[00:00:04] Welcome to another episode of Tech
[00:00:05] Unhinch where tech gets human. I’m your
[00:00:07] host Rabia Javeed and joining me today
[00:00:09] is Christopher Pen who is a co-founder
[00:00:11] and chief data scientist at Trust
[00:00:13] Insights and eighttime IBN champion in
[00:00:16] data and AI. He has been working in AI
[00:00:19] since 2013, a decade before Chad GPD
[00:00:21] made it mainstream and has built AI
[00:00:23] powered marketing systems for brands
[00:00:25] like McDonald’s, T-Mobile, Twitter,
[00:00:27] Cisco. He is also the best author of
[00:00:29] several bestselling marketing books
[00:00:31] including almost timeless 48 foundation
[00:00:33] principles of generative AI. Welcome to
[00:00:35] the show Chris.
[00:00:36] >> Thank you for having me.
[00:00:36] >> Chris, we we’ll start off with some ice
[00:00:38] breakers. Um you are a martial artist
[00:00:41] who throws knives, you can read turrets,
[00:00:43] and you’ve built AI attribution models.
[00:00:46] How do these two worlds connect? Because
[00:00:48] that’s not a normal resume that one
[00:00:50] would just see. The thing about the
[00:00:52] martial arts is that it teaches you,
[00:00:54] particularly the what the art I
[00:00:55] practice, which is an old Japanese
[00:00:56] battlefield martial art. It teaches you
[00:00:58] that you have to win even under
[00:01:00] circumstances where you’re being
[00:01:02] strongly encouraged not to win. You can
[00:01:03] do so by any number of ways, but the the
[00:01:06] art I practice is called ninjutsu. And
[00:01:08] contrary to what you see in a lot of
[00:01:09] movies, the the whole Japanese ninja
[00:01:12] thing was about information gathering,
[00:01:13] intelligence gathering, so that you
[00:01:15] could avert crisis before they happen.
[00:01:17] This translates really well into things
[00:01:19] like a marketing analytics, data science
[00:01:21] and AI because if you have the data and
[00:01:25] you know what to do with the data, you
[00:01:26] know how to validate data and you know
[00:01:28] how to to put data into action, you have
[00:01:31] a decided advantage over other people
[00:01:34] who do not because in the age of AI, all
[00:01:37] of these, you know, AI models and these
[00:01:38] technologies are data hungry. We’ve all
[00:01:41] seen articles about how they’re
[00:01:42] consuming, you know, every piece of data
[00:01:44] on the internet. Well, if you know how
[00:01:45] data works, if you know what to do with
[00:01:47] it, you can build a durable competitive
[00:01:51] advantage over other people because you
[00:01:54] know what to do with data. And even if
[00:01:56] two people have the same access to the
[00:01:58] same public data, if you’re better at
[00:02:00] working with it, it’s like, you know,
[00:02:01] two people who are given the same
[00:02:03] ingredients, one’s a professional chef,
[00:02:04] one’s not, the professional chef is
[00:02:06] going to come up with a better dish.
[00:02:08] >> You started working in AI in 2013, a
[00:02:11] full decade before Chad GPD exploded.
[00:02:13] What’s the biggest misconception people
[00:02:15] have now about AI that you even knew
[00:02:17] back then?
[00:02:18] >> All these technologies are predictive
[00:02:20] technologies. They predict based on the
[00:02:22] data that they have, right? You look at
[00:02:24] classical machine learning, regression
[00:02:25] analysis, classification, look at
[00:02:27] generative AI. When you look at what’s
[00:02:30] happening behind the scenes when an AI
[00:02:32] model is running, if you install for
[00:02:34] example LM Studio or Lama.cpp CPP and
[00:02:37] you install a local model like you know
[00:02:38] Quinn or uh Gemma 4, you can watch the
[00:02:42] predictions happen as it’s assembling
[00:02:44] words and that should tell you a lot
[00:02:46] about what these tools are and are not
[00:02:48] capable of. If they have never seen
[00:02:50] something, if they have never ingested
[00:02:52] data before about that specific thing or
[00:02:55] topic, they can’t predict it. You cannot
[00:02:57] predict what you don’t have data for.
[00:02:59] People expect these tools to be kind of
[00:03:01] like these magical oracles, allseeing,
[00:03:03] all knowing. No, they are prediction
[00:03:05] engines. And so if you don’t give it
[00:03:06] data to predict from, you’re not going
[00:03:08] to get as good a prediction. So a
[00:03:09] classic example of this is in time
[00:03:11] series forecasting, right? If you are
[00:03:13] predicting from bad data, let’s say
[00:03:15] you’re doing a a time series forecast to
[00:03:17] predict revenue or predict marketing
[00:03:19] qualified leads and you use a 5-year
[00:03:21] time horizon, you’re screwed. Why?
[00:03:23] Because 2021 was the middle of a
[00:03:26] pandemic. Customer behavior was wildly
[00:03:28] different then. By 2023, it changed. So,
[00:03:31] if you don’t know anything about stats
[00:03:34] and prediction, you’re going to bring in
[00:03:35] 5 years worth of back data and your your
[00:03:37] forecast will be horrible because
[00:03:39] they’re predicting a world that no
[00:03:40] longer exists. Just like if you’re using
[00:03:42] data from before 2020, you are trying to
[00:03:44] predict a world that no longer exists.
[00:03:46] If you’re doing fuel prices and oil
[00:03:48] prices, right, it is May of 2026 as we
[00:03:51] record this. The last 78 days have been
[00:03:55] wildly different in terms of what’s
[00:03:57] happening with oil and petroleum in the
[00:03:59] world because of geopolitical events. If
[00:04:01] you are trying to predict these things
[00:04:04] using data from before that time, guess
[00:04:06] what? Your prediction is going to be
[00:04:07] wrong. As it translates to generative
[00:04:08] AI, a lot of people will say, I want,
[00:04:11] you know, I want AI to write me this
[00:04:12] blog post about this topic. And then
[00:04:14] they’re like, well, it didn’t spit out
[00:04:16] anything that was innovative. Well, no
[00:04:18] it didn’t because by definition, if you
[00:04:20] want something that’s innovative, it has
[00:04:21] to be something that’s never been seen
[00:04:22] before, which means it can’t predict it.
[00:04:24] So, you have to provide it. And one of
[00:04:26] the biggest things that I would
[00:04:27] encourage people to do is use this thing
[00:04:29] that we all have, right? This mobile
[00:04:31] device, turn on the voice memo recorder,
[00:04:33] spit out what’s in your head that is
[00:04:35] hopefully unique, and have the machines
[00:04:38] generate from your ideas because your
[00:04:40] ideas are probably more unique and less
[00:04:43] seen than what the models have been
[00:04:45] trained on. So you know Christopher back
[00:04:47] in 2019 you gave a keynote called five
[00:04:50] ways AI is changing marketing forever.
[00:04:52] You know that was very impressive that
[00:04:54] you were probably doing it in back in
[00:04:56] 2019 but we are in 206 now. If you were
[00:04:59] to give that talk today what are the
[00:05:01] five ways now that are relevant and you
[00:05:05] know what has still stayed the same from
[00:05:07] those five ways?
[00:05:08] >> Well I mean the things like having good
[00:05:10] data and putting it to use that never
[00:05:11] changes. What people need to understand
[00:05:14] now is that we we have what we call the
[00:05:15] five levels of AI enablement. This is a
[00:05:18] trust insights thing and maps to product
[00:05:19] market fit. Done by you, done with you,
[00:05:21] done for you, done without you, done
[00:05:23] ahead of you. And briefly, done by you
[00:05:25] is what everybody and their cousin knows
[00:05:26] as chat GPT. You sit there, you’re the
[00:05:28] typing person. You’re copy pasting,
[00:05:30] right? You’re doing all the work. Done
[00:05:31] with you are things like GPTs or gems or
[00:05:34] clawed projects where you have a
[00:05:36] standard operating procedure in place,
[00:05:37] but you’re still doing like half the
[00:05:38] typing and half the copy pasting. done
[00:05:40] for you is agentic AI and this is a big
[00:05:43] leap. This is where we are today. Aentic
[00:05:45] AI tools like claude code, claude
[00:05:47] co-work, open work, open AI codeex,
[00:05:49] Google anti-gravity. These are tools
[00:05:51] that you are now learning how to
[00:05:53] delegate. You’re learning how to be a
[00:05:55] manager of AI as though you were
[00:05:57] managing a junior employee. And this is
[00:05:59] a big big head change because you now
[00:06:02] have to get good at delegation. You have
[00:06:04] to get good at planning. You have to do
[00:06:06] uh my my CEO and co-founder Katie Rober
[00:06:08] and I were talking on the Trust Insights
[00:06:10] podcast this week about how we now have
[00:06:12] this ridiculous term called requirements
[00:06:15] engineering which is sort of an
[00:06:16] evolution of context engineering which
[00:06:18] was an evolution of prompt engineering
[00:06:20] because tech bros just love unnecessary
[00:06:22] jargon. But at level three which is
[00:06:24] where companies should be today you
[00:06:26] should be delegating work to AI. That’s
[00:06:28] what’s changing marketing right now. You
[00:06:30] can hand claude co-work a lot of data
[00:06:33] and say build me a marketing strategy
[00:06:35] for the next quarter and as long as the
[00:06:37] data is good and the instructions are
[00:06:38] good it will generate something. It will
[00:06:40] generate the assets. It may even connect
[00:06:42] to your CRM and execute things. Where
[00:06:43] this is going is in tools like openclaw
[00:06:46] Hermes agent etc. These are fully
[00:06:49] autonomous systems where you are now
[00:06:51] instead of delegating a project plan you
[00:06:53] almost delegating a job description. So,
[00:06:55] I have one, for example, running in
[00:06:57] another window right now that is finding
[00:07:00] a thousand qualified prospects for my
[00:07:03] company based on specific criteria. It’s
[00:07:05] using a bunch of tools, including some
[00:07:06] that that it built based on my
[00:07:08] directions. And it’s going to produce
[00:07:10] not only the list, but also the assets
[00:07:12] and things. And if I wanted to, I’m not
[00:07:13] going to because it’s that would be
[00:07:15] absurdly dangerous, but if I wanted to,
[00:07:16] I could have it execute and actually run
[00:07:18] that. And as the tools get better, that
[00:07:20] will be a reality in the in the very
[00:07:22] near future of these tools being able to
[00:07:24] execute the entire plan. You can have it
[00:07:26] do it now, but it’s dangerous.
[00:07:27] >> Also, Chris, your book is called Almost
[00:07:29] Timeless. Um, and if you um had to talk
[00:07:33] Oh, okay. We can see it as well. Would
[00:07:35] love a copy myself. What’s the one
[00:07:37] framework or principle from it that will
[00:07:39] still be relevant in the 10 years even
[00:07:41] that tools are changing every day?
[00:07:43] >> Well, there’s 48 of them. So, you have
[00:07:45] your choice. In terms of the one that I
[00:07:47] use the most and I talk about the most
[00:07:49] is uh plan big act small and this is
[00:07:52] this relates to different AI models are
[00:07:55] skilled at different things. So for
[00:07:57] example if you go into claude there’s
[00:07:58] three models opus which is the most
[00:08:00] expensive the biggest model sonnet which
[00:08:02] is the middle road model and haiku which
[00:08:04] is the fast model. If you go into open
[00:08:06] eyes any of open eye stuff you have GPT
[00:08:08] 5.5 there’s five versions of it. Go into
[00:08:10] Google Gemini there’s flash light flash
[00:08:13] and pro. So everyone’s got, you know,
[00:08:14] small, medium, large effectively. Small
[00:08:17] models are fast, efficient, very
[00:08:20] sustainable. They they use much less
[00:08:22] energy, but they’re dumber. The big
[00:08:23] models are slow, incredibly resource
[00:08:26] intensive, compute intensive. They use a
[00:08:28] lot of electricity, they use a lot of
[00:08:29] fresh water, but they deliver great
[00:08:31] results. So what people should be
[00:08:32] thinking about and this is true across
[00:08:34] the board and it will continue to be
[00:08:36] true is you want to use the biggest best
[00:08:38] tools to do your planning and then hand
[00:08:40] off the plan to a small lightweight tool
[00:08:43] to execute the plan because as long as
[00:08:44] the plan is good and has things like
[00:08:46] definitions of done and measures of
[00:08:48] success. We use what we call the 5P
[00:08:50] framework by trust insights purpose
[00:08:51] people process platform performance. If
[00:08:53] you define these things in your plans,
[00:08:55] then when you hand that off to a small
[00:08:57] model or tool, it will execute on it
[00:09:00] greater speed and lower costs than any
[00:09:03] than the big model. A lot of people kind
[00:09:05] of make the naive approach to either
[00:09:06] just use the small model all the time
[00:09:08] which gets them bad results and then
[00:09:09] they have to repeat work a lot which is
[00:09:11] a waste or they use the big model for
[00:09:12] everything which gets great results but
[00:09:14] it’s an incredible incredibly resource
[00:09:16] intensive. If you plan big, act small.
[00:09:18] This also applies to human delegation,
[00:09:20] right? The senior most people in an
[00:09:21] organization do the planning. The junior
[00:09:23] most people do the execution. This is a
[00:09:25] a timeless principle and it will remain
[00:09:27] true no matter what happens with the AI
[00:09:29] technology. What will probably happen in
[00:09:31] the next couple years is you’ll see the
[00:09:33] big heavy models be used for planning
[00:09:35] and then local models that run on your
[00:09:37] hardware or in your infrastructure will
[00:09:39] do the execution
[00:09:40] >> 100%. You’ve also talked about
[00:09:42] extensively social media ROI given that
[00:09:45] in 2026 with organic reach at all-time
[00:09:47] lows and AI agents potentially doing the
[00:09:49] buying and how should marketers think
[00:09:52] about ROI differently. How how do you
[00:09:55] comment on that?
[00:09:56] >> ROI is a financial formula. There is no
[00:09:57] flexibility is earned minus spent
[00:09:59] divided by spent. If someone is doing
[00:10:01] anything other than those two values,
[00:10:03] they’re not doing ROI. Second, if you’re
[00:10:04] not measuring ROI now, you can’t measure
[00:10:07] changes in it. So, you need to be
[00:10:08] measuring it. And third, earn minus
[00:10:10] spent divided by spent is ROI. It is a
[00:10:12] financial equation. It is one of the
[00:10:14] hardest things for marketers to measure
[00:10:15] because they very often can’t measure
[00:10:17] what they’ve earned, right? And you have
[00:10:19] to do really good multi-touch
[00:10:21] attribution to determine what credit and
[00:10:23] earning means. And then what you spend
[00:10:26] is where everybody goes off the rails.
[00:10:28] There’s hard dollars and soft dollars or
[00:10:29] hard currency, soft currency. Hard
[00:10:30] currency is you are paying currency to
[00:10:32] another vendor. Uh you’re paying
[00:10:34] currency out of your wallet to somebody
[00:10:36] else, right? So that is going to be
[00:10:38] things like software and tools, servers,
[00:10:40] etc. If you’re using we’re fans of
[00:10:42] Agorapulse for example as a social media
[00:10:44] monitoring and scheduling software
[00:10:45] that’s hard currency. The biggest spend
[00:10:47] on social media marketing is soft
[00:10:49] currency which is what you pay people to
[00:10:51] work on it right. So what is you take
[00:10:53] somebody like maybe they earn you know
[00:10:55] $100,000 a year or whatever their hourly
[00:10:57] wage is roughly $50 an hour give or take
[00:11:00] which means that every hour they spend
[00:11:02] working on social media they are
[00:11:04] spending in pay you know whatever their
[00:11:06] hourly rate is here’s the ugly secret
[00:11:08] nobody wants to talk about pretty much
[00:11:10] all organic social media is negative ROI
[00:11:13] bottom line right because attribution
[00:11:15] being what it is and what human beings
[00:11:17] cost if you are using human beings to do
[00:11:19] organic unpaid social media you are
[00:11:21] probably almost always at negative ROI
[00:11:24] and probably substantial because human
[00:11:25] beings are expensive. Now what a lot of
[00:11:27] companies are starting to do is say okay
[00:11:28] well how much can we have AI do because
[00:11:30] if you have the tools doing this the
[00:11:32] equivalent wage of like a dollar an hour
[00:11:34] then suddenly you might be more in that
[00:11:36] territory however you also have to have
[00:11:39] outputs that are actually decent in
[00:11:41] general organic social media is kind of
[00:11:43] one of those things that’s you kind of
[00:11:44] got to do it to be present but you
[00:11:47] should not be expecting great results
[00:11:49] out of it
[00:11:49] >> Christopher we also keep hearing AI
[00:11:51] powered in every SAS pitch these days
[00:11:53] but drivers are numb to it when a SAS
[00:11:55] company comes to trust insights let’s
[00:11:57] say and says help us stand out what do
[00:12:00] you tell them to stop doing or start
[00:12:02] doing
[00:12:03] >> AI powered is like saying electricity
[00:12:04] powered or internet powered right it’s
[00:12:06] table minimum now so the question is
[00:12:08] what do you do and how do you do it
[00:12:10] bigger better faster or cheaper than a
[00:12:12] competitor I mean that’s what it boils
[00:12:13] down to that’s what we still all care
[00:12:14] about we want bigger better faster
[00:12:16] cheaper and if you are not checking at
[00:12:18] least two of those boxes you’re going to
[00:12:20] have a hard time convincing somebody to
[00:12:22] buy whatever the thing is that you’re
[00:12:24] that is for sale so in the age of AI,
[00:12:27] particularly for SAS. SAS is in a lot of
[00:12:29] trouble right now because you can sit
[00:12:32] down with Clawed Code, which costs 200
[00:12:34] bucks a month for the top level
[00:12:35] subscription, and say, “I don’t like my
[00:12:37] CRM. Here’s the five things I don’t like
[00:12:39] about it, write me a new one.” And if
[00:12:41] you don’t care about, you know,
[00:12:43] switching costs and things like that,
[00:12:44] you’re probably going to get a pretty
[00:12:45] decent substitute because as the tools
[00:12:47] get better and better, as Aentic AI gets
[00:12:49] better and better, and particularly at
[00:12:51] coding, which is what what all the AI
[00:12:53] companies have optimized for, you will
[00:12:55] get really good results. You will get
[00:12:58] stuff that maybe is unique to your
[00:13:00] company in a way that a major vendor
[00:13:02] could never be because they have to
[00:13:04] satisfy a lot of customers whereas your
[00:13:05] code has to just satisfy you. So for
[00:13:08] bigger, better, faster, cheaper and have
[00:13:10] some kind of compelling value that makes
[00:13:13] it better than someone telling Claude
[00:13:15] code, “Hey, make the thing.”
[00:13:16] >> Chris, you’re everywhere yourself. Be it
[00:13:18] YouTube with, you know, 2500 plus
[00:13:20] videos, LinkedIn, Substack newsletter
[00:13:23] with 98,000 subscribers, your own
[00:13:25] podcast for, you know, more than 19
[00:13:27] years now. How do you manage that? And
[00:13:30] if if you if you have to be honest, how
[00:13:32] much is it AI versus you? I manage it by
[00:13:35] not having many friends and I don’t have
[00:13:37] a television. Um so a lot of the time
[00:13:40] sucks the uh of things like watching
[00:13:42] television don’t exist. I for my content
[00:13:45] I rarely use AI. I mostly so for example
[00:13:48] my daily YouTube videos I shoot them on
[00:13:51] Sunday mornings while I’m cooking uh the
[00:13:53] the Sunday meal. So you will notice in
[00:13:55] all of them I’m usually making something
[00:13:57] and it’s not a cooking show but it is
[00:13:58] definitely uh you know something that
[00:14:00] I’m working on and I just strap on a DJI
[00:14:03] microphone and I record and you know the
[00:14:04] video quality is not great but it’s
[00:14:06] guaranteed authentic. Um same for my
[00:14:08] newsletter. So my newsletter is
[00:14:10] typically what I do is while I’m driving
[00:14:13] um to and from my martial arts school on
[00:14:16] Saturday morning I have uh my lavalier
[00:14:19] mic on. I have my voice memos app
[00:14:20] recording running on my phone and I just
[00:14:22] foam at the mouth for a full hour which
[00:14:24] you know is like 6,000 words of talking
[00:14:27] and then as I work out my thoughts and
[00:14:28] stuff like that then I take it to Claude
[00:14:30] and I say okay let’s organize my words
[00:14:33] let’s remove the speech disfluencies
[00:14:34] like the stumbling and the swearing at
[00:14:36] other drivers and then I will turn it
[00:14:38] into a script and then I follow the
[00:14:40] script and make continue to make changes
[00:14:41] and then I record it. So in general very
[00:14:44] little of my content is AI generated
[00:14:45] because it’s not how I work that’s not
[00:14:48] how my brain works. stuff that I I wish
[00:14:50] that Claud and and the various tools did
[00:14:52] a better job, but for whatever reason,
[00:14:55] and I’ve done, you know, styometry tests
[00:14:57] and things, my particular style is very
[00:14:59] hard for it to imitate. Well, usually
[00:15:02] because there’s some weird quirk or some
[00:15:05] analogy that the machine just can’t
[00:15:06] forecast well. For example, I use this
[00:15:08] when I talk about writing style. Good
[00:15:10] writing is inherently low probability,
[00:15:13] right? So, if I say, you know, I the the
[00:15:15] patient was in gastric distress, that is
[00:15:17] a high probability phrase. Not
[00:15:18] interesting, not compelling, doesn’t
[00:15:20] catch your attention. If I say it looks
[00:15:21] like they power wash their restroom with
[00:15:23] Nutella, that’s a very low probability
[00:15:24] phrase. It’s gross, but a machine is not
[00:15:26] going to predict that well and stuff.
[00:15:28] So, these kinds of weird low probability
[00:15:30] phrases that make good writing are
[00:15:32] inherently harder for machines to
[00:15:34] predict. So, unless I provide them, the
[00:15:36] machine has a harder time forecasting
[00:15:37] and then by that point, I may as well
[00:15:39] just write the thing myself.
[00:15:41] >> No. Yes. One of the biggest shifts in
[00:15:43] B2B right now is a founder distribution.
[00:15:46] CEOs and CTOs building in public. You’ve
[00:15:49] been doing it for years, too. What’s the
[00:15:51] difference between a founder who uses
[00:15:53] social media effectively versus the one
[00:15:55] who just looks like they’re trying too
[00:15:57] hard?
[00:15:58] >> In terms of using social media
[00:16:00] effectively, it comes down to
[00:16:01] motivation. Why are you doing it? Why
[00:16:03] are you out there at all? Obviously, you
[00:16:06] have self-interest, right? I want to
[00:16:08] attract more customers. I want to
[00:16:09] demonstrate that I know what I’m talking
[00:16:11] about. etc. Those would be, you know,
[00:16:13] self-led motivations. However, you also
[00:16:15] have to have some sort of extrinsic
[00:16:17] motivation. So, I want to teach people
[00:16:19] how to do this thing. I want to be
[00:16:20] helpful to people. And a lot of folks
[00:16:22] who don’t do well don’t have that
[00:16:25] motivation of I’m going to give away my
[00:16:27] knowledge and I don’t care whether you
[00:16:29] buy from me or not because I’m going to
[00:16:31] show you what’s going on. That’s a good
[00:16:33] motivation. A bad motivation is I’m
[00:16:34] going to gatekeep or I’m going to come
[00:16:36] across as, you know, I’m going to to
[00:16:39] tease bits and pieces, but I’m not going
[00:16:41] to be actually all that helpful to you.
[00:16:43] That comes across pretty easily. And you
[00:16:45] just know like going out to a a bar,
[00:16:47] right, and you’re talking to a person,
[00:16:48] you have a fairly good idea of that
[00:16:50] person’s intentions very, very quickly.
[00:16:52] And some people you want to hang out
[00:16:53] with, some people you don’t want to hang
[00:16:55] out with. And that comes across founders
[00:16:56] who want to be successful building in
[00:16:58] public. What’s your motivation? And is
[00:17:01] your motivation equally balanced between
[00:17:03] helping yourself and helping other
[00:17:05] people? If it’s not, you’re not going to
[00:17:06] do well.
[00:17:06] >> For a CTO or CIO trying to build a
[00:17:09] leadership profile in 2026, what’s your
[00:17:12] take with social platforms still matter
[00:17:14] and which are dead weight?
[00:17:15] >> It depends on your audience, which is a
[00:17:17] completely unsatisfying answer. But for
[00:17:20] leadership stuff, your best bet is still
[00:17:23] email, right? Is still because it is the
[00:17:25] one channel that is not owned by any
[00:17:27] large corporation. Everything else is
[00:17:29] owned by some large corporation and they
[00:17:30] can turn the dials on you at any time
[00:17:32] and so you can spend a lot of time
[00:17:33] building a platform there and it
[00:17:35] vanishes. For example, I use Substack
[00:17:36] because of one button and that button is
[00:17:39] export all. I can hit that button, I can
[00:17:41] take my list with me and go somewhere
[00:17:42] else. I would not be on that platform if
[00:17:44] I didn’t have the ability to remove my
[00:17:46] list immediately and go somewhere else.
[00:17:48] I participate on LinkedIn because like
[00:17:50] that’s where a lot of my audience is. Uh
[00:17:52] I participate on threads mostly to
[00:17:54] complain out loud. I participate on
[00:17:56] Instagram just mostly for my friends.
[00:17:58] It’s not for I mean I I share publicly
[00:18:00] but I don’t create content there for the
[00:18:02] purposes of business. It depends on your
[00:18:03] audience. Where is your audience? Right?
[00:18:05] If your audience is on Tik Tok, you need
[00:18:07] to be there. Sorry, but that’s you know
[00:18:08] the truth. If your audience is on Waybo,
[00:18:10] if your audience is on Pinterest or 500
[00:18:13] pixels or any of the, you know,
[00:18:15] thousands of social media social
[00:18:17] channels there are, you got to go where
[00:18:19] they are and you have to, you know,
[00:18:20] provide useful stuff there. your
[00:18:22] audience is on what you know some of the
[00:18:23] the crazier places on the internet and
[00:18:25] if that’s where your audience is you got
[00:18:27] to be there like if you’re hawking NFTTS
[00:18:28] still uh you’re probably on a place I
[00:18:30] don’t know like Truth Social or
[00:18:32] something any place you can get somebody
[00:18:33] to buy one of those things uh in 2026
[00:18:35] you should from an AI perspective you
[00:18:38] should be on YouTube because YouTube is
[00:18:40] the number one training source for all
[00:18:42] AI models and so if you want to be doing
[00:18:44] stuff like GEO etc you kind of need to
[00:18:46] be on YouTube.
[00:18:47] >> Yeah. Yeah. and and Chris for um you
[00:18:49] know given if we say that LinkedIn is
[00:18:51] the default for you know B2B SAS perhaps
[00:18:54] perhaps and if you had to pick you know
[00:18:56] the second platform for a SAS company
[00:18:58] that they should invest in within 2026
[00:19:01] what is it and why
[00:19:02] >> email still email
[00:19:03] >> 100% I mean
[00:19:04] >> in fact LinkedIn is second emails first
[00:19:07] >> all right yeah and what should these
[00:19:10] people be careful about in their emails
[00:19:13] given that you know with your one prompt
[00:19:15] now claude can churn or any AI can
[00:19:17] return thousands of emails for you but
[00:19:20] you would still not get any results back
[00:19:22] on them how you should be doing emailing
[00:19:24] in 2026
[00:19:26] >> it still goes back to the thought
[00:19:27] leadership angle right if I open an
[00:19:29] email as a reader what’s in it for me
[00:19:31] yeah
[00:19:31] >> right because most emails there is
[00:19:33] nothing in it for me it is emails are
[00:19:35] being sent from a very selfentric
[00:19:37] selfish perspective hey I want you to
[00:19:39] give me money right what’s if I open up
[00:19:41] your email what is in it for me if
[00:19:43] there’s nothing of value and by the way
[00:19:45] your product does not count as value to
[00:19:48] me. If there’s nothing in it for me, I’m
[00:19:50] not going to read it and I’m going to
[00:19:51] unsubscribe or I’m going to mark it as
[00:19:52] spam. Have to figure out a lot of people
[00:19:54] love the expression uh you know giving
[00:19:56] value or creating value and they have
[00:19:58] the foggiest idea of what that means. Um
[00:20:00] it’s a sort of a a buzzword, a trope in
[00:20:02] in marketing. What is in it for me the
[00:20:05] reader? Assuming I never ever ever buy
[00:20:08] anything from you. What is in it for me?
[00:20:10] Right. One of the biggest things that
[00:20:12] for example my my Substack email list is
[00:20:15] now up to almost 300,000 readers, right?
[00:20:17] So it in terms of subscriber base, it’s
[00:20:20] literally a small city. When I send that
[00:20:22] out, I assume literally maybe five
[00:20:24] people will ever buy something from any
[00:20:26] given issue, right? Out of 300,000, the
[00:20:29] vast majority of people are never going
[00:20:30] to buy anything. So I have to provide
[00:20:32] them value enough to eventually refer
[00:20:35] somebody else that will buy something or
[00:20:37] refer somebody else to subscribe to the
[00:20:39] newsletter. So I have to create value in
[00:20:42] by giving away knowledge, by giving away
[00:20:44] information, by giving away perspective.
[00:20:46] The issue I just did this past weekend
[00:20:48] was 18 ways to reduce your token usage
[00:20:51] with AI, right? No one’s going to hire
[00:20:52] me to do that. No one should hire me to
[00:20:54] do that. Just read the newsletter. But
[00:20:56] that’s there’s value there. There’s a
[00:20:57] lot of value there. Particularly for
[00:20:58] companies that are concerned or
[00:21:00] individuals who are like, I have the $20
[00:21:01] a month subscription and I’m tired of
[00:21:02] seeing you’ve hit your your usage limit
[00:21:05] on, you know, day one of seven. the
[00:21:07] stuff in that newsletter will help those
[00:21:08] people and eventually some small
[00:21:11] percentage of them will say I wonder
[00:21:12] what else they do and potentially buy
[00:21:15] something whether it’s a book a course
[00:21:17] hiring us as a consulting agency but the
[00:21:19] biggest mistake everybody makes in
[00:21:21] marketing is they see their audience as
[00:21:24] a bunch of walking wallets and there’s
[00:21:26] like how can I pick this person’s pocket
[00:21:27] as quickly as possible and it’s no
[00:21:29] wonder like yeah nobody reads your
[00:21:31] emails because everybody who reads your
[00:21:32] emails knows there’s nothing in it for
[00:21:34] them
[00:21:34] >> yeah and you know um If we dive a bit
[00:21:37] into the human versus machine questions
[00:21:39] now, will marketing and content be fully
[00:21:42] taken over by AI or do you think there
[00:21:44] is a ceiling? And if there is, then
[00:21:46] where’s that line?
[00:21:48] >> So AI is is destroying social media
[00:21:51] pretty hilariously, right? So you just
[00:21:53] see more and more AI slop, bad use of
[00:21:55] social media. You see, for example,
[00:21:56] LinkedIn being overrun by comment bots
[00:21:58] and you could tell cuz that the same
[00:22:00] person leaves the exact same formatted
[00:22:02] structured comment on every blog post.
[00:22:03] And so you should plan as a marketer for
[00:22:05] social media to get even less and less
[00:22:08] effective as AI continues to just
[00:22:10] pollute it, right? Because you have
[00:22:11] inept marketers who are looking for an
[00:22:13] easy button just running bots, you know,
[00:22:15] 24/7. In terms of content marketing, the
[00:22:19] internet is being overrun. Uh something
[00:22:21] I I saw stat I want to say it was it was
[00:22:24] some company. It was some content
[00:22:25] company, but roughly 50% of the content
[00:22:28] being published on the internet now is
[00:22:30] machinemade. So things like search are
[00:22:32] going to be impacted more heavily.
[00:22:34] You’re seeing Google saying like, “Yeah,
[00:22:35] we can differentiate between
[00:22:36] non-commodity and commodity content.”
[00:22:38] And commodity content is what machines
[00:22:41] tend to spit out because they’re not
[00:22:42] well they’re not well structured and
[00:22:44] built. As this happens more, your
[00:22:45] marketing channels you have access to
[00:22:47] will get less and less valuable. you’ll
[00:22:49] be able to attract fewer and fewer
[00:22:50] customers from it. Which means the only
[00:22:52] mechanism you have for growing your
[00:22:55] reach is through the human beings who
[00:22:57] are going to refer other human beings to
[00:23:00] you. Which again goes back to things
[00:23:01] like email, like what’s in it for me? If
[00:23:04] I interact with you as a company on any
[00:23:06] platform or in any space, what’s in it
[00:23:08] for me? If the answer is nothing, then
[00:23:10] until I have a need, which is like, you
[00:23:12] know, 1% of the time, I’m not interested
[00:23:14] in interacting with you. And if you’re
[00:23:16] constantly asking me, hey, are you ready
[00:23:17] to buy from us yet? The answer is going
[00:23:19] to be goodbye. And that will be that.
[00:23:20] Marketers have a very very difficult
[00:23:23] task ahead of them, which is a being
[00:23:25] heard over the title wave of slop being
[00:23:28] created by inept use of AI. And I I I
[00:23:31] keep saying that because you can do very
[00:23:33] well with skillful use of AI. You can
[00:23:36] provide value to people with skillful
[00:23:38] use of AI. What we see most marketers
[00:23:40] doing is not skillful use of AI.
[00:23:42] >> There is also a hot debate right now,
[00:23:44] Chris Grace. Um, will AI agents scale
[00:23:46] SAS or just force SAS to evolve? And you
[00:23:49] worked with major brands on this? What’s
[00:23:50] your take on it
[00:23:51] >> in terms of like AI just replacing SAS?
[00:23:54] >> Yeah.
[00:23:55] >> Yeah. Well, think about it. How stupid
[00:23:56] it is, right? In terms of a of a
[00:23:58] business model, if I can have if I if
[00:24:00] you’re going to charge me money month
[00:24:02] over month for a service that had better
[00:24:04] do that has to do better than what I can
[00:24:06] do myself now.
[00:24:07] >> Yeah.
[00:24:07] >> So, what is your value proposition as a
[00:24:09] SAS company that a machine can’t do?
[00:24:12] That’s the key question that a SAS
[00:24:14] company has to ask. It is not the code.
[00:24:16] Your code is a commodity. Your fancy
[00:24:18] features are a commodity. And in
[00:24:20] literally 20 seconds, I could take a
[00:24:22] screenshot and a voice recorder and say,
[00:24:24] “Hey, Claude, make my version of this,
[00:24:26] but make it better.” I saw there was
[00:24:28] they actually saw this not too long ago.
[00:24:29] There was this one event where there was
[00:24:31] a keynote and there was it was a paid
[00:24:33] keynote for a SAS company and they’re
[00:24:34] like, “Our software is this this is
[00:24:36] exactly what marketers need. You know,
[00:24:38] you’ll get a one month free trial and
[00:24:39] it’s only whatever month after that.”
[00:24:41] I’m like, “Okay, screenshot, voice
[00:24:42] prompt.” And within two hours, Claude
[00:24:45] had replicated their entire stack
[00:24:47] because there there was nothing special
[00:24:48] about the software. It was it was a
[00:24:50] completely commodity. Their sales pitch
[00:24:53] their what they thought their value was
[00:24:55] was it was low priced enough that your
[00:24:57] average marketer would buy it. No,
[00:24:59] Claude can make that, right? Quinn can
[00:25:00] make that. Chat GPT, OpenAI can make
[00:25:02] that. You can make pretty much anything.
[00:25:04] So, if you think about the value chain,
[00:25:06] if you have commodity, which is the
[00:25:08] lowest bar, all code is commodity. Your
[00:25:10] software is commodities. Sorry to say
[00:25:12] it, it’s true. Your code has no value by
[00:25:14] itself. Which means that you have to do
[00:25:16] things up the value chain like create a
[00:25:18] brand around that. What is and and how
[00:25:20] do you do that? Well, that means
[00:25:21] extending commodity to some kind of
[00:25:23] service that is a pain point that the
[00:25:26] commodity itself can’t answer. So,
[00:25:28] anyone can code, right? Anyone can have
[00:25:30] Claude build and maintain software. What
[00:25:32] is the next thing up in the value chain
[00:25:34] that somebody does not want to do?
[00:25:35] Maintaining code still takes time. It’s
[00:25:37] boring. It’s not fun. Fixing bugs is not
[00:25:39] fun. And yes, tools like clawed code and
[00:25:41] open code and stuff make that very easy
[00:25:42] now, but it’s still not fun. So your
[00:25:45] value might be that level of service.
[00:25:47] Your value might be the training to help
[00:25:49] the humans operate the software better.
[00:25:51] Your train your your value might be the
[00:25:53] experience that somebody has using the
[00:25:55] software that can be augmented in some
[00:25:57] other way beyond the software. But in
[00:25:59] 2026, if you think your software is a
[00:26:01] competitive advantage, you are
[00:26:02] delusional.
[00:26:03] >> You’ve warned about desklling, the
[00:26:05] offloading of executive function to AI.
[00:26:08] You use AI yourself to challenge your
[00:26:10] thinking, not replace it. Can you tell
[00:26:12] us how you structure a prompt to make AI
[00:26:15] push back on you instead of just
[00:26:17] agreeing with you?
[00:26:18] >> Yeah, it’s one sentence. Ask me
[00:26:20] questions until you have enough
[00:26:21] information to complete the task. Right?
[00:26:22] That’s it. That is the sentence. Put it
[00:26:24] at the beginning in your prompts. If
[00:26:25] you’re using Agentic systems, there are
[00:26:26] even modules like brainstorming skills
[00:26:28] and things that uh can come pre-built
[00:26:30] into a system like Claude Co-work or
[00:26:32] Open Code or or Anti-gravity where you
[00:26:35] say to the model, “Hey, ask me
[00:26:37] questions. Let’s brainstorm this a
[00:26:39] challenge me on these assumptions. What
[00:26:41] have I forgotten? I have learned so much
[00:26:44] more from AI than I have ever learned
[00:26:46] before in in terms of coding because I
[00:26:48] will say, “Hey, we’re doing this
[00:26:49] project. Here’s what I think it should
[00:26:51] be. What libraries or packages don’t I
[00:26:53] know about that would be a good fit for
[00:26:55] this? Use your web search tools.” And it
[00:26:56] comes back and I’m like, “Holy crap, I
[00:26:58] didn’t even know this thing existed.” Or
[00:26:59] in some cases, I I will do and we’ll
[00:27:01] say, “Hey, by the way, I got bad news
[00:27:03] for you. Your idea already exists. Why
[00:27:05] don’t you just go use that?” like, “All
[00:27:06] right, well, I will just go use that
[00:27:08] because that’s better.” But it requires
[00:27:10] the thing that has to happen is that the
[00:27:13] human being has to have some form of
[00:27:14] self-awareness to go, “Huh, what don’t I
[00:27:17] know?” And or what if I don’t treat this
[00:27:19] like the easy button? And that is where
[00:27:21] most people go wrong. They’re like,
[00:27:23] “Hey, just do this thing for me.” And it
[00:27:24] does. That’s great. You’re no longer
[00:27:26] planning. You’re no longer organizing.
[00:27:27] You’re no longer making decisions.
[00:27:29] You’re no longer solving problems, which
[00:27:30] are the four core tenants of executive
[00:27:33] function. When you hand that over to a
[00:27:35] machine, you desill. When the machine is
[00:27:37] asking you to plan, when the machine is
[00:27:39] asking you to organize, when the machine
[00:27:41] is asking you to make decisions, when
[00:27:43] they ask the machine’s asking you to
[00:27:44] solve its problems, you are upskilling,
[00:27:46] not downskilling. So in your use of
[00:27:48] these things in in your planning
[00:27:50] prompts, for example, you should be
[00:27:51] saying, “Hey, you need to keep asking me
[00:27:53] questions until you fixed my blind
[00:27:55] spots.” You will get great results out
[00:27:57] of the machines that way and you will
[00:27:59] get great results out of your own brain.
[00:28:00] No, that’s that’s very true because I’
[00:28:02] I’ve done this myself as well and I’ve
[00:28:04] learned a lot in the process of it. But
[00:28:06] I’ve also figured that as you’ve also
[00:28:08] said that if you were giving AI the same
[00:28:10] instructions you’d give to a contractor,
[00:28:12] the company doesn’t need you perhaps.
[00:28:14] And and that’s very brutal and honest,
[00:28:16] which is kind of needed by most of the
[00:28:18] marketers. But what should marketers
[00:28:19] today be doing to stay indispensable?
[00:28:22] >> The easiest answer, what do you do that
[00:28:25] is not templated? So as a marketer,
[00:28:27] think about all the things that are
[00:28:28] templates, right? your website design,
[00:28:30] your email nurture campaign, your CRM
[00:28:34] strategy, your social media posts, your
[00:28:37] presentations, your reports, all the
[00:28:39] stuff. If it’s I’ve been saying this
[00:28:41] since 2015, I said this on stage at
[00:28:42] Social Media Marketing World in 2015, no
[00:28:44] one believed me. If you do it with a
[00:28:46] template today, a machine does it
[00:28:47] without you tomorrow. Look honestly.
[00:28:50] Look, take a hard honest look at your
[00:28:53] work. What of the things you do every
[00:28:55] day is a template or a standard
[00:28:57] operating procedure in some way in some
[00:28:59] fashion or form. If it is a template
[00:29:01] today, it will be gone from your plate
[00:29:04] tomorrow. So if your job is 95%
[00:29:07] templates, I’m doing the social media
[00:29:08] template now. I’m following this
[00:29:09] procedure. I’m doing the blog template
[00:29:11] now. You’re gone. You there, you are not
[00:29:13] indispensable. A machine can do what you
[00:29:15] do. If you are in Canva all day, I’m
[00:29:17] making templates in Canva. Yeah, you’re
[00:29:19] you’re going that that work is going to
[00:29:21] go away. What is left of your job that
[00:29:23] is not templated and and is valuable,
[00:29:26] right? Cuz yeah, you can sit in meetings
[00:29:27] all day, you can know chat on Slack all
[00:29:30] day, but what is left of your job that
[00:29:32] is not templated and valuable? And that
[00:29:34] is a very painful question for a lot of
[00:29:36] people answer because our organizations
[00:29:38] overall have pushed us to templatize and
[00:29:42] operationalize everything in the name of
[00:29:44] productivity for the last 30 years.
[00:29:46] Well, the natural logical conclusion to
[00:29:48] that for 20 years was outsourcing. we’re
[00:29:50] going to send this to someplace where
[00:29:51] the wages are cheaper and someone else
[00:29:52] can follow the template. And now the
[00:29:54] machines were saying, well, now the
[00:29:55] machines are even cheaper than
[00:29:56] outsourcing. So, we’re going to send
[00:29:58] this all to the machines, right? One of
[00:29:59] the the biggest the largest populations
[00:30:01] of people that will be harmed by AI are
[00:30:04] developing nations that have hosted
[00:30:05] things like call centers and stuff
[00:30:06] because human beings, yeah, you paid
[00:30:08] them a dollar an hour, but now a machine
[00:30:10] can do it for a penny an hour. And so
[00:30:11] you were going to have a large
[00:30:13] withdrawal of income from developing
[00:30:16] places where they were used to low paid
[00:30:18] but reasonable machine level work. Now
[00:30:20] it’s just entirely machine-based. And so
[00:30:22] that be a a substantial geopolitical and
[00:30:25] socioeconomic shift. But for the
[00:30:27] marketer, what do you do that is not
[00:30:29] templated and valuable? That’s what will
[00:30:31] make you indispensable. And if the
[00:30:33] answer is nothing or it’s less than 50%
[00:30:35] of your job, you are in trouble.
[00:30:37] >> Chris, let’s talk a bit about legacy
[00:30:39] thinking in a modern world. A lot of
[00:30:40] companies especially in enterprise are
[00:30:43] sitting on legacy marketing stacks,
[00:30:45] legacy processes, legacy assumptions
[00:30:48] about how buyers behave and it’s a lot
[00:30:51] you know and at rather a large scale and
[00:30:53] you’ve worked with brands you know big
[00:30:55] brands like McDonald’s and T-Mobile when
[00:30:57] you walk into a company which is still
[00:30:59] running its marketing like it’s 2019 um
[00:31:02] what’s the first thing you modernize
[00:31:04] that’s one and two is there any
[00:31:06] framework that you would recommend them
[00:31:08] to start with and from where The
[00:31:10] framework that we use that we live by is
[00:31:12] the 5P framework by trust insights. It’s
[00:31:13] the one I mentioned several times.
[00:31:15] Purpose, people, process, platform,
[00:31:16] performance. When you go into anywhere
[00:31:18] to do anything, you say what are you
[00:31:20] doing? Why are you doing it? Who’s doing
[00:31:22] it? How are you doing it? What are you
[00:31:23] doing it with? And how do you know that
[00:31:25] you succeeded? And when you look at most
[00:31:26] organizations, they have failed to
[00:31:28] define most of those categories for
[00:31:31] anything legacy or modern, right? They
[00:31:33] will say, you know, we’re we’re doing
[00:31:34] this email campaign, you know, this is
[00:31:36] our this is our marketing automation
[00:31:37] system. and you look at it and go,
[00:31:38] “Okay, why are you doing this?” Cuz we
[00:31:40] should. No, that’s not a good reason.
[00:31:42] Why are you doing, you know, why are you
[00:31:44] doing these drip campaigns? Are they
[00:31:45] working? How would you know? I love when
[00:31:47] people bust out the, “Oh, uh, you know,
[00:31:49] we need to have positive ROI.” Well,
[00:31:51] you’re not measuring it or you’re not
[00:31:52] measuring it properly. So, that that’s a
[00:31:54] meaningless goal. Anytime you go into
[00:31:56] any organization, legacy or modern, you
[00:31:58] run the 5P framework by trust insights
[00:32:00] to say, “What are you doing? Why are you
[00:32:02] doing it?” And most of the time, people
[00:32:04] do not have good answers. And that’s
[00:32:06] where you start any kind of
[00:32:07] modernization effort is to say, you
[00:32:09] know, what is your actual goal? I mean,
[00:32:10] fundamentally, if you’re a business, the
[00:32:12] the end goal is you have to make money.
[00:32:14] And you probably is to make more money
[00:32:15] rather than less, right? So that’s the
[00:32:17] bare minimum goal. And then you start to
[00:32:18] say, okay, well, how do you make money?
[00:32:20] Who do you make money from, right?
[00:32:22] What’s in it for them and stuff? You
[00:32:25] look at a restaurant business for
[00:32:26] example, how do you make money? People
[00:32:28] come in to eat. Okay? Well, how could
[00:32:30] you make more money? You can, you know,
[00:32:31] charge more for the food. You can give
[00:32:33] them less food. That’s called
[00:32:34] shrinkflation. Or you can give them
[00:32:36] ideally different food than they can get
[00:32:39] elsewhere. That makes you that that is a
[00:32:41] differentiator for you. In that case,
[00:32:44] okay, then now then now you can say,
[00:32:45] well, what can we modernize or what can
[00:32:47] we change that help you achieve those
[00:32:50] things of doing something bigger,
[00:32:51] better, faster, cheaper, right? We’re
[00:32:53] we’re back to the the main four. That’s
[00:32:55] the blueprint, right? That is the
[00:32:56] blueprint for everything and anything.
[00:32:58] No ma what people get wrong. My CEO
[00:33:00] Katie River uh says this all the time.
[00:33:03] People lead with the technology and
[00:33:05] that’s the wrong place to start. It’s
[00:33:07] like leading with appliances. They say,
[00:33:08] “We’re going to use blenders. We are all
[00:33:10] in on blenders. Blenders are the thing.”
[00:33:11] And you’re like, “Yeah, but we’re a
[00:33:12] steak restaurant. The only thing we can
[00:33:13] blend is sauce, right? No one wants a
[00:33:15] blended steak. Uh, that’s disgusting,
[00:33:17] right?” And so if you were starting with
[00:33:19] the technology, “Yeah, we’re all in on
[00:33:20] blenders.” Yeah, but you’re a steak
[00:33:21] restaurant. Blenders are the wrong tool.
[00:33:23] If you want to be all in on something,
[00:33:24] be all in on a better grill. And you’re
[00:33:26] like, “Yeah, but nobody grills aren’t
[00:33:27] sexy. Nobody’s talking about grills.”
[00:33:29] Yeah, but that doesn’t matter to your
[00:33:30] business. Your business is a steak
[00:33:32] restaurant. You need better grills, not
[00:33:34] better blenders. I don’t care what
[00:33:35] everybody else is doing with the the
[00:33:37] smartest new automated blenders. It
[00:33:39] won’t help your business. That’s why she
[00:33:41] created the 5P framework to say purpose
[00:33:43] and performance are bookending this
[00:33:45] thing. Forget about digital
[00:33:46] transformation. Forget about people
[00:33:48] process technology. Forget about the
[00:33:50] fancy flashy stuff. Focus on why you’re
[00:33:53] doing it and how do you know you’re
[00:33:54] succeeding? And if you can’t clearly
[00:33:56] articulate those two bookends, that’s
[00:33:58] where you start your modernization
[00:33:59] efforts.
[00:34:00] >> Yeah. But Chris, you know, most of the
[00:34:02] companies that they don’t just initially
[00:34:04] start with a mission or a vision. They
[00:34:06] would rather just start and then things
[00:34:08] might work out for them and then now
[00:34:10] there comes a point after some years and
[00:34:12] they then they sit together and they’re
[00:34:14] like okay what’s the mission or what’s
[00:34:16] the vision and what do we actually want
[00:34:18] to get out of this right and then that I
[00:34:20] I believe that that’s relatable for many
[00:34:22] of the CIOS or CTOs or perhaps IT
[00:34:25] directors out there who are now founders
[00:34:27] of their ventures too. So, how would you
[00:34:30] um kind of comment on that?
[00:34:31] >> You start by saying, “What are we doing
[00:34:33] and why are we doing it?” You have to
[00:34:34] say, “Okay, why are we doing it?” And
[00:34:36] then you go to the people part. Who’s
[00:34:38] going to buy it and why? And if you’re
[00:34:39] if you are a founder and you are
[00:34:40] struggling, it’s probably because you
[00:34:42] haven’t answered one of those two
[00:34:44] questions. Why are we doing it? And why
[00:34:45] would people buy from us? And if you
[00:34:47] can’t articulate that, that’s where you
[00:34:49] start. That’s where you start fixing
[00:34:51] things to say, how can we provide value
[00:34:53] to people that would actually convince
[00:34:55] them to give us money? And it may be
[00:34:56] that may be a pivot, right? Think about
[00:34:58] some of the really big pivots. There was
[00:35:00] a company in Japan in occupied Korea
[00:35:03] during World War II uh called
[00:35:05] Mitsuboshi. They sold dried fish and
[00:35:07] seaweed. They were an exporter importer
[00:35:09] import export business after the war and
[00:35:11] after they were liberated from uh the
[00:35:13] Japanese occupation. They went back to
[00:35:16] their proper Korean name and today they
[00:35:19] still exist. They are known as Samsung.
[00:35:22] Right? Samsung there. You will not find
[00:35:24] dried fish or seaweed anywhere in a
[00:35:26] Samsung store. you’ll find, you know,
[00:35:27] tablets and phones. They are still an
[00:35:29] importer exporter. They still have that
[00:35:31] DNA of of making cool stuff and and
[00:35:33] shipping it everywhere in the world, but
[00:35:35] they are no longer dried fish, right?
[00:35:37] So, they pivoted, but they kept the DNA
[00:35:40] of their company. You as a founder have
[00:35:42] some sort of DNA that makes your
[00:35:44] business uniquely yours, but you will
[00:35:46] pivot in its application if you are
[00:35:49] unclear about the five Ps. You should
[00:35:50] pivot so that you can provide the value
[00:35:53] that your customers are going to be
[00:35:54] willing to pay money for.
[00:35:55] >> Absolutely. Chris, there’s this growing
[00:35:57] trend of productiz services that
[00:35:59] companies are starting with a fixed
[00:36:01] scope service learning the workflow then
[00:36:03] turning the repeatable parts into SAS
[00:36:06] and then you’ve seen this play out. When
[00:36:08] do you think that this really works?
[00:36:10] >> It works less and less now because again
[00:36:12] if you look at agentic AI you can hand
[00:36:14] those parts or even better you can hand
[00:36:16] the outcomes of the parts to a machine
[00:36:18] and have it do it better. When you work
[00:36:21] with things like agentic AI, if you are
[00:36:23] clear and this is again why the 5P
[00:36:24] framework by trust insights matters
[00:36:26] because if you’re clear about the
[00:36:26] purpose and you’re clear about the
[00:36:27] performance and you know what the
[00:36:28] outcome is, it can reverse engineer how
[00:36:30] do I get from start to finish without
[00:36:32] the messy middle. There are entire
[00:36:34] companies there building clean rooms
[00:36:36] using AI to say here’s our competitor’s
[00:36:38] product. Here’s the outcome we want.
[00:36:39] Reverse engineer how to get there but do
[00:36:42] it your way AI instead. And so we’re not
[00:36:44] we’re using zero lines of code from
[00:36:46] somebody else’s product, but we’re going
[00:36:48] to get to a better outcome. Right? So if
[00:36:51] we think about how these tools can
[00:36:53] accomplish this task of achieving a
[00:36:56] specific outcome, then when a, you know,
[00:36:58] somebody says, I want to take these
[00:36:59] components and and turn it to SAS. It’s
[00:37:01] just as easy for a person skilled at AI
[00:37:04] to say, I’m going to take your SAS and
[00:37:05] turn it back into components, and I’m
[00:37:07] going to do it my way and do it better
[00:37:08] than than you did. Because a lot of
[00:37:10] folks who are in SAS tend to have
[00:37:13] software development blind spots, pretty
[00:37:15] big ones, as evidenced by the fact that
[00:37:17] you are seeing more and more cyber
[00:37:18] security issues. Security is one of the
[00:37:20] things SAS folks think about the least.
[00:37:23] Uh, and this is evidenced by the number
[00:37:25] of lawsuits and you know, oops, our
[00:37:28] database got compromised. U, go look at
[00:37:30] shiny hunters, the hacking group. You
[00:37:32] can go on to their website on the dark
[00:37:34] web and see how many businesses the
[00:37:36] crown jewels of the business are
[00:37:37] literally sitting on a website for
[00:37:39] someone to download because nobody
[00:37:41] thought about security um or that they
[00:37:43] did it was an afterthought. The
[00:37:44] challenge with folks taking, you know,
[00:37:46] chunks of stuff and turning into SAS is
[00:37:48] that if you don’t do it with great
[00:37:51] overall development practices, you make
[00:37:54] a halfbaked product that has is filled
[00:37:56] with holes and then in, you know, should
[00:37:59] your company take off, you get
[00:38:01] compromised right away.
[00:38:02] >> Yeah. Let’s talk a bit about AI slob.
[00:38:04] What’s your once intense rule for making
[00:38:06] AI generated content sound human?
[00:38:09] >> Uh, learn styometry. With thirdparty
[00:38:11] cookies dying and AI agents doing
[00:38:13] research on behalf of users, how do you
[00:38:15] even measure what’s working in marketing
[00:38:16] anymore?
[00:38:18] >> Uh the number one way to do this and the
[00:38:21] best way to do is the way that is
[00:38:22] durable and will endure pretty much
[00:38:24] anything is at the point of purchase ask
[00:38:26] someone how they heard of you. Ask
[00:38:28] people. You have to talk to people. And
[00:38:30] the thing about AI is now you can just
[00:38:32] say to somebody, hey, how did you hear
[00:38:33] about us? Right? and just give them a
[00:38:34] free form text box or, you know, give
[00:38:37] them space on the phone to talk and then
[00:38:40] take those responses, use AI to process
[00:38:42] them, to classify them and organize them
[00:38:44] and turn them into useful marketing
[00:38:45] insights because if somebody says, “Oh,
[00:38:47] I had Claude go do some research or I
[00:38:50] used Open Viking or Deerflow to do some
[00:38:53] research.” Like, okay, great. That’s
[00:38:54] good to know. Now when you get to fully
[00:38:56] agentic buying then you start looking at
[00:38:58] things like you know browser agents and
[00:39:00] stuff like what piece of software
[00:39:02] executed this purchase on our behalf.
[00:39:04] Cardinal sin that most companies make is
[00:39:06] that they do not ask and even if they do
[00:39:09] they’re in many cases they don’t
[00:39:11] incentivize asking in some fashion or
[00:39:13] they don’t make it easy for someone to
[00:39:15] to tell them they you know or they worse
[00:39:18] they have that legacy checkbox you know
[00:39:20] drop down like how did you hear about us
[00:39:21] and there’s five choices that don’t
[00:39:22] reflect reality anymore. Um Chris there
[00:39:24] is a um telling stat 60% of teams using
[00:39:27] AI report two to 3x returns but only
[00:39:31] when measurement is mature. What does
[00:39:33] measurement maturity actually mean?
[00:39:36] Because I think a lot of companies think
[00:39:38] they are measuring but they are really
[00:39:40] just counting vanity metrics.
[00:39:42] >> So be careful with that word. Vanity
[00:39:44] metrics if you don’t know anything about
[00:39:47] uh statistics vanity metrics could be
[00:39:49] misleading. Very often they are the
[00:39:50] start of a process but not the end of
[00:39:52] the process. And people conf people
[00:39:54] people get confused and they say okay
[00:39:55] clearly this is the end of the process
[00:39:56] like number of followers on social media
[00:39:59] they say that’s a vanity metric. Well
[00:40:00] that’s the start of the process because
[00:40:01] the if that number is zero you’re doing
[00:40:03] something wrong right so clearly it’s
[00:40:05] not without value it’s just not your end
[00:40:07] goal. So part of what you have to do
[00:40:09] with measurement is figure out what is
[00:40:10] the end goal measurement that you
[00:40:12] actually care about right and then use
[00:40:14] the AI tool of your choice to say build
[00:40:16] me multivariate regression analysis of
[00:40:19] what higher level metrics have the
[00:40:21] strongest causal inference for the
[00:40:23] outcome we care about and use those
[00:40:25] exact words with your AI coding tool of
[00:40:27] choice and it will build it for you. Uh
[00:40:28] it may ask you some questions but it
[00:40:29] will build it for you. There are very
[00:40:31] few metrics that are absolutely
[00:40:32] worthless. It’s just the context that we
[00:40:34] use them in. Now to the topic of AI
[00:40:38] measurement. AI is how is part of how
[00:40:40] you do certain things, not the outcome.
[00:40:42] The goal is not to use AI. The goal is
[00:40:43] to do more stuff better, faster, right?
[00:40:45] Bigger, better, faster, cheaper. So if
[00:40:47] you’re not meas if you don’t have some
[00:40:48] form of measurement for what bigger is,
[00:40:50] what better is, what faster is, what
[00:40:52] cheaper is, you can’t measure AI. And
[00:40:54] there’s a good chance you’re not
[00:40:55] measuring anything right now either
[00:40:56] because a lot of people like to say,
[00:40:57] what’s the ROI of AI? Well, if you’re
[00:40:59] not measuring ROI now, you can’t measure
[00:41:01] the ROI of AI. So how do you measure
[00:41:03] bigger? How do you measure better?
[00:41:04] Right? You have NPS scores. If you
[00:41:06] don’t, then you probably have a problem.
[00:41:08] What is faster? You have time sheets.
[00:41:10] What is cheaper? Well, you have wages
[00:41:12] and software costs. You should see a
[00:41:14] delta in whatever bigger, better,
[00:41:16] faster, cheaper means to you when you
[00:41:18] apply AI to it. And you can do this with
[00:41:20] something called propensity score
[00:41:21] modeling or uplift modeling that we get
[00:41:23] from bioinformatics. AI is a treatment
[00:41:26] and you have a control group and you
[00:41:27] have a treatment group and you measure
[00:41:29] what is the difference between the
[00:41:30] control group and the treatment group.
[00:41:31] when you apply the treatment of AI to
[00:41:33] any given process and if you don’t know
[00:41:35] how to do that ask your AI for help.
[00:41:37] >> Yeah. Um I think people will still
[00:41:40] hesitate to even ask these basic
[00:41:41] questions you know from their chat bots
[00:41:44] and they would rather
[00:41:45] >> they will because people don’t want to
[00:41:46] know the answers right if you ask the
[00:41:48] question like what are we doing this
[00:41:49] bigger better faster cheaper and it said
[00:41:51] and you know you do the analysis and go
[00:41:52] wow we suck and nobody these days in a
[00:41:55] risk averse zero mistakes tolerated
[00:41:57] environment wants to do that analysis
[00:41:59] because it will the honest truth is you
[00:42:02] suck. So you know uh Christopher most
[00:42:04] SAS companies um software companies are
[00:42:06] using AI for top of the funnel content
[00:42:09] blog posts socials ads but you know you
[00:42:12] we are seeing major um stronger results
[00:42:14] when AI is used in customer life cycle
[00:42:16] marketing. What does that actually look
[00:42:18] like and give us a concrete example of
[00:42:20] AI in the middle or bottom of the funnel
[00:42:23] that is probably working. One of the
[00:42:25] things that we do is we build ideal
[00:42:26] customer profiles that people can use
[00:42:28] with AI to do stuff like synthetic
[00:42:30] synthetic customers so that you can
[00:42:32] train people like sales folks on how a
[00:42:35] customer is likely to react. So anam a
[00:42:37] simple example is we’ll create an ideal
[00:42:38] customer profile. will then create
[00:42:40] role-play personas from it. And we’ll
[00:42:42] give one persona attributes of an ideal
[00:42:44] customer, but we’ll say this customer
[00:42:46] persona is actively hostile, right? They
[00:42:48] don’t like you. You have to figure out
[00:42:50] how to persuade them. And they will keep
[00:42:52] coming back over and over again telling
[00:42:54] you exactly what they think of you. And
[00:42:55] you as a sales professional have to
[00:42:57] figure out how to navigate that. You
[00:42:59] have to figure out how to navigate all
[00:43:00] of the objections. You have to figure
[00:43:02] out how to navigate all of the pain
[00:43:03] points until you can persuade this
[00:43:06] system which has been programmed to be
[00:43:08] hostile to you to grudgingly accept your
[00:43:10] help. Right? So you can use and you
[00:43:12] should be using these tools for things
[00:43:14] like sales enablement for analyzing the
[00:43:17] the number one question that nobody
[00:43:19] asks. What could go wrong? Right? So if
[00:43:23] in the buying process at every stage of
[00:43:25] your salesunnel you should be able to
[00:43:27] answer the question what could go wrong
[00:43:29] here? Why do people drop out of this
[00:43:31] stage of the funnel and then use the
[00:43:32] machines to help you understand it
[00:43:34] better and build solutions to fix it?
[00:43:37] >> You know, um I read it that you u make
[00:43:39] your Substack posts um get run through
[00:43:42] four different personas before you make
[00:43:44] them live.
[00:43:44] >> Exactly the same process. So I have an
[00:43:46] ideal customer profile and I have
[00:43:48] personas for it. I have, you know, one
[00:43:50] who’s a supporter, one who’s a critic,
[00:43:52] one who’s a conspiracy nut that’s kind
[00:43:54] of just like pretty far out there, uh,
[00:43:56] and one who’s just a general audience
[00:43:58] member for that particular persona. And
[00:44:00] each one of them looks at a newsletter
[00:44:02] issue and says, “What do they like about
[00:44:04] this? What do they not like about this?
[00:44:06] What’s missing from this that would be
[00:44:07] useful to them? And what do they think
[00:44:09] is unnecessary?” And based on that
[00:44:11] feedback, I will make adaptations to the
[00:44:13] content to say like, “Yeah, this I
[00:44:15] forgot about this part.”
[00:44:16] >> Absolutely. So um another interesting
[00:44:18] take performance max which is a Google’s
[00:44:20] AI powered advertising product now
[00:44:22] accounts for over 60% of Google ad spend
[00:44:24] for marketers who spent years learning
[00:44:27] how to optimize campaigns manually
[00:44:29] that’s basically Google telling them
[00:44:31] let’s the AI drive it is that a good
[00:44:33] thing or are we just handling the
[00:44:34] control to blackbox
[00:44:36] >> well this goes back to what we were
[00:44:37] talking about earlier which is if it’s a
[00:44:39] template a machine’s going to take it so
[00:44:41] you are writing what is what is Google
[00:44:43] ads Google ads is an entirely templated
[00:44:45] system right you have your headline, you
[00:44:47] have your description, you have your
[00:44:48] call to action, you have your image. And
[00:44:49] so many people have been automating it
[00:44:51] for years anyway outside of the system.
[00:44:54] Now, where systems like PMAX and stuff
[00:44:56] run into trouble is that because they
[00:44:58] are machine learning systems, they need
[00:45:00] time to calibrate. They need time to
[00:45:01] calibrate. They need money to calibrate.
[00:45:03] So, if you do not have a lot of money to
[00:45:06] spend it, the system may never
[00:45:08] calibrate, right? It may continue
[00:45:09] showing ads to wildly incorrect
[00:45:11] audiences uh and burn your budget
[00:45:14] because you didn’t provide enough budget
[00:45:16] for it to get valuable results from a
[00:45:19] machine learning perspective. And the
[00:45:21] other thing that we have to all remember
[00:45:23] is that these systems are very optimized
[00:45:26] and very tuned to make the ad companies
[00:45:29] money. That they make us some money
[00:45:30] along the way is incidental, but their
[00:45:33] primary function is to make the ad tech
[00:45:35] company make money. Um so they will
[00:45:37] sometimes make recommendations and
[00:45:39] decisions. This is you know called
[00:45:41] multi-objective optimization. They will
[00:45:43] sacrifice some efficiency for us to make
[00:45:47] their corporate masters money which is
[00:45:49] one of the reasons why it’s very
[00:45:50] important that we you not give up those
[00:45:52] skills and that you be able to work with
[00:45:55] outside traditional digital ad systems.
[00:45:58] There is a lot of value still in
[00:46:00] something as simple as like a pizza box
[00:46:02] flyer in in a certain geography. If you
[00:46:05] know a geography and you know your
[00:46:07] customers really well to the point where
[00:46:09] you could you know uh exactly where they
[00:46:11] get their pizza from, you can
[00:46:13] potentially influence them in ways a
[00:46:15] digital system never would. I’ll give
[00:46:16] you a real simple example. There are
[00:46:18] four pizza chains that serve the
[00:46:20] Pentagon, right? The US Department of uh
[00:46:23] defense uh main facility in Alexandria,
[00:46:26] Virginia. Four companies when it’s a
[00:46:27] late night, they’re running pizzas over
[00:46:29] to the Pentagon as fast as possible. If
[00:46:30] you watch influence decision makers at
[00:46:32] the Pentagon, you would be better off
[00:46:34] buying ad space on those pizza boxes
[00:46:36] because those pizza boxes are going into
[00:46:38] the most secure locations on the planet,
[00:46:41] right, with your little paper flyer on
[00:46:43] them that no ad system is going to be
[00:46:45] able to get into, right? Because so much
[00:46:46] stuff is locked down. You got to think
[00:46:48] outside the box, like pay attention to
[00:46:50] what people do. Pay attention to where
[00:46:52] people get their stuff and then work
[00:46:54] within in in those boundaries instead.
[00:46:57] Yeah, Pmax is great, but if particularly
[00:46:59] if you’re not a huge business with who’s
[00:47:01] willing to spend 10 grand a day uh with
[00:47:03] Google,
[00:47:04] >> Chris, if I land on um say a software or
[00:47:07] a manufacturing’s homepage and I don’t
[00:47:08] understand in 5 seconds about what do
[00:47:11] they do? Whose fault is that? The
[00:47:13] marketer who wrote it or the product
[00:47:15] team who couldn’t articulate it?
[00:47:17] >> Yes, it’s everyone’s fault, right? Um
[00:47:20] and again, this is where things like
[00:47:22] ideal customer profiles and AI can help
[00:47:24] you, right? You can load it up with
[00:47:26] things like the web content
[00:47:27] accessibility guidelines and say, “Audit
[00:47:29] this page. How how good is the basic
[00:47:32] design. Does it communicate basic
[00:47:34] concepts? Does it use color and contrast
[00:47:35] correctly?” You can audit the language
[00:47:37] and say, “Is it clear from the
[00:47:39] customer’s perspective?” Because the
[00:47:40] number one mistake marketers make is
[00:47:42] marketers think a that they are the
[00:47:43] customer. They’re not. And b that the
[00:47:46] customer knows what’s in their heads and
[00:47:48] very often they do not. So if you have a
[00:47:50] page that is not converting, you have to
[00:47:52] look at three things. So, this goes back
[00:47:54] to 1968 to Bob Stone’s direct marketing
[00:47:57] uh list which is his thing was list
[00:47:59] offer creative. Do you have the right
[00:48:00] people, right? The wrong people. It
[00:48:02] doesn’t matter how good your design is.
[00:48:03] If you’ve got the wrong people visiting
[00:48:05] who are never going to buy from you,
[00:48:07] your marketing efforts are wasted. So,
[00:48:08] do you have the right people? Do you
[00:48:09] have the right offer? You may have the
[00:48:11] right people, but they do not want to
[00:48:12] buy what you are selling. It doesn’t
[00:48:14] matter. Yeah. If I have if I am uh
[00:48:17] selling swimming pools, right, and I’m
[00:48:19] attracting people from Gnome, Alaska,
[00:48:20] where you cannot put a pool because it’s
[00:48:22] frozen most of the year, right? I may
[00:48:24] have the right people. I may have the
[00:48:25] right income, the right customer base,
[00:48:27] you know, the right demographics, but
[00:48:28] they ain’t going to buy a pool ever. And
[00:48:30] then finally, you get to the creative
[00:48:32] design, messaging, and all that stuff.
[00:48:34] Most marketers, they default right to
[00:48:37] like design and creative and wording
[00:48:38] without ever considering, do we even
[00:48:40] have the right people that we’re
[00:48:41] attracting to begin with? Moving towards
[00:48:43] the um end of our conversation with
[00:48:45] these last two questions first. If
[00:48:46] someone’s just started with AI and
[00:48:48] marketing today, what should they learn
[00:48:50] first?
[00:48:50] >> If you’re just getting started, learn
[00:48:52] how to prompt intelligently, learn how
[00:48:54] to write project plans because that’s
[00:48:57] what modern prompting is. It is not a
[00:48:59] conversation. It is project planning.
[00:49:01] >> All right. So um Chris, if a CTO or IT
[00:49:03] director is listening right now, someone
[00:49:05] who’s been ignoring social media,
[00:49:07] skeptical of AI hive and frankly
[00:49:09] exhausted by all the noise, what’s uh
[00:49:12] one advice you would give them and one
[00:49:14] actionable thing that they must do to
[00:49:16] stay relevant?
[00:49:17] >> Uh you should be learning about on
[00:49:18] premise deployment of AI models because
[00:49:21] that’s going to directly affect your
[00:49:22] world. Uh, and it could potentially save
[00:49:24] your money a your company a whole bunch
[00:49:27] of money because at some point every
[00:49:30] major cloud-based AI company in the west
[00:49:32] is running at a substantial loss, right?
[00:49:35] That what they’re charging for their
[00:49:36] services is not sustainable. OpenAI, for
[00:49:39] example, is bu burning billions of
[00:49:41] dollars a month. At some point, that
[00:49:43] money runs out and then they prices go
[00:49:44] up by like adding a zero. If you are
[00:49:47] dependent on a cloud vendor, you’re
[00:49:49] basically held hostage to them. If you
[00:49:51] are considering looking at buying
[00:49:53] hardware like an Asus GB10 or uh an
[00:49:56] Nvidia GGX Spark or a Huawei Ascend 950
[00:50:00] uh cluster, then when that when that
[00:50:02] party stops, right, and you know the the
[00:50:05] when the music stops and everyone’s
[00:50:06] going for a chair, you’ve built your own
[00:50:08] chair, right? You have hardware on
[00:50:10] premise that can continue to serve up
[00:50:12] today’s
[00:50:14] local models. The open weights models
[00:50:16] are 95% of the capability of the
[00:50:19] state-of-the-art models at 0% of the
[00:50:21] cost as long as you have the
[00:50:23] infrastructure. So if you’re an IT
[00:50:24] director, you’re a CIO, you’re a CTO,
[00:50:27] you should have an on-remise solution
[00:50:28] ready to go so that if things change or
[00:50:32] you know the the chief AI officer comes
[00:50:34] in and says, “Hey crap, we you know our
[00:50:36] AI budget just went from six digits to
[00:50:38] seven digits.” You can say, “I got you.
[00:50:39] We’re going to switch over to using this
[00:50:41] local router and we’re going to have our
[00:50:43] employees use these local models first
[00:50:45] and then your price of AI goes down to
[00:50:47] the cost of electricity to run those
[00:50:49] boxes. And that’s you are going to be
[00:50:51] the the biggest hero in your company.
[00:50:54] >> All right. So, uh Chris, uh before we
[00:50:56] wrap up, I have this quick rapid fire
[00:50:58] for you. One AI tool you can’t live
[00:51:00] without right now.
[00:51:01] >> Claude Code.
[00:51:02] >> One AI trend you think is overhyped.
[00:51:04] That’s tough because a lot of them are
[00:51:07] the hype is stupid, but the underlying
[00:51:09] technology is good. I would say
[00:51:11] overhyped right now probably is fully
[00:51:14] autonomous agents. They’re not ready for
[00:51:16] production yet.
[00:51:17] >> Um, one social platform you wish you
[00:51:20] would just end already.
[00:51:22] >> X
[00:51:22] >> one LinkedIn post format that needs to
[00:51:25] be retired immediately.
[00:51:27] >> All of them. Again, if you’re if it’s a
[00:51:29] template today, a machine’s going to do
[00:51:30] it tomorrow. We all know the post. Cashy
[00:51:32] headline. It’s silly question, bullet
[00:51:35] list, you know, uh pointless, you know,
[00:51:38] hook question at the end, emoji. We’ve
[00:51:40] we know them all. We’ve seen them all.
[00:51:42] They don’t offer they they tend not to
[00:51:43] offer a lot of value.
[00:51:44] >> Five years from now, what’s the one
[00:51:46] thing about marketing that AI will never
[00:51:48] be able to do? I that’s a good question
[00:51:51] because I don’t know the the the models
[00:51:55] are increasing their capabilities
[00:51:56] exponentially which is very difficult
[00:51:59] for human beings to wrap their brains
[00:52:00] around and we are getting to a point
[00:52:02] where buying agents are a real thing. So
[00:52:06] I don’t know other than like basics of
[00:52:08] marketing like does your is your product
[00:52:10] worth buying?
[00:52:11] >> Well Chris thank you so much for for
[00:52:13] this grounded and insightful
[00:52:14] conversation and to everyone listening
[00:52:16] thanks for joining us on Tech Unhinged.
[00:52:18] >> Thank you.
