Delivering Email People Actually Want – Laura Atkins, Email Deliverability Expert

In this episode of Collateral Dialogue, host Ken Simpson delves into the intricate world of email deliverability with Laura Atkins, a renowned expert with over two decades of experience in the field. Laura, an anti-spam and email deliverability specialist, shares her journey from molecular biologist to a leading figure in the email industry, highlighting her contributions and experiences that have shaped today’s email landscape.

Starting with her early encounters with spam in grad school, Laura narrates her transition into the world of email, discussing her role at MAPS (Mail Abuse Prevention Service), her insights into the development of email technologies, and her eventual foray into consulting. She offers a unique perspective on the evolution of email deliverability, discussing the early days of fighting spam, advising major companies, and the significant changes in technologies and strategies over the years.

Laura also sheds light on her current consulting work, helping large organizations and government entities navigate the complexities of email strategy and compliance. She emphasizes the importance of sending communications that recipients actually want, a simple yet profound principle in her approach to deliverability.

Listen here

Watch here:

Read the transcript:

[00:00:00] Laura: The idea is, and the way I approach deliverability, and this is my anti spam background coming out, is that I don’t care what you send. I just care that your subscriber wants that. That the person that you’re sending mail to wants it. You know, I want them to get the mail that they want, that they need, helps them get the information that they want.

[00:00:24] Laura: That’s who’s side I’m on. Yeah. There are bad guys out there that, I mean, you talk to some of the folks from Google and they will tell you absolutely that, you know, it was a security compromise that changed the 2016 election in the U. S.

[00:00:39] Ken: My friends, today we have a very special guest joining us, Laura Atkins, an expert in the world of email deliverability and anti spam with over two decades of experience.

[00:00:51] Ken: Laura has seen it all, from the early days of fighting spam in the 1990s. To advising major companies and government organizations today [00:01:00] on email strategy and compliance. She offers a unique perspective that I’m excited to explore. The viewpoint of someone fighting to keep inboxes free of unwanted messages so that people can communicate efficiently.

[00:01:14] Ken: Laura has been at the forefront of the email industry’s evolution for over 20 years. She has seen technologies come and go, companies rise and fall. Her principles remain steady, that we should send communications people want to receive. So simple yet profound. I can’t wait to explore her unique journey and perspective on today’s email landscape.

[00:01:37] Ken: The stories she has to share are guaranteed to enlighten and intrigue. So let’s jump right in.

[00:01:57] Ken: Laura, why don’t we start, if you don’t mind, telling [00:02:00] us where you are sitting right now. Where in the world are you?

[00:02:05] Laura: So I am in, uh, South Dublin, Ireland. So we moved here back in 2018 and we bought an old Victorian and spent a lot of COVID, um, dealing with contractors trying to renovate it and finally moved in the end of last year.

[00:02:23] Ken: Wow. Excellent. And, uh, what is your favorite?

[00:02:31] Ken: Um,

[00:02:33] Laura: I really like the people, you know, everybody is generally friendly. Everybody is generally helpful. It’s, it, it, it feels much more of a community and it feels much more relaxed than it feels much more, um, softer almost than where we were before, which was in the Bay area and, you know, it was tech bro central cause we were right outside Palo Alto and, And it’s nice not being quite so tech centric.

[00:03:03] Ken: You know what, that actually gives us a perfect place to start, which is talking about your history in Silicon Valley before you made the wise choice to move to the Emerald Isle. Um, so, uh, can you maybe spend a little time sharing some insights into your journey? Uh, from Silicon Valley to Ireland. And, and can you tell us what the deal is with Terskar?

[00:03:28] Ken: Um, anyway, maybe start by, maybe start by talking about Silicon Valley and maps and, you know, how you got into this whole email space. Absolutely.

[00:03:38] Laura: I can do that. I think we need to actually start back a little bit before I managed to make it to Silicon Valley. Um, I’m actually a molecular biologist by training.

[00:03:47] Laura: So I, when I was in, um, high school, right after I got out of high school, I got an internship at the FDA doing, um, fancy, this fancy molecular biology research. Um, and [00:04:00] then kind of went through college, graduated from college, worked at the FDA for a year, then ended up in grad school up at the University of Delaware, uh, developing respiratory vaccines for chickens.

[00:04:14] Laura: So I’m, I’m well into the respiratory vaccine thing, um, and then, you know, kind of one thing really led to another, but while I was at the University of Delaware, I got my first piece of spam. So, you know, I had had an email address for five, six years by then, um, but I got my first piece of spam and it was fairly early on, it was about 1995.

[00:04:39] Laura: Um, maybe 96,

[00:04:41] Ken: I don’t think I got any spam in 1995.

[00:04:46] Laura: I was outside Philadelphia, which was where Sanford Wallace was. And so there was kind of like this whole cluster of local things. And I had a udell. edu address, so it was local. Um, and Delaware at that time didn’t do usernames. They did numbers. So you could just like literally sequence through the numbers and hit everybody’s.

[00:05:07] Laura: Um, so anyway, so I was in grad school at University of Delaware and got this first piece of spam and then started with, okay, we have to stop this because like, spam is annoying, um, and kind of as a hobby, joined some of the anti spam groups on Usenet and some other things, um, learned how to read headers, did all of that, um, meanwhile, I now left the University of Delaware and ended up in Wisconsin, the UW Madison, um, molecular biology again, but this time we were studying embryology and fetal alcohol syndrome.

[00:05:40] Laura: Um, but again, still kind of in that whole space of fighting spam, working with folks to… Protect email. We didn’t have email filters early back then. There was no filter. There was no spam folder. The spam folder didn’t come out until like 2000 or 2001. So this was like, this was before we had any of the tools that we have.

[00:06:03] Laura: And it was all kind of on the individual to police their mailbox and do all, and report the spam and read the headers and do all of that. In the course of all of that, I met Steve. Who is now my husband. Um, and we actually met, we did actually meet at an anti spam IRC channel. And… Wow. Decided that he was, he, at the time he was in Boston.

[00:06:26] Laura: So he was working at, um, AMD? No, that wasn’t his AMD. That was Hewlett Packard. Or whoever they were before they were Hewlett Packard. Um, anyway, he was designing microchips in Boston. Um, so I ended up getting a job at one of the hospitals. associated with one of the large universities in Cambridge, Massachusetts, um, to, again, do molecular biology.

[00:06:55] Laura: And after we had been there for a while, and again, you know, we were kind of all doing this anti spam stuff on the side, Steve ran the samspade. org website, and all of that kind of thing, um, He got offered a job in California. So this was, this was early 2000. He got offered a job out in California doing architecture and design for a company, for a startup.

[00:07:16] Laura: So we got up, moved to California. Um, and then I kind of poked around in the biotech industry a little bit, but I’d never actually finished my PhD. And so it was kind of like, I didn’t have quite, you know, all of the letters after my name to get a really interesting job. And so I was kind of being picky about things because I had some really good research positions and had my own projects, even though I didn’t have a PhD and all of that.

[00:07:41] Laura: Um, and then I got offered a job at MAPS. And MAPS is the Male Abuse Prevention Service. They were the first commercial block list. Um, they ran the real time blockhole list. They ran a couple of the other, the first dial up list. They [00:08:00] ran an open relay list, um, that they kind of acquired in from various folks.

[00:08:04] Laura: Um, and it was, uh, it was about, by the time, at our biggest, we were about 35 or 40 people. Wow. And we were all kind of united in this idea that spam was bad and spam was a problem and we should address it. And we were addressing it kind of in two different ways. So, there was the block list side, um, that was run by, uh, a bunch of folks, and, you know, they would take in nominations, at that point, you could, like, write up a nomination, and nominate somebody to be listed on the RBL, and then they’d go through the nomination, and they’d contact the company, and they’d try and work with them, and if it worked great, they didn’t get listed, if the company was like, no, no, we’re gonna keep spamming, um, then they would get listed.

[00:08:52] Laura: I was actually working on a different side of things. I was working on the, uh, customer services side. So our [00:09:00] whole idea was that we were going to stop people from spamming before they started. So that we were going to come in and we were going to teach people how not to spam. So that they didn’t even get near the block list kind of thing.

[00:09:12] Laura: Um, and I did that for about a year and we did, we built out a good team. We built out, um, we actually ended up The first round of stuff that we did was running an abuse desk that we were running the abuse desk for a major network services provider. We had customers like GeoCities. We had Napster as a customer.

[00:09:29] Laura: We had eBay as a customer. Um, so we had, you know, we were a big network provider sitting in Palo Alto and Um, I was running the abuse desk. At that point, we were in early 2001. So we’re, we’re at the point where the Nasdaq is crashing and the dot bomb is happening and everybody’s getting laid off. And that happened to Maps.

[00:09:53] Laura: Um, and Maps decided that the customer service side wasn’t really working and it wasn’t a liable business model. So they were going to lay off the customer service side. So what happened then was the chairman of the board of the company that Steve was working for is Rodney Jassy. Who is big in the marketing space and he does a lot of things with marketers and he had a lot of contacts in marketing.

[00:10:17] Laura: And so he said, I’ve got some friends who are having problems with the RBL. Why don’t you, why don’t you go consult for them? And so that was basically my first consulting gig was, you know, from Rodney. And I did that. It was a company in Georgia and they were sending lots and lots of email, lots and lots of email at that time.

[00:10:37] Laura: Um, I spent four days down in Georgia and fixed all their problems and we went through a whole bunch of stuff that they were doing and thought, well, I could do this on my own. You know, we couldn’t make it work at maps, but maybe I can make do this on my own. So that was the start of the consulting side of the business.

[00:10:54] Ken: Um, so literally we got to, you were literally there at the [00:11:00] dawn. Of what we would, you know, perhaps called deliverability consulting, right? Yeah. Uh, and what an interesting time. I just want to acknowledge what an interesting time to be in Silicon Valley. I mean, I, I remember going down to the Valley in 2002.

[00:11:21] Ken: And it was, I was in South San Francisco for a meeting and literally every office was empty. And the guy that I was meeting with. He got his office space for free because the landlord just needed to fill it with someone. I mean, I guess it’s a little bit reminiscent of the, of the pandemic period, the post pandemic period, sadly, but at that time it was, it was crazy.

[00:11:43] Ken: There was no traffic on the one on one. It was absolutely a ghost town. And so many interesting things got born at that time. Obviously one of which was your consulting. Yeah. So you started consulting, uh, in email deliverability, helping, uh, companies. So, you know, where, where did it go from there? What were some of the next steps, uh, as some of the things that happened early on, like how did it evolve?

[00:12:11] Laura: So, from that point, you know, I, I was picking up some, some different gigs. Um, you know, I had colleagues that had moved on and done other things, um, were recommending me to various. of their companies. Um, I got a couple very early, big customers, um, one of which was actually, one of my very first customers was ExactTarget.

[00:12:35] Laura: Um, and we came in, and they just, basically, it wasn’t even deliverability consulting, they just needed somebody to kind of, what we were doing is, we were looking at logs, so we had a log parser, and we were paying attention to what was delivering and what wasn’t. Um, and we helped them set up some of the policies.

[00:12:55] Laura: For their compliance desk and how do we make sure that our customers are doing the right thing and you know, it really, at that point in time, deliverability wasn’t a thing. Um, you know, and it was, it was, you’re either listed on the RBL or your delivery is fine. This was kind of the very early spam folders.

[00:13:17] Laura: I remember I was, um, speaking at a conference in Aspen, Colorado in 2003, I think it was. Might have been 2004. Um, but Yahoo! had just… Released the, this is spam button. And when they just released the, this is spam button, one of the things they did to, to teach people to use it was they had a contest, so they had it and it, and entering the contest was reporting something as spam, Oh, the marketers at this conference, cause I was up and I was on a panel with a bunch of people talking about inboxing and deliverability, not that we called it that.

[00:13:57] Laura: Then, but, and they’re like, how can Yahoo is incentivizing people to report our mail and it’s like, well, yeah, but that doesn’t mean that that’s actually going to affect where your mail is delivered or all of that kind of thing. And of course, this was pre FBL, so there were no FBL. So you didn’t even know who was reporting it.

[00:14:16] Laura: It was all Yahoo internal. So, yeah, it was a really interesting time and there’s a huge amount of stuff that happened and, you know, like between about 2000 and about 2005 or six that have influenced what is happening today, but it’s really not as much people don’t realize kind of where we were back then and what tools we had and what, what tools we didn’t have.

[00:14:45] Ken: Right. I, my recollection is that. The spam, spam sort of became a problem, a real problem around 2001 or 2002. Like for me, that’s when I started actually seeing it in my inbox [00:15:00] and being like, Oh, this is actually a bit of an issue. And that’s when the first companies really popped up to fight spam. Uh, you know, companies like proof point, for example, emerged around that time.

[00:15:10] Ken: Um, and, and I was working for a software company that. Almost by accident made a spam filter just as a sort of demo for a, for a customer, and then that became their whole business and they got acquired for that. Um, which is how I ended up in this space. Uh, but yeah, that was very interesting. You mentioned there weren’t even.

[00:15:32] Ken: FBLs, or for people who don’t know what that means, feedback loops. So feedback loops are a way that all the major providers, uh, tell senders about the spam that they’re seeing, the spam reports that they’re seeing from that sender’s, uh, IP addresses and domains. And That’s a pretty key piece of technology, because without that feedback information, uh, senders really don’t have a lot to go on, um, in assessing whether the mail they send is, is wanted or not.

[00:16:02] Ken: I’m sure you’ll get into this more later, but there are other things you can do. Um, and, and it’s interesting, sort of, I’m skipping ahead, but very recently, there have been some big changes in how feedback loops work. Um, you know, for one thing, you’re going to have to start paying for many of them if you’re a, uh, a big sender.

[00:16:21] Ken: Um, but, uh, okay, cool. So carry on, carry on with the storytelling. This is fascinating.

[00:16:32] Laura: So we really did kind of start to develop these kinds of tools and we started to develop kind of some structure around what we were doing because a lot of it was really ad hoc and a lot of it was, who do you know?

[00:16:44] Laura: Um, and, you know, it wasn’t as much that, you know, and I, I knew a lot of people because when maps kind of broke up. Um, a lot of the folks at MAPS kind of spread out, um, JD Falk was one of my MAPS colleagues who ended up at Microsoft and Yahoo and then eventually Return Path, um, and he was kind of, he was really influential and in fact he predates me in the, in the industry, um, by a couple years, I first met him back in the 90s on One of the anti spam groups, um, you know, but there were a lot of people in MAPS who kind of then moved out into the anti spam and, and the deliverability space, not quite as much.

[00:17:26] Laura: Um, I think I was really the one who kind of ended up there and everybody else kind of went over to work at abuse desks or a lot of them went to the ISPs and to actually block spam on the inbound.

[00:17:36] Ken: Or some of them went and worked at Facebook and just became rich and are retired now, in my recollection.

[00:17:44] Laura: No, he was at AOL. He didn’t go, he wasn’t at MAPS. Oh, okay.

[00:17:48] Ken: Okay. Miss some, some other people. Yeah. I think,

[00:17:53] Laura: I, I don’t think any of the MAPS folks actually ended up at Facebook. Yeah. Um, [00:18:00] but yeah, and so we all kind of knew each other and we all kind of, um. You know, a lot of, a lot of the core originally really did grow out of that Usenet community and out of SpamAl and some of the other private mailing lists that were invite only and had to be approved and all of this kind of thing.

[00:18:19] Laura: But a lot of those folks, you know, the people that I was hanging out with kind of both socially and professionally at the time. You know, they’re now VPs. They’re now, you know, Chief Privacy Officers. They’re now founding anti spam companies. They’re, they’re really going on. And so, Maps was kind of, in a lot of ways, it was influential in that it gave a lot of people a start.

[00:18:42] Laura: Now, I think, in some ways, folks would have ended up in this space anyway, because they came from the tech side. So, they came from, you know, I run a mail server side. They came from, I wrote this piece of code side. Um, me, I would probably still be in a lab somewhere slinging DNA around.

[00:19:01] Ken: So also, also, you know, would have been a very healthy pursuit and very interesting and all that stuff, you know, making a difference in the world, uh, just in a different way.

[00:19:10] Ken: Yep. Um, so, okay. So, you know, coming up to, uh, the present, uh, What do you do now? Uh, you know, who’s your, who’s your typical client? I honestly, I’ve been in this industry, um, since I guess pretty much technically 2003, 2004, uh, and running a service since 2012. Um, so really exposed to deliverability issues.

[00:19:41] Ken: For a little over a decade, you’ve got way more exposure. And I feel like there’s a whole side of the industry that I just don’t even understand yet, uh, amazingly. And that’s, what is the life of, uh, what is the life of people who have to deliver large amounts of mail that they really hope people want to [00:20:00] receive, but it’s not always the case.

[00:20:02] Ken: And how do you help them? Like, what do you do? Who are your customers?

[00:20:07] Laura: So more recently, a lot of my customers have been, you know, Big organizations, um, you know, uh, branches of government level, um, big social media companies. Um, groups like that who it’s not as much about deliverability, you know, I’ve, I’m kind of not in the day to day.

[00:20:29] Laura: Oh, let’s look at your bounce logs. Let’s do this. Let’s do that. What I’m more doing is coming in and they’re saying, okay, so we’re on these three ESPs and we want to consolidate down to one or. We are going to buy our own IP space and we need to set that up and manage the transition across. Um, or, you know, we are a large, uh, app provider that, you know, sends 200 million emails a week.

[00:20:58] Laura: Um, because we send, you know, reminders every day that you need to log in and do your thing. Um, And we’re having some problems with deliverability, and we need somebody to come in and look at what’s going on and how we can fix things, um, and what kinds of changes we need to do. So it’s, it’s been as much about business consulting, you know, what are your processes?

[00:21:24] Laura: What are your, um. goals as a company. And I know, you know, it’s like to make money and to get bought out or to IPO or whatever. But it’s more about, you know, who, who are you as a company to your customers? What, you know, what persona do you want to project to them? And so it’s You know, I also, I’ve also had a couple come in from the other side from, who want compliance work.

[00:21:50] Laura: So they’re, we’re an ESP or we provide, we’re a SaaS provider and we let people send email through our systems. And how do we manage that and [00:22:00] how do we interact with our customers such that they’re not going to do anything really bad and destroy our reputation or their reputation? Um, but really kind of that idea of, we want to make sure that what we have in place Supports the idea of reaching the inbox rather than, um, you know, okay, well we’re going to publish these four DNS records and we’re just going to go ahead and do things, but it’s a, it’s a broader, more holistic view of deliverability as a business problem and a business value rather than, you know, we got this much mail to the inbox.

[00:22:44] Ken: So, uh, yeah, so if I can reflect on what you said. It, it seems like perhaps in the past, the reasons why companies might’ve come to you is because, oh, we’re getting blocked by SpamHouse or, you know, we’ve got this recurring issue where we’re getting problems delivering to Yahoo. Can you please help us? And but nowadays it sounds like things have evolved to a higher level, kind of a more, uh, holistic, as you said, holistic, uh, business problem where you help them define, well, what, what are you to your customers?

[00:23:18] Ken: You know? Are you the kind of company that is going to cold prospect people and, you know, even if it’s kind of spammy and, uh, or, or what kind of, uh, how are you, how do you want to come across to the market, uh, and then derive an email strategy based on that? Because I think some of the basics of email deliverability from a technology perspective, they seem to be table stakes.

[00:23:43] Ken: Now, like, anyone can go online and, you know, get some tips on how to have a good, you know, clean, uh, email sending approach, how to prune your lists to make sure that you don’t have really old subscribers on there, how to look at your engagement statistics to see which segments aren’t performing well and clean them up, that kind of thing.

[00:24:03] Ken: I feel like that’s more table stakes now, but the, but maybe the strategy is where the problem is. That’s what you’re tackling.

[00:24:12] Laura: Yeah, you know, it was, it was probably about a decade ago that I started looking at, okay, what the idea is, and the way I approach deliverability, and this is my anti spam background coming out, is that I don’t care what you send.

[00:24:32] Laura: I just care that your subscriber wants that, that the person that you’re sending mail to wants it, and wants is kind of this really broad, broad thing. It could be as much, you know, yeah, okay, so I like this company and I’ll get their mail and every, you know, six months or a year or whatever, I’ll buy something.

[00:24:55] Laura: Or, you know, whatever it is, I don’t mind that I’m on [00:25:00] this list. I don’t mind that they continually mail me. Right. The idea, in fact, I was on a, I was on an industry call once, and we were talking about something that the ISPs were doing, and senders were kind of cranky, and I was like, well, but, okay, let’s reframe this, and let’s go ahead and change all of this.

[00:25:17] Laura: And somebody snapped at me, whose side are you on anyway? Because I was like, trying to explain, you know, why the ISPs might be making these decisions and asking for these things. And I’m like, you know what? I’m actually on the side of the recipient, you know, it, it is, it is that end user who is in a Gmail mailbox or who is just trying to live their life and is trying to do things and uses email as a tool and that’s whose side I’m on, you know, I want them to get the mail that they want, that they need, that they, you know, helps them buy things they want, helps them have experiences that they want, helps them get the information that they want.

[00:25:55] Laura: That’s the side I’m on.

[00:25:58] Ken: These days, I think it was, it must have been like at least seven years ago that, uh, there was a talk by the lead engineers from a major mailbox provider at a conference that shall be nameless. And they were talking about how their machine learning gurus, uh, gave them this new neural network, this magic box.

[00:26:25] Ken: And it took their filtering accuracy up by a factor of 10, uh, and how it basically worked is they threw in all of the statistics about how people were interacting with the mail they received and it would just go into this magic box and based on that the magic box would say you should filter this to the spam folder or you should allow this to be delivered and they didn’t really know what happened inside the box.

[00:26:52] Ken: Um, but I feel like, you know, if it’s been seven years since that talk, things have only gotten a whole lot more powerful and [00:27:00] really what these neural networks are doing on the receiver end is they’re acting as a proxy for, uh, the likeability of your content, right there, they deeply understand somehow whether what.

[00:27:14] Ken: You’re sending to the inbox is stuff that that particular recipient really wants to receive. It can anticipate that it can anticipate the behavior of the human, which is incredible, but you look at things like GPT for, and it doesn’t make, it’s not surprising that that’s possible these days. So, yeah, as, as if you’re, if you’re someone trying to deliver stuff, I like your idea of just, uh, trying to, uh, teach companies that you should send stuff that people actually want to receive.

[00:27:43] Ken: They may not be jumping up and down. excited about it, but it’s like, yeah, once in a while, uh, I get a mail from this company and I’m like, cool, I can buy a new kind of vacuum cleaner or whatever. I’m interested in that. And how do you then, but then how do you actually do that? How, how do [00:28:00] companies, um, make sure that they’re sending stuff that their customers want or that their audience wants to receive?

[00:28:09] Laura: So that’s getting a lot harder these days. Um, you know, and, and one of the things that I’ve always done, um, particularly when I’m dealing with SPL listings and, uh, for a long time, and I think Spamhaus still kind of. I’m not going to suggest this, but for a long time it was, you have to do everything CLI, you know, you have to confirm all your addresses.

[00:28:29] Laura: And what I would do was instead of, you know, like taking the entire database and confirming the entire database. Um, we’d split it up into addresses that we, we had some sort of activity, whether that be a purchase, whether that be a click, whether that be opens. Kind of, sort of. There were always caveats in my opens, but, um, but we know that this group of customers, that this group of recipients is somebody who is actually, you know.

[00:29:02] Laura: A user of your brand, right, you know, and so we could put those aside and we can say we’re not going to confirm those because we have enough activity. We have enough signals that they’re confirmed and then we could look at other things, you know, on the other end of things where, like, we’d never had an open.

[00:29:20] Laura: We’d never had a click. We’d never had a successful delivery, right? You know, this was free. All of that data hygiene stuff and like companies like never removed addresses and, you know, And all of that kind of thing. And we could say, you know what, this group is completely valueless. They’re like, they have never done anything.

[00:29:36] Laura: They’ve never logged into the account that they created. They never did any of this stuff. We’re just going to throw them away. We’re not even going to worry about them. And then that kind of left you a pile in the middle. And then we could kind of sort a little bit finer of, okay, you know, we have say 150, 000 or 150, 000.

[00:29:57] Laura: You know, 15 million or whatever addresses in this pile and we’ll confirm some of them and we’ll make some assumptions about others and we can do this, but that was kind of how I approached this fan house listings as a way again to keep those folks who we were pretty sure actually wanted the mail on the list without having to bother them because CLI is a bit of a pain for Thank you.

[00:30:19] Laura: Consumers, you know, they’ve got to remember that they got to click on the link and I’m like always getting mail and it’ll pop up and I’ll look at it and I go, Oh, I’ll do that later. And then like, it gets lost in my mailbox. And, you know, 2 months later, it’s like,
[00:30:31] Ken: Oh, wait, wait for the audience to see. That means confirmation of interest? Confirmed opt in. Confirmed opt in. Okay, thank you. See, I didn’t know that. I should know this. I should know this.

[00:30:44] Laura: So, in Spanhouse’s definition, because there are multiple definitions running around, but in Spanhouse’s definition, it’s you send them a message and you say, Hey, you know, do you want to stay on the list?

[00:30:55] Laura: Do you want to continue receiving mails? Right, right, right. And they need to click yes.

[00:30:59] Ken: Got it. Um, yeah, I’ve definitely seen a lot of that since the introduction of the Canadian anti spam legislation because it mandated, you know, all, anyone who sends mail to Canadians had to confirm, uh, that their recipients, uh, were opted in.

[00:31:15] Ken: And if you hadn’t done that in the past, you had to now do it. So there was this period of time in which, uh, I got tons of messages and I’m sure you did as well. You know, Hey, uh, do you want to keep? Being on this list because if we don’t hear from you, we’re going to drop you from our list. And I remember thinking this is great.

[00:31:33] Ken: Like I, I don’t actually want to be on many of these lists. This is fantastic. But in other cases, it caused a moment of anxiety. Um, yeah. So what you’re saying is, uh, for companies that want to have good deliverability these days, one of the first things that you do, um, is basically confirm. Or, you know, take a look at their recipient lists, segment out the ones that never do anything.

[00:31:59] Ken: I mean, a good fraction of those are probably spam traps anyways, and at best, they’re just uninterested entirely, so what’s the point? And then take the ones that are interested, that have some engagement with your list, and then do COI. On many of those, if not all of them, um, and that’s your, that’s your base.

[00:32:20] Ken: That’s what you’re going to keep going with, right? And that must be a tough conversation because I’m sure a lot of organizations, um, who might be naive to the modern world of deliverability. They don’t want to let go of those email addresses. They, they worked hard to get them. They paid money for them.

[00:32:36] Ken: They did, you know, million dollar ad campaigns to get those signups onto their lists. And now they’ve got to torch them. Like, how does that conversation go?

[00:32:45] Laura: Well, often that conversation is a lot easier than you might think because they’re coming to me with a spam house listing. Right, got it.

[00:32:53] Ken: The fire, the house is burning, got it.

[00:32:56] Ken: Right,

[00:32:56] Laura: you know, and so it’s kind of that piece of… Well, you can do this or you can stay listed, um, you know, and if you stay listed and you’re on many of the current ESPs, um, you’ll lose your ESP and you’ll have to go to a new one and then ESPs won’t onboard you if you have a current SPL listing. So there’s, there’s a lot of incentive when you can’t send an email, right?

[00:33:22] Laura: Um, when it’s when it’s somebody who’s coming to me, like, say, this big app or this big social media company, um, what we really look at instead is not necessarily doing the confirmation step because, you know, again, the confirmation step has problems. There are people who don’t care and who are happy to receive your mail, but actually won’t take the action to click.

[00:33:45] Laura: So, um, so it’s more about. What do we value? What are we looking for in terms of our optimum subscriber? And what actions do we want to take? And where are our values? You know, do we want to keep on as big a list as possible? And fundamentally, if you’re not having delivery problems, if your mail is going to the inbox, it doesn’t matter.

[00:34:15] Laura: But you can kind of end up being okay, and then you slowly, slowly, slowly, slowly tick down in reputation. And then you kind of hit the threshold and your mail starts going to spam. But, there are, unlike a decade or 15 years ago, there are, there are warning signs. You know, you’ll start to see more of your mail, particularly mail to…

[00:34:42] Laura: New recipients or to recipients that haven’t engaged very much with your mail that that mail will start going in the spam folder, but overall, most of your recipients are still getting it in the inbox. And this is kind of that, that individualized delivery that you were talking about. And so it’s [00:35:00] like, if we don’t know, if we as Google, or we as Yahoo, don’t know what an individual will think about this mail, then we’ll apply our judgment, which is based on these machine learning filters.

[00:35:13] Laura: And… Um, we’ll put it in the inbox or the spam folder. If we know about you as a recipient, and you’ve received mail from this sender before, you’ve received mail similar to this sender before, um, then we’ll apply that logic on top of ours and see if we sh if our decision for you is incorrect. Hmm. And so, there’s kind of these two levels of filters.

[00:35:36] Laura: There’s the filter that is the judgment of the ISP for everybody at the ISP. And those are the ones that we talk about when we talk about engagement and when we talk about complaints and when we talk about, you know, everything we do is that filter. Those individual filters, we all, we almost can’t affect.

[00:35:59] Laura: There’s not [00:36:00] anything we can do.

[00:36:02] Ken: Huh. Interesting. So yeah, I mean, like getting into a conversation about reputation. Uh, for a long time, there’s been this notion of IP reputation. Uh, you send, I mean, email is still broadly sent out of IPv4 addresses. You know, yes, there’s IPv6, but nobody really talks about it.

[00:36:20] Ken: It’s, it’s like, I think 10 years ago, it was an interesting thing to build into your MTA platform and there was a wave of implementations and then now it’s kind of there, but it’s IPv4, so there’s, there’s like 4 billion IP addresses. Each IP, you know, has a sort of rating of some sort. It’s either on a block list or it’s not.

[00:36:40] Ken: And, and, uh, it used to be that that was the main thing people worried about. Am I sending out a clean IPS? You know, people would pay a lot of money to be on IPS that are clean, are brand new, that had never sent anything before. I remember when we bought our, we bought a slash 20, uh, for mail [00:37:00] channels and.

[00:37:02] Ken: Uh, I sent the block to SpamHouse ahead of time and I asked them, has, have you ever seen this before? And they said, no, that one’s completely new. And, and so then I was like, okay, cool. I’ll buy this one because I had been warned by friends that when you buy IP addresses, you got to check their history because even if they were involved in a major spamming operation five years before, they’re going to have problems.

[00:37:25] Ken: Um, anyways, what’s your impression of. What reputation is now, like, is it still important? Uh, how has that all changed?

[00:37:36] Laura: Oh, the reputation discussion. Um, so yeah, and just for a little history for folks who may not know, the reason IP reputation was the top of the pops, and this is, you know, this is what we do, is because IP reputation was very difficult to forge.

[00:37:57] Laura: Because IP wa, the IP address that you were, that was connecting to you was the one piece of information that you could trust was true because it was actually from your own machines. Now there are ways to fake ips and they are still being done to this day. And there, you know, this was not a percent attacks and stuff.

[00:38:18] Laura: Yeah. Um, there was asymmetric routing where you sent your traffic out through like. Big pipes and then like had an AOL dial up to get the, um, to get the return packets. It was, it was crazy. It was absolutely crazy. Um, but so, and that’s why IP representation was a thing because that was the one piece of data that we could trust.

[00:38:43] Laura: Now, you know, again, in, in the 2000s, we started looking at things like. Okay. How can we make other bits of data more robust? And how can we make sure that we can trust other bits of data? So we came up with SPS and we came up with the Kim, um, and there were a bunch of other things that we tried and there were a bunch of other things that we talked about, um, that never quite made it into the wider deployment and we’re never quite codified or implemented.

[00:39:11] Laura: But what that gave us was that gave us a, more of a way to, to start trusting the domain name. So we could start trusting that a message that was signed with a DKIM, with DKIM, that whatever domain was in that D equals in the DKIM signature was the domain that was taking responsibility for that mail. And then we could hang reputation off of that.

[00:39:33] Laura: Right. Um, SPF is a little squishier, but again, you know, It definitely is. the idea was there.

[00:39:40] Ken: It definitely is. I could tell stories. Uh, for re recent history, I don’t know if you’ve heard, but yeah, uh, SPF is a challenge. SPF is a challenge, but DKIM is kind of like pretty strong cryptographically. The DNS itself is very.

[00:39:56] Ken: I’m not going to say perfect, but it’s pretty bulletproof now with DNSSEC. Um, the infrastructure is pretty good. So, so you’re saying, you know, domain reputation has become quite important. IP reputation, it’s still a thing, but domain reputation is very important.

[00:40:14] Laura: I treat IP reputation as you must be this tall to ride this ride, you know, and it, but it doesn’t matter how much taller than, you know, you’re just this tall.

[00:40:25] Laura: And so kind of, there’s this baseline of, you have to have. You know a decent enough reputation and these days that includes things like you have to have reverse DNS set up correctly and you have to You know not be in an a be in a block of bad IP So, you know, they’re they kind of There, there are other things to do with IP reputation, but in general, if, you know, you have a minimum IP reputation, then your mail is accepted.

[00:40:53] Laura: And once your mail is accepted, IP reputation is basically not an issue. Yeah. Um, and I’ve been saying this for a really long time, that once, once your message gets into the spam filter, rather than the, the MTA, which is the, uh, machine receiving the mail. Is it? That once the mail has been accepted by the receiving system.

[00:41:15] Laura: IP doesn’t matter, right? And that’s when the domain reputation comes in. And the domain reputation, any domain that’s in your message, anywhere in the message, has a reputation. So anything in your links, anything in your images, anything in your headers, you know, and there’s kind of like, we can really trust this or we’re not sure about this.

[00:41:39] Laura: So there’s strength to the reputation. So that domain that’s in your DKIM signature, That’s a really strong reputation. Right. Um, and if that domain in your, your signature, in your Deakin signature is in the same domain space as, you know, the links in your message, then that [00:42:00] reputation is really strong.

[00:42:01] Laura: Right. Um, but if it’s not, you know, the reputation of the links in the message is also important. You know, and the domains and the links, that’s really important too. And certainly we’ve seen over the years that there’s certain domains that end up in messages that are bad enough that they can just tank your delivery.

[00:42:22] Ken: Wow. So, for example, if you were using, uh, you know, if the links in your messages were pointing at… Some, uh, you know, some service that had, uh, that had lots of people using it. And they were using those links in their messages and the reputation of, of those messages was poor. That would kind of transfer over to the, those domains.

[00:42:42] Ken: So, I mean, to give a silly example, if your email, uh, contains URL shortener links, like that might not be the best idea because we know that spammers use URL shortening services all the time. And, and so it’s likely that those domains are bad. Uh, so if you’re, [00:43:00] you know, pro tip, if you’re sending out mail for your major corporation, you might want to have your links come from your domain, or at least a domain that you very carefully control.

[00:43:10] Ken: And you’re pretty sure is not used in spamming campaigns. Yep. Yeah. Huh. Pro tip. Pro tip. Yeah, yeah, exactly. Um, yeah, because one of the things that I think we all know now is that the ma the major email receivers. When they process a message, assuming it’s gotten through the IP reputation filters and they’ve actually got the message, uh, you know, one of the really important features is the domains.

[00:43:36] Ken: So I’m imagining that there’s all these things about domains that get thrown into the neural network, like the age, uh, you know, how often is this domain visited, you know, how many other messages is this domain used in, do people engage with those messages, all these things go into the bag and then a judgment.

[00:43:57] Ken: Is made and the domains are really a big [00:44:00] part of that, that judgment.

[00:44:02] Laura: Yeah. Domains are really are, are, I think the, the key to deliverability these days, you know, if you’re sending through a, a, a moderately reasonable service, um, your domain, your domain is going to trump

[00:44:20] Ken: everything. Interesting. Right. So if you’re, and if you’re sending through a poor service, then, then we can’t help you.

[00:44:28] Ken: Is that the idea?

[00:44:31] Laura: Well, I mean, you know, the first, the first kind of thing is, and in fact, I’ve got, I’ve got a client right now who I’m, I’m working through and, um, Their, their ESP, their compliance desk isn’t the greatest. And so this ESP, their shared IPs are going on and off of SpamCop fairly regularly, right?

[00:44:52] Laura: And it’s a B2B thing. And SpamCop is actually used by Mimecast, which handles a lot of B2B mail and hosts a lot of domains. So we’re seeing, you know, a lot of delivery problems based on these SpamCop listings, but their shared IPs, it’s not like, you know, I mean, it might be my client, but I don’t know that.

[00:45:12] Laura: Um So the idea that we have to kind of be aware of what we’re doing and who we’re doing business with. Now, there’s contracts involved and it’s not something that we can simply move them off. You know, there, there are, there are business considerations to making these kinds of big changes. Um, ESPs is like, I hate moving, you know, I hate moving a client off an ESP because we got to start.

[00:45:42] Laura: You know, from warm up and with it’s a lot of work. It’s a huge amount of work, you know, moving all of your campaigns and moving all of your logic and moving all of your reputation and all of this. It’s a huge undertaking. So I don’t like to recommend that. Um, but by the same token. There’s this piece of, well, this, this ESP is not really taking care of business and they’re just letting customers do stuff or it’s getting them listed on SpamCop and yeah, they get delisted after 24 hours, but then three days later they’re listed again, you know?

[00:46:18] Ken: Yeah, and that kind of leaves a stain. Yeah, that’s kind of, yeah. So, uh, you know. Talking about ESPs, I, I mean, I like, you would think I would know what ESPs do, right? Uh, I mean, we technically are an ESP, kind of. We relay mail for web hosting providers. Really, we’re like an anti spam service that also delivers mail.

[00:46:41] Ken: Um, we don’t have, you know, big enterprise clients sending to lists. Directly working with us, but you know, millions of hosting users, each sending to tiny lists of their own with absolutely no control over anything and no idea what they’re doing. Uh, because these are tiny SMBs, uh, they sent through our service, but what, you know, does the perfect ESP exist today?

[00:47:08] Ken: If you were starting from scratch, what would the ideal ESP look like? Like, could you…

[00:47:15] Laura: I’ve always said I don’t want to run an ESP because it’s just a pain.

[00:47:16] Ken: No, And, you know, but if you, if you did have, you know, uh, if you got 2 billion, uh, instead of that money going to generative AI, and you got to create Laura’s ESP, what would, what would be the characteristics of that ESP?

[00:47:35] Ken: What would be like, wow, everybody’s got to use this service because they’ve got it dialed from day one?
[00:47:42] Laura:See, I’m not sure something like that exists. You know, it’s, it’s not, there are different ESPs that focus on different things. So you get your Mailchimps and your Constant Contacts that focus on kind of the smaller folks who are sending to [00:48:00] 500, 600, 800, you know, addresses.

[00:48:04] Laura: And, and, and they don’t want to get bigger. You know, they’re a small company and they’re happy with their volume and all of that. And, and that kind of sender has a vastly different focus. need set than say, my big, you know, uh, social media, social networking company, who is sending 250 email, 250 million emails a day.

[00:48:27] Laura: And what they’re doing more is, you know, they need somebody who has hefty iron under the covers that can like. Send the mail out. They need really strong reporting. They need, you know, good bounce handling and good feedback and all of that kind of thing. Um, so that, you know, the experts inside the company can have the data that they need to drive the company.

[00:48:57] Laura: Um, you know, so, and there’s certainly, there, there’s certainly multiple ESPs in that space, you know, that have the big iron and that have the big pipes and that have the really, um, focused groups and who can, who are sending, you know, a billion emails a day, you know, across their entire customer base. Um, and I think that, you know, the MTA and the deliverability style features, you know, the ability to track opens the ability to, um, track clicks the ability to, you know, sign with independent d Kim keys and be able to, to multiple sign messages, um right to oversight, all of that kind of thing.

[00:49:44] Laura: Yeah.
[00:49:49] Ken: If you can’t do that, you shouldn’t be in ESP, but you know, this is like table stakes now.

[00:49:51] Laura: Yeah. Yeah. Um, but there’s a lot of other stuff that on top of that, that improves usability. So do you have an API? Do you let people bring their own IP space that you can route for them? Cause there’s a lot of groups out there that like.

[00:50:07] Laura: You know, you can buy a slash 20 or a slash 24 and you can take it to X and they’ll route it for you. And so all those, I, you’re, you’re using their infrastructure, but it’s your IP, right? Right. And you can take those IPs with you when you move.

[00:50:19] Ken: Right.

[00:50:21] Laura: Um, you know, but, but so I don’t think there’s one. You know, and I, I actually think that the diversity in the marketplace is really good.

[00:50:31] Laura: I think it is better that we have folks who are focusing on specific needs and focusing both on the, you know, for the sender side, but also, you know, how do we, how do we interact with the recipients? What kind of experience do they have when they want to come unsubscribe from something? And we manage the unsubscribe page.

[00:50:50] Laura: So I don’t think there, I don’t think there’s one answer to that.

[00:50:52] Ken: Right, got it. So yeah, so there’s, there’s a healthy mix of ESPs in the market already, is what you’re saying, basically, and, uh, and they each serve different segments. Uh, and if you want, you know, tons of data so that you can have your own internal machine learning people crunch through it and figure out what lists should be doing what, you can have that.

[00:51:15] Ken: And if you just want to send a mailing list to your knitting group, you can do that too. Uh. Yeah. Yeah. Yeah. And I guess the problems faced by a company like MailChimp are going to be way different than the problems faced by an ESP that’s servicing Fortune 500 companies, uh, you know, uh, they’re not necessarily, you know, the big, the ESP servicing the Fortune 500 isn’t going to be so worried about abuse because while they’re definitely trying to make sure that the Fortune 500 doesn’t do really stupid stuff, they’re probably less worried about, you know, a fake login, uh, and someone doing a drive by campaign.

[00:51:53] Ken: Is that true? Or, or, you know, do those big iron ESPs also worry about abuse? Oh, but they should. They should? Okay, sure.

[00:52:03] Laura: You know, um, there certainly have been a lot of, there was, there was an ongoing set of attacks. Um, it’s been in the last five years, because the last time I heard about it, it was actually here in Dublin. Um, but that there was, there was a significant attempt to, um, compromise employee credentials at a couple IS, ESPs. Wow. Um, uh, some of them made the news and I’m not going to name them because fundamentally, you know, you can look it up.

[00:52:41] Laura: But, um, and what it ended up being was, it was, And, and they’re not, they’re names you would recognize, but what I heard through the grapevine was that, you know, there was a lot of, there was a lot of work being done on the social engineering side. It was, you know, people were looking up Facebook friends and family and sending texts to the employees pretending to be family members.

[00:53:10] Laura: Wow. And friends. Um, to compromise their credentials and it worked, you know, it did work in a couple cases. Um, and I think it’s kind of interesting because the, the ones I remember more recently, it was always, they got in and they stole people’s crypto lists, right? So it was, it was all crypto coins and yeah, cryptocurrency.

[00:53:35] Laura: Um, but it’s not the first time. Right. That these companies have been, um, compromised and in fact, um, I still, there was, there was a big compromise, I don’t know, 12, 13 years ago, maybe. And I ended up having to call return path on the Wednesday before Thanksgiving and go, um, guys, I hate to tell you this, but your database has been compromised.

[00:54:06] Laura: Oh. You know, and I still get, every once in a while, I will see, um, I will see Tom and he’ll just look at me and he goes, You really ruined my Thanksgiving.

[00:54:14] Ken: Oh, that sucks. And I know how important Thanksgiving is to people in the States. Um, I mean Canada. We have Thanksgiving earlier. And it’s just, it’s not that big of a deal.

[00:54:27] Ken: You know, you have, you have a turkey dinner with your family, that’s it. But in the States, it’s like, everybody flies home. Everybody takes a couple, two, three days off. And it’s hardcore. It’s almost bigger than Christmas, to be honest. Um, so to have your Thanksgiving disrupted because of a database breach is, that’s problematic.

[00:54:49] Ken: I can see, yeah. Yeah, I, I mean, we, we got a taste of how, uh, not a breach, but in our case, you know, someone discovered what they perceived to be a, an exploit, uh, where they could. Um, they could get, they could send mail, uh, through somebody who used us and there, and because of SPF, that would allow them to impersonate domains that were sending mail in other parts of our infrastructure and being a unique kind of ESP that sends mail for the hosting industry.

[00:55:30] Ken: We can’t, uh, force verification of every single domain. Because there’s too many of them that we send mail for more domains than any other transactional provider. There’s literally millions of them. Um, and they’re popping up and down all the time. So they thought that this was a really cool exploit and they talked about it at Def Con.

[00:55:49] Ken: Um, and it was, it was hard work dealing with that. I learned a lot about PR. I made a lot of mistakes. I’ll be the first to admit I made myself look really bad. Um, and I’ll never do that again, but, uh, it just goes to show when you. are working in an adversarial world, uh, one slip up can be really painful.

[00:56:11] Ken: Uh, and ESPs are like, they, they’ve got a treasure chest of IP reputation, uh, that they are guarding for their customers. And if that gets disrupted, it is very painful. It can ruin your Thanksgiving, you know?

[00:56:28] Laura: Yeah. And, and I think, I think that’s something that a lot of marketers don’t quite understand. Um, is, is just how adversarial and how scary the security is around email.

[00:56:47] Laura: I mean, you know. Email is the root of all, I won’t say all, but, you know, the vast majority of the major compromises that have happened over the last 20 years have [00:57:00] all started with email. Yeah. And that is, and that is why the filters do what the filters do. You know, and that, and if, and I, I get this question occasionally, it’s, what will happen if spam goes away tomorrow?

[00:57:13] Laura: Well, if spam goes away tomorrow, it doesn’t matter. The filters are still going to be there, you know, because. A, we’ve built them up and they’re useful to people because a lot of people get a lot of mail and find things like tabs or sorting or any of this stuff. The filters will still be there. But there’s still this idea of email is dangerous.

[00:57:35] Laura: And there are, there are, there are bad guys out there that, I mean, you talk to some of the folks from Google and they will tell you absolutely that, you know, it was a security compromise that changed the 2016 election in the U. S. Right. Right. You know, and, and there’s a whole lot of companies who are going to want those appliances and you’re going to want those filters in front of their mail, even if there’s no spam.

[00:58:02] Ken: Right. Yeah. Uh, this brings me to a topic I wanted to talk to you about, which is, which is AI generative AI specifically. I must say, uh, when, uh, when open AI released GPT four last, uh, in March, uh, my immediate thought was. Oh, God, this is going to be used by the bad guys to create very human, like, lifelike conversations with, with the victims over email and SMS.

[00:58:32] Ken: Um, and we’re not going to have to wait very long until we start seeing this in, in the real world, in our email boxes. And I have to say, I didn’t see it for a while. I didn’t see anything that was like, obviously GPT generated, but now I do every day. And, uh, a very recent thing was. Okay, you know, it’s GPT when the email starts with I hope this email finds you.

[00:58:56] Ken: Well, that seems to be its play It opens every email with I hope this email finds you. Well and What it was trying to get me to do was to you know, really pay attention because they had found some Security breach that they wanted to talk to me about and it was you know from some brand Purportedly in reality.

[00:59:19] Ken: It was from a gmail account and I, I knew that it was, I knew that it was a phishing attack, but, um, by using generative AI, they were able to make that communication highly personalized. So they had some information about me that they must have drawn from LinkedIn or another public source. They knew what my role was, and the email was perfectly crafted around that.

[00:59:42] Ken: Um, I, I’m pretty sure a human didn’t write it. AI wrote it. And you got to think if this is happening millions of times a day, uh, and these models are only getting more powerful. That’s gonna just… make the anti abuse job so much more [01:00:00] difficult because the spam filters, it’s not a template. It’s an approach.

[01:00:04] Ken: How do you filter that stuff? Right. And they’re coming from Gmail.

[01:00:09] Laura: Well, I’m going to, I’m going to argue that that’s not a, that’s not generative AI writing those emails because I’ve been getting them for much longer than before March. I’ve been getting them for years and years and they always do. I hope this email starts.

[01:00:22] Laura: Oh, funny. Well, yeah. Um, and yeah, and it’s, it’s always very clear to me, you know, that they’ve, they’ve done things like, you know, scraped my website. So they’ve pulled an old blog post off or, you know, they found something on my LinkedIn or whatever. But this is all, this is all just, um, it’s basically like WordPerfect macros rather than I think GPT.

[01:00:47] Laura: And this is something that we’ve been dealing with for a long time. Now, what, what, what is actually happening and where the threat model is, if, you know, we [01:01:00] go down the threat model end of things, where the threat model is these days, what happens is that you’ll get somebody approaching you on Facebook or on, um, Facebook messenger or email, and eventually they’ll try and move you over to WhatsApp or Signal.

[01:01:14] Ken: Yes, absolutely.

[01:01:15] Laura: Because that is where everything’s encrypted, and it cannot be, it cannot be gotten into by either Facebook or the Signal folks. Ha. Um, and that’s why they’re doing it, is so that they can, so that they, it’s easier for them.

[01:01:30] Ken: They can get you into a dark alley, basically, and then they can mug you.

[01:01:34] Laura: Basically, yeah, you know. Where the, where the cameras can’t see you, where the, you know, the intercepts can’t happen. Um, so I think, but I also think that some of that, that some of that generate, you know, it could be being generated because somebody has, has found their stash of 50, 000 templates that they use and fed it into the generated bot AI.

[01:01:56] Laura: Right. And so, you know, but there are tools that will personalize messages based on LinkedIn profiles. Yeah. Um, and they’ve been around for a while. I think that that’s always been a challenge because spear phishing has always been an issue. Um, you know, and you get the, you get the messages impersonating the CEO telling the, um, accountant to make a wire transfer.

[01:02:23] Laura: Sure. Or you get these other kinds of… Things that’s always been an issue, but I think that’s that’s one of the reasons that um mail into Businesses is a little more complicated than mail into like gmail or yahoo because They have many more layers, and it’s not just this single, is it spam, is it not spam?

[01:02:44] Laura: It’s, it’s, is this dangerous? Is this impersonating somebody? What are they intending?

[01:02:48] Ken: There’s all of these kinds of things. Yeah. Uh, fascinating. I mean, I, like, talk, not, not talking about phishing or security yet, you know, going over to the marketing side, um, We’re certainly seeing a lot more cold prospecting type activities where either people are coming to us and Asking if they could use our service to send out cold prospecting Oh to verified lists only but it’s cold meaning like it’s spam, right and we turn them all down obviously but many of them are talking about how they use AI to generate the cold prospecting messages and They can even automate the handling of replies from the prospect.

[01:03:38] Ken: So they, they’ll, you know, take a list. Uh, they have a sort of overall intention and they’ll go back and forth with the target, uh, until a human, until it sort of makes sense to forward it to a real human. And I, I just wonder if that’s, what’s going on kind of in the dark corners of the internet already, like our big companies starting to engage with generative AI to run email campaigns.

[01:04:02] Ken: Are you seeing that in your practice where they’re. They’re saying, Hey, you know, we want to use AI to automate the generation of content or even the handling of whole conversations. How do we do that in the email world and not get blocked? Right? Like, how would you, is that happening? Yet Have you seen that?

[01:04:20] Laura: So, I mean, it might be happening. Um, I, I generally, I’m not a fan of, of, of any of these. A you know, I am not on the AI bandwagon, let’s just put it that way, . Um, and so it’s, then you probably need to find a different consultant because Idon’t have any answer if that’s what you want to do.

[01:04:35] Ken: Yeah. Yeah. And why would you, why would you say that?

[01:04:41] Ken: Like, why are you not on the bandwagon?

[01:04:47] Laura: Not that it’s the right bandwagon, but, well, because it’s inherently. I mean, how the large language models have been generated has been by stealing content from people. You know, my website is in multiple large language models because they scraped all of my content.

[01:05:09] Ken: Totally. It knew, I, I asked GPT, do you know Laura Atkins of Word to the Wise? And it was like, yes, I absolutely. What do you want to know about her? And I said, well, generate some questions for, uh, an interview and it happily did. So, and I’m not going to lie. That’s how I generated the list of initial questions.

[01:05:25] Ken: And then I, you know, iterated on it, but. Yeah, they know. They know about you.

[01:05:32] Laura: Um, and I have, you know, and I have friends that are, you know, authors, and like, authors of movies one would hear, one would know of, and books that have been on the New York Times best selling list. And their content has been pirated and put in these large language models.

[01:05:48] Laura: And so, the models themselves are built on theft. Um, and they’re built on stealing intellectual property. And… From my perspective, you know, what a large language model is, and you know, okay, so I came from the, I came from the squishy biology end of science, but I did have some statistics, and I do have some, you know, familiarity with the space.

[01:06:12] Laura: It’s basically autocorrect gone global. Right. Right. You know what a large language model and these generative a, um, a AI engines tell you is if you have this word, or if you have these, you know, seven words, then they tend to be in this order. Or if you ask about this topic, these are words I have seen about this topic.

[01:06:36] Laura: Yeah. Um, and I think part of, you know, and, and it makes shit up absolutely. There’s a couple people who, I mean, there was, there was, there have been a few that have made the press, but, you know, somebody tried to write a legal brief and it made up fake cases for this guy’s legal brief. Right, right, right. Um, and it’s done the same with science.

[01:07:00] Laura: People have asked it science questions and it has made up, um, peer reviewed published articles. With real people’s names in there. And so now these scientists are getting requests for this article that they wrote that they never wrote because… Oh no. GPT. Um, you know, and so there’s, there’s, there’s a lot of…
There’s a lot of reasons I don’t like it. Um, and I don’t think it is as transform I think people want it to be transformative because they want to be on the beginning of the next great bandwagon. Um, but it’s not AI. It is statistical predictions. And so, I don’t, you know, and I’m just not, I don’t think it’s gonna make that big a difference.

[01:07:49] Laura: You know, there is, there, all it’s going to do is increase the amount of garbage in the world. Yeah. And garbage writing in the world. And one of the problems, one of the problems that we’re starting to see is actually that, um, the AI, the AI makes stuff up and it puts together some of these nonsensical things.

[01:08:11] Laura: And then it publishes it, somebody publishes it somewhere, and then the AI takes in that content that it has made that is incorrect. And so now we’re getting that it’s iterating on its own garbage.

[01:08:24] Ken: Yeah, right, right. Of course, that’s a problem. I actually think that’s, well, there’s a few perspectives on that.

[01:08:33] Ken: One, one perspective that I’ve heard, uh, and I think from people at OpenAI, okay, maybe not surprisingly, Is that they actually don’t see that as a problem because the AI generated content that is published to the web is being curated by humans who are saying, Oh, this stuff’s good enough to be published.

[01:08:53] Ken: And so there’s sort of like, uh, it, it, yes, it’s getting more training data that’s regurgitated from its own language model, but the, the, the humans are selecting the stuff that gets put up largely. And so they feel like it’s going to improve the quality of the model instead of making it worse. I guess the other thing is.

[01:09:13] Ken: They, they’ve frozen a snapshot of what the web looked like before it was all polluted. And so they can compare the before and after, and even though they might not know what new stuff is AI generated, at least they know that up until March of 2023, the web was largely human authored. And so we can rely on that for the basis of our model.

[01:09:35] Ken: Um, anyway, I That’s so adorably naïve! Isn’t it? Yeah. This adorably naive stuff is, is backed with, you know, tens of billions of capital and another interesting side of generative AI, um, is the generation of images, uh, and. You know, that’s hugely controversial with artists, um, who are now finding ways that they can either block their art from the AI models, or they’re even creating adversarial art to, to ruin these, uh, models.

[01:10:07] Ken: So they’re, they’re like generating work that they say is theirs, um, but it’s garbage in order to taint the model. And so, you know, people will be less incentivized to generate images in the style of whatever artist. I mean, if you go into, um, many of these tools, these art generation tools, uh, You can kind of have it generate just some sort of random stuff.

[01:10:31] Ken: Uh, you can give it a picture and you can say like, Well, just make this into art, right? And it’ll generate a few different ideas for you. And each one will have an in the style of artist name. And so they’re literally not even trying to hide the fact that they are copying the style of artists. And those people aren’t getting paid.

[01:10:48] Ken: That’s always really, you know, bothered me. Um, uh, so yeah, I kind of get what you’re, where you’re coming from, but, but it’s with us, right? It’s this stuff is now with us and, [01:11:00] uh, I had just wondered is, is generative AI something that people in the email sending side are, are starting to use in a serious way to, to send out, you know, to generate campaign content, for example.

[01:11:15] Ken: Personalized campaign content. I, I, I hypothesize that if you could get a language model to write very personalized content to each user, then you might have better engagement. If it’s not just a template, that’s the same for everyone. From what you’re saying so far, it sounds like that kind of thing has been going on already for a long time,

[01:11:40] Laura: right?

[01:11:41] Laura: Yeah, I mean, what, what the large language models do is it kind of, it’s nothing more than, you know, hiring a group of English speakers in the Philippines or in the Far East who, to generate individual emails based on, you know, and that’s been the, the boiler rooms, you know, and they’re paying pennies to have, to have people write this. And, you know, so I just, you know, and I know, I know there’s a lot of marketers that are thrilled and think it’s great and think that this is, you know, candy and it’s going to change everything. I think there are societal and ethical questions that, you know, are important to me, um, and are things that I think. Should be dealt with and, you know, a lot of these big companies that have these large language models have fired their AI ethics teams.

[01:12:41] Ken: Yeah, right.

[01:12:42] Laura: I forgot about that. And that is, you know, and it’s like, this does not strike me as a good pathway to be going

[01:12:50] Ken: down.

[01:12:52] Ken: I wonder if it’s going to end up being a little bit like Airbnb, where, wow, great startup idea, made tons of money. And initially, it’s so cool. We’re all traveling around, you know, I remember. Circa 2011, 2012, going to Berlin and staying in someone’s apartment and thinking, this is so great. Wow. It’s like, I’m, I’m a real Berliner, you know, and now cities are chucking out Airbnb entirely.

[01:13:17] Ken: They’re banning it like New York, right? Uh, where I live, the province has essentially banned it in the entire province, um, because we found this tiny problem with Airbnb. That it completely eliminates rental housing for actual people who need to live there. And so communities are getting cleared out and people are homeless.

[01:13:39] Ken: With AI, I think it’s kind of similar. It’s very glitzy. It’s very exciting in, in this early phase, but the real impact of it, uh, we’re not going to feel for a while. And when we do feel it, people are going to get really mad. And they’re going to create policies that, that make it a little bit harder to, to do this stuff.

[01:14:00] Laura: I mean, I…

[01:14:04] Laura: This was a clear endgame for me. I mean, I, you know, the Airbnb thing. How did you not think that was gonna happen, you know? Yeah. I mean, And I, I think that’s where, you know, the ethics teams and kind of the broader societal questions and the things that we need to address and wrestle with are actually important.

[01:14:26] Laura: And we ignore a lot of them all the time. Um, and, you know, I And I say this again as somebody who, you know, was trained as a scientist and avoided every bit of humanities that they didn’t force me to take when I was getting my degree.

[01:14:45] Ken: Me too. But,

[01:14:46] Laura: um, you know, it’s, it’s those humanities questions where you kind of have to wrestle with stuff that’s not black and white and you kind of have to weigh the differences and weigh the, the pros and the cons and who benefits and who is harmed and so where does this move and where does this go?

[01:15:05] Laura: How do we make that better? And how do we fix that? And, and how do we, how do we even ask and talk about these hard questions? And I think, I think, you know, the AI discussion is just one of the, one of the places where, you know, we’ve asked it. I also think it’s suspicious that, well, it’s not suspicious, it’s very clear what’s going on, uh, that as, as crypto started to, as, as cryptocurrency started to die.

[01:15:35] Laura: came in. And so, you know, it’s, it’s all the same people trying to use all of the hardware that they were using to mine Bitcoin and all of that are now in the, oh, and let’s do AI and let’s, you know, so there’s a huge profit motive there. And I am not, I am certainly not a person that is, that is in the, if it makes money, it’s good.
A lot of people make money doing very bad things. A lot of people make a lot of money doing very bad things. And so I don’t think money is, is, is a good answer to, but this harms people. You know, this, this is, this is harming people. This is, you know, causing problems. These are the problems, it’s gonna cause problems down the line.
And, you know, the answer I get back is, well, but I can make billions of dollars.

[01:16:28] Ken: Yeah, or, or like, it’s inevitable. Yeah, but, you know, come on, stop being such a Luddite. This is inevitable. It’s going to happen. You know, get with the program kind of thing.

[01:16:39] Laura: Yeah, but they told me that, you know, fiat money was going away, and everybody would be using Bitcoin, you know, a decade ago, so it’s like, yeah, well, no, actually,

[01:16:48] Ken: that’s not how it works.

[01:16:49] Ken: I feel like I’m, I’m one of those people who’s inexorably attracted to the latest and greatest thing. Um, and back when in early 2011, uh, a friend of mine who is really well connected in the security space, he said over coffee one day, Hey, have you heard about Bitcoin? You got to look into Bitcoin. It’s fascinating.

[01:17:11] Ken: And I went to Hong Kong for two weeks on a business trip and I sat in a cafe and I, and I just did nothing but read about Bitcoin, you know, and it was, it was so engaging, so fascinating what was going on. Um, And, but over the, over the ensuing sort of 12 to 18 months, as I got more engaged with that world, what I found was that there were definitely some cool people working in that world.

[01:17:37] Ken: There were a few people who were like, these are my people there. They seem to be ethical. They’re doing it for the engineering, you know? Um, but the vast majority of people were like weird hucksters and, uh, and it just had this scammy. Feel to it, um, this get rich quick scammy feel to it. And, and it, that just got worse and worse over time.

[01:18:00] Ken: Um, and I’m really glad that that phase has kind of broken and now, uh, you know, it’s still there bubbling away, but governments have finally caught up and they’re saying, you know what? That behavior was outright illegal and we’re going to now prosecute these people. So I feel like it’s sort of like now it’s, uh, it’s a contained issue.

[01:18:23] Ken: Um, but AI, AI is not, is not yet there. It’s still in that it’s still in that early phase where everybody’s looking at it. Everybody’s fascinated tinkering with it. Um, and we, we ain’t seen nothing yet. Like it’s going to get seriously scary before, uh, before. A serious amount of investment goes into the other side of the adversarial game with AI.

[01:18:48] Ken: Um, just like we, you know, just like we saw in the anti abuse space. I think people who work in our industry really understand this, you know, adversarial technology. You see it coming and you’re [01:19:00] like, that’s going to be a problem. Everybody else is like, no, it’s not. This is great. This is fantastic. You know, you won’t, you won’t need a bank account.

[01:19:06] Ken: Everybody can be like self sufficient at, you know, uh, And you’re like, no, this is really going to be a problem. Like they’re going to be funding terrorism soon. Just watch. And sure, it happened, you know, same with AI.

[01:19:19] Laura: Well, and you know, and in some ways it’s kind of. It’s kind of like the whole spam thing. I mean, you know, we were telling people back in 96 and 97 that spam was going to be a problem.

[01:19:29] Laura: Yeah, and this was going to cause real issues And they’re like it’s four messages just hit delete and yeah, it was You know in a big spam run those days. It was a couple hundred thousand messages Well, of course, you know, this was like pre 2000. So the internet was a whole lot smaller But by the same token, you know And these are questions that we kind of don’t have answers to.

[01:19:48] Laura: Um, and these are issues that I think people should think more about. Um, but AI is going to do what I, you know, marketers are going to do what [01:20:00] marketers are going to do. And I’ve never found a place with. Where marketers think they’ve gone too far, they always think they haven’t gone far enough, right?

[01:20:11] Ken: Right. Yeah. Well, yeah, I can see that, but if you’re a marketer, your job, your job at a company is to make certain numbers happen. You got to have leads coming in. The salespeople have to be busy. That’s, that’s your whole metric that you’re measured by. Right. And so you haven’t gone far enough because you’re always going to have higher expectations.

[01:20:34] Ken: From one quarter to the next that doesn’t surprise me and and to the extent that people build AI tools Whether they are whether they work really well or not. People are gonna try them.

[01:20:44] Laura: Yeah, I I will say that Here I don’t and every time I’ve like looked into creating account and playing with them They’ve started asking me for personal information that I’m not giving to an AI generator So I’ve actually not played with them because no you don’t get my phone number.

[01:21:00] Laura: No, you don’t get my address No, you don’t get all this data about Um And I don’t know if, if everybody else is just like, Oh, I don’t care, they can have that data or, or what. But, you know, or if it’s just something about the legislation on the side of the Atlantic.

[01:21:15] Ken: Right. Yeah, it kind of, it kind of makes you think differently, doesn’t it?

[01:21:21] Ken: Um, well, listen, Laura, we’ve, we’ve spent a long time talking together. I, I could talk to you for another two hours and maybe we should do another session at some point, but I think we, we should wrap it up. Uh, okay. I thank you so much for coming on the podcast, uh, and, uh, and I look forward to speaking with you again.

[01:21:42] Laura: All right, it’s been wonderful. It’s been great. And I’m happy to come back.

[01:21:46] Ken: Awesome. Awesome. We’ll do that.