be more efficient, and that often involves learning new technical skills.

257
00:43:26.570 –> 00:43:33.320
Jerry Hoff: I love it, I love it. We only have about 10 seconds left, so I have one last question. So this is lightning round, right? Just 5 seconds each.

258
00:43:33.440 –> 00:43:37.720
Jerry Hoff: Kevin, what should remain under human ownership or oversight?

259
00:43:38.920 –> 00:43:50.300
Kevin Lee: My 5-second answer would be high… anything that’s high stakes and irreversible decisions that might come into, like, policy, strategy discussions, we still need humans in that mix, for sure.

260
00:43:50.570 –> 00:43:54.560
Jerry Hoff: Love it. Barrington, what needs to remain under human ownership or oversight?

261
00:43:54.560 –> 00:44:05.479
Barrington Lloyd-Lovett: Humans should always understand the core work, and anything that is two escalation levels higher can take the customer off the product, or can take the product offline.

262
00:44:05.860 –> 00:44:15.180
Jerry Hoff: Love it. That is all the time we have for today. Thanks, everybody, who joined us. We had a great crowd, and we’ll see you at the next Blueprint series. Thank you all very much.