YouTube transcript
Video ID: RbIK694S7_I
Language: en
Captions: auto-generated

[0:00] Angela, good to see you. Another week
[0:02] has gone by and we had a few short share
[0:04] forms to catch up on and we were talking
[0:06] about how you had some really good
[0:08] questions back in the share forum I
[0:10] think a month ago month or two ago and I
[0:12] would love to hear what the top
[0:13] questions were and we can discuss
[0:14] through them recapping what people
[0:16] thought and our own insights on it.
[0:18] >> Sounds good. We are going to go back to
[0:21] the May share forum and right now we're
[0:23] in July. So we're going to hit some
[0:25] technical accounting issues that come
[0:27] out of the weird AI era and then we'll
[0:29] maybe bridge to some costs. So in May we
[0:32] talked a lot about some of the new
[0:35] issues coming out of the fact that
[0:37] everybody's trying to figure out how to
[0:38] sell AI right now, right? So if all the
[0:41] product and strategy teams are coming up
[0:43] with new selling models, that means
[0:45] companies are releasing features that
[0:47] may or may not be priced separately all
[0:49] the time. And so one of the questions
[0:51] that came up is we have all these
[0:53] existing customers and of all these
[0:55] existing customers, we decided that they
[0:57] all got access to this new AI feature.
[1:00] And this new AI feature, we're going to
[1:02] also sell maybe consumption based to
[1:06] our to our self-serve or product led
[1:08] customers. So it's going to be a
[1:10] separately priced product for those
[1:12] customers. But for our existing
[1:14] enterprise customers, we're just going
[1:15] to give it to them for free so that they
[1:17] can try it out and see if it sticks.
[1:19] Totally makes sense, right? Because
[1:20] nobody really knows how to sell this
[1:23] yet. Some people are selling by agent,
[1:25] which I think is interesting given me
[1:27] and you have both built and we're like,
[1:29] that's so weird. Aren't Doesn't the
[1:31] model decide how many agents there are?
[1:33] >> Yeah.
[1:35] >> And some people
[1:36] are just trying to do it consumption
[1:38] based or some people are trying to fit
[1:39] it into their straight line normal SaaS
[1:42] plain vanilla selling. So the question
[1:44] kind of comes up in a couple different
[1:46] ways, but I'll stop there. Are you
[1:48] seeing the same thing where people are
[1:50] just like throwing stuff against the
[1:51] wall? They're treating their enterprise
[1:53] and their self-serve customers a little
[1:56] different and that means that there's a
[1:58] little bit of
[1:59] accounting chaos under the rumble,
[2:01] right? Because then people are trying to
[2:02] figure out what does that look like in
[2:04] accounting? What's the performance
[2:05] obligation? Did it change the
[2:06] performance obligation? Is it a
[2:08] modification when you give that new
[2:09] thing? I think there's all these fun
[2:11] little things that that pop up because
[2:13] of that question.
[2:14] >> Yeah, I'm saying the same thing, no
[2:15] matter what you say. I think we saw a
[2:17] lot of the revenue leaders that we
[2:18] talked to in your latest event and some
[2:20] of the other, you know, in the share
[2:22] form as well are facing this as well.
[2:24] It's like It reminds me, and I think
[2:26] that's what we talked about previously,
[2:27] of like I just the speed of product is
[2:31] as product life cycles are just evolving
[2:33] so fast or product to launch and
[2:35] probably life cycles, too, the decay of
[2:36] products as well, so fast where we used
[2:38] to be in a world where it might might
[2:40] have taken a year to launch a CD or
[2:42] on-premise software or develop updates.
[2:44] And that it seems like ages ago when I
[2:46] think I mentioned it and someone was
[2:47] like, "Well, how old are you?" when I
[2:48] talked about that at at the conference,
[2:50] but that was like 10 years ago, like
[2:52] 10-15 years ago, wasn't that long ago.
[2:54] And then we moved into the SaaS world
[2:57] where you'd launch continuous updates,
[2:58] but still like the updates would kind of
[2:59] be batched or if it was an update it
[3:01] happened once every couple of months,
[3:03] and maybe minor updates more frequently.
[3:05] To now, we're seeing like the big AI
[3:09] companies change models almost weekly or
[3:11] to make some kind of refinement or
[3:13] updates weekly. And then you get talked
[3:14] to Anthropic about that, you know, like
[3:16] them out. And everyone's mirroring.
[3:18] Where almost every company I'm talking
[3:20] to, the product team's like, "We can
[3:21] launch something this week." And it
[3:23] changes the problem like to your to your
[3:25] point.
[3:26] Accounting or accounting function has
[3:28] never had to deal with this before.
[3:30] Normally, you have time to assess a
[3:31] product as a combined with your current
[3:33] like we assess it as a performance
[3:34] obligation, the what performance
[3:36] obligations are embedded. What promises
[3:38] are they? Are they combined or not?
[3:39] Under six, but now we're in this world
[3:41] where, well, if you have this
[3:43] ever-changing road map of features
[3:45] changing within the products, you're
[3:47] selling them differently, how do you
[3:48] actually account for them? What are your
[3:50] performance obligations? And I think
[3:51] it's tripping a lot of people up. So,
[3:53] it's pretty interesting. Did I Did I get
[3:55] that right? Am I missing anything there?
[3:57] >> Yeah, I think that's exactly right. I
[3:58] think that anyone that was selling like
[4:01] a fixed amount of stuff for a year,
[4:04] every software company or SaaS company,
[4:06] or whatever you want to call it, but an
[4:07] AI company, AI native everything, right?
[4:10] >> Yeah. Yeah.
[4:11] >> That they're now selling everything on
[4:13] the menu, or they're being expected to
[4:16] sell everything on the menu, or to
[4:18] continuously turn out features at a rate
[4:21] that is much faster because if you can't
[4:23] keep up with the changes, and the
[4:25] changes are quite literally, you know,
[4:27] weekly, daily, then what does that mean
[4:29] with your accounting conclusion that you
[4:30] made at the beginning that you said I
[4:32] had, you know, three performance
[4:34] obligations. Let's just do kind of a
[4:35] standard if you're a SaaS company
[4:37] transitioning over to a SaaS plus AI
[4:40] company, right? So, maybe you had one
[4:43] platform fee that was ratable straight
[4:44] line. You had maybe some consumption
[4:47] component, but probably pretty new like
[4:49] X amount of storage, however many
[4:52] summaries you did, which is a very
[4:54] potentially AI token-intensive process,
[4:57] right? If you're pulling in tons of
[4:58] context and spitting something out. Um
[5:00] and then another piece that was
[5:02] professional services. Those are kind of
[5:03] like the standard three things that you
[5:05] may or may not sell. Now, all of a
[5:07] sudden, that middle one, that that
[5:09] ratable platform piece is like, "Oh, no.
[5:12] Now, instead of just selling the storage
[5:13] bucket, we're going to sell like 15 uses
[5:15] of our little agent, right?" And maybe
[5:17] that's every month. And it's like, "No,
[5:18] no, no, not just that agent. We have the
[5:20] summarization agent, but then that
[5:22] thing's going to code you a mini app,
[5:23] which is not that crazy right now." And
[5:25] that mini app is going to be housed on
[5:26] this weird little extra part of our
[5:28] platform, and we're going to maintain
[5:30] this thing and keep it, I'm going to say
[5:32] socks compliant, but we all know that
[5:34] hasn't been figured out yet, right? So,
[5:35] like, okay, now you started with three
[5:37] things. You now have five things. Is
[5:38] that a contract modification? What is
[5:40] that, right?
[5:41] >> And especially if you put on your
[5:42] contract, "Hey, by the way, we're going
[5:44] to launch new things every single week."
[5:46] Like, if you have a road map and your
[5:48] road map's constantly changing. It's not
[5:50] feasible at some point to continue to
[5:52] reassess all of your performance
[5:54] obligations, but you can also get very
[5:55] funky, right? Not to jump the gun there,
[5:57] but it It's like you can't just assess,
[5:59] well, the individual launch anymore. You
[6:01] have to assess, well, contractually, are
[6:03] you always giving all of like who are
[6:06] you giving all of these new offerings
[6:07] to? Are you giving them to your existing
[6:09] customers? Because if they are, is that
[6:11] more of a stand-ready obligation where
[6:13] you have a platform and you're just
[6:15] changing everything, changing adding
[6:16] things to it and subtracting things and
[6:17] making it new, and is it bigger than the
[6:19] sum of its parts? Or are you really
[6:21] going to the business of like Amazon or
[6:23] shoes where it's like, "Hey, we're
[6:25] actually selling a bunch of different
[6:26] types of shoes, and each thing you buy
[6:27] is a different pair of shoes." I mean,
[6:30] it's debatable. I don't think anyone
[6:31] really knows where we're going because
[6:32] we're figuring this this out as we go,
[6:35] but I think there's some things that
[6:36] make sense under the like accounting
[6:38] standards. Or and then go away from the
[6:40] even FASB accounting standards of like
[6:42] the basic the basic topics, like cut
[6:45] off, accuracy, completeness, timeliness.
[6:47] I think we got to got to go back to
[6:48] fundamentals in a lot of ways and make
[6:50] sure we're purporting revenue correctly
[6:53] in spirit and consistently, but
[6:54] >> Yeah. Um you bring up a great point, and
[6:57] it was that legal piece. And I'm going
[7:00] to translate that to legal
[7:01] enforceability, which is part of our
[7:03] step one of ASC 606, which is basically
[7:06] do you even have a contract, right? And
[7:09] so
[7:10] we know like I I think it's fair to say
[7:12] that every company is trying to release
[7:14] new things all the time to keep up with
[7:17] what is expected of them in this AI
[7:19] world. Then there is a split between who
[7:21] has actually contractually changed that
[7:24] and who is just trying to release,
[7:26] right? Now, if you have that split of
[7:28] who has contractually controlled this
[7:31] thing, there's a couple different ways
[7:32] to contractually control. One of them is
[7:34] to say what you said, which is you're
[7:36] always going to get what we have, which
[7:38] sounds a lot like we have an open menu
[7:40] and you have access to our open menu,
[7:42] which feels a lot like straight line.
[7:45] >> Yep.
[7:45] >> And then there's the people that what
[7:46] they do in practice is still that, but
[7:49] they haven't contractually caught up
[7:50] with it yet.
[7:51] >> Yep. Yep.
[7:52] >> So, if they haven't contractually caught
[7:53] up with it yet, and then you do another
[7:55] split, it's like do they get to keep
[7:57] that feature forever, or have you set an
[7:59] expectation that they get to keep that
[8:01] feature forever? That is super super
[8:03] gray, but if that you do, then you might
[8:05] actually might be in a contract
[8:07] modification.
[8:08] >> Mhm. Mhm.
[8:10] >> And we have so many trees to split, but
[8:12] I'm just going to pick that one. Now,
[8:13] let's say that everybody is expected to
[8:16] be able to keep that feature, which I
[8:17] think is a little crazy, because I don't
[8:19] expect to keep any crazy AI feature that
[8:20] I get popping up in any, you know,
[8:22] software that I use right now, cuz I'm
[8:23] like, "Oh, that's cute." And I'm like,
[8:24] "Is it going to stick?" I don't know.
[8:25] Then I play with it a little bit, maybe.
[8:27] But, if it's a contract modification,
[8:29] accounting world goes in chaos, right?
[8:30] For those of people that don't know it,
[8:32] you have to end a contract and begin a
[8:34] new one, typically. And then if you do
[8:35] that, anything where real life was off
[8:38] from accounting life, meaning if you
[8:39] have like deferred revenue, or you have
[8:41] a contract asset where revenue is ahead,
[8:44] you have to take those weird differences
[8:46] and then roll them into the next
[8:47] contract, which is not the same as what
[8:49] you're billing for. And you have to
[8:50] track this weird little difference
[8:52] forever until the next modification
[8:54] happens. So, I'm going to stop there,
[8:56] cuz I think that that's like the crazy
[8:57] thing that happens in accounting that
[8:59] might be trackable painfully on a
[9:02] spreadsheet for many existing companies,
[9:04] but really painful if you're doing it
[9:06] for all of your customers suddenly,
[9:07] because you pushed out a new AI feature
[9:09] and didn't paperwork it correctly, or
[9:12] paperwork it safely for your accounting
[9:14] operations.
[9:14] >> Yeah, yeah. And I think even more, we're
[9:16] saying accountants and finance leaders
[9:18] need to take go into the front end of
[9:21] the business and work on product
[9:23] launches and I think a lot more or I
[9:25] think we were talking about it when
[9:26] you're in the last year form about how
[9:28] we're seeing a lot of like a lot of
[9:31] revenue and go-to-market roles with
[9:34] accountants in them become DP plus
[9:36] roles, because we need these leaders in
[9:39] this space to make sure that product
[9:42] launches can happen compliantly and
[9:44] follow best practices and it tends to be
[9:47] the CPA in the room or the accountant in
[9:49] the room is the person who can bridge
[9:51] all of those gaps. Like the legal gap
[9:53] like hey that there's a legal issue here
[9:54] we need to document the contract,
[9:56] there's a finance issue, there's an
[9:57] accounting issue and there's a product
[10:00] launch issue and an economic issue. We
[10:02] need it turns out the CPA in the room is
[10:04] the one who can bridge those uniquely
[10:06] which is which is a cool thing about our
[10:07] profession but a problem for businesses.
[10:09] Like you said it's like there isn't I
[10:11] think there are some best practices and
[10:12] maybe we can talk about those here but
[10:15] they aren't widely distributed and most
[10:16] companies don't know about them. And
[10:18] like you said every there's disparate
[10:19] practices through disparate ways or
[10:21] people are approaching these businesses
[10:22] and there needs to be a more uniform
[10:24] there doesn't need to be but it it if
[10:25] they should take a more uniform approach
[10:28] to this in order to make their lives a
[10:29] little easier. Because like you said in
[10:31] a real world if the the goal is we're
[10:33] just going to start launching a bunch of
[10:34] products all the time that are going to
[10:35] be add-on to our existing product. Well
[10:37] yeah we're probably more of an in a
[10:39] stand ready kind of role then that's a
[10:40] little easier to account for. Maybe
[10:41] that's going to change your earnings and
[10:43] your revenue timing and whatnot. But
[10:45] like better to do that consciously and
[10:48] to understand it and then figure out how
[10:49] the pricing plays with that. Because the
[10:51] price pricing on a consumption based but
[10:53] your product's more of a stand ready
[10:55] type that's going to cause some
[10:56] heartburn with your auditors and you're
[10:58] going to need to get your story straight
[11:00] pretty well. If the opposite if your
[11:02] strategy is I'm really going to nickel
[11:03] and dime and try to launch a new product
[11:05] and sell thousands of products and
[11:07] become a new Amazon or Nike of the world
[11:09] but in SaaS land then then your
[11:11] contracts and everything else should
[11:13] follow.
[11:14] >> Yeah you you said something about you
[11:16] know the dream that everyone had before
[11:18] was to get straight line revenue which
[11:20] is interesting right? But that is a
[11:22] disconnect from the reality of I think
[11:25] how people are thinking about pricing
[11:27] now which is consumption based which
[11:28] isn't just because it's like fun and
[11:30] like it's like the customer gets the
[11:32] value that they use, which is true, but
[11:34] it's also because of the cost side. So,
[11:36] before I go deep into the cost side,
[11:38] which I so badly want to,
[11:39] >> Yeah, I yeah. I see it. I see it in you.
[11:41] Yeah.
[11:43] >> Let's quickly close out that that
[11:44] straight line piece. The straight line
[11:46] piece is you wanted it cuz you valued
[11:49] well a certain way, right? You had
[11:50] predictability, less volatility. You
[11:53] were able to say we were going to do X
[11:55] amount more next year because you had
[11:57] some renewal uplift that you can
[11:59] relatively control and forecast for. I
[12:02] think there is a debate out there, and
[12:04] I'm not actually sure but I'm curious to
[12:05] get your thoughts. Like, what is the
[12:07] investor world thinking, right? It's
[12:09] like, what's better now? Is straight
[12:11] line better? Is consumption better? Do
[12:13] we not really know? Are you just trying
[12:15] to get close enough because if you
[12:16] accidentally get straight line but
[12:17] you're actually consumption and you
[12:18] flip, that's kind of dangerous, too,
[12:20] right?
[12:21] I'm not actually sure if straight line
[12:23] is as cool as it used to be.
[12:24] >> Yeah, and I agree I I think there's a
[12:26] big challenge with AI where the many of
[12:29] the investor community is forecasting
[12:32] based on consumption-based basis. If you
[12:34] mention AI or tokens or anything
[12:37] anything usage-based pricing, a lot of
[12:39] them it's lost on them if you're
[12:42] recognizing items ratably, but you might
[12:44] be metering things on a usage-based way,
[12:46] which is an okay convention or okay path
[12:49] in a in accounting CPA land.
[12:51] But I've seen companies that disclose
[12:52] we're recognizing revenue ratably, and
[12:54] then investors talk about
[12:55] consumption-based revenue recognition.
[12:57] And so, there is this disconnect on do
[12:59] they actually understand how the numbers
[13:02] are being reported? And there's two
[13:03] benefits, right? Like, so if I'm a CFO
[13:05] and I'm trying to play with the numbers,
[13:06] which I don't recommend to anybody, but
[13:08] if Yeah, don't do it. But, if you wanted
[13:11] to, there's two way Like, hey, if you
[13:12] pull consumption-based, well, that tells
[13:14] you a better earning story in the long
[13:16] run if you have back-ended growth
[13:17] because your consumption-based revenue
[13:19] will be exponential on like, let's say
[13:21] you have a growing customer base.
[13:23] Because if you start low and your
[13:24] products ramp over time, you'll have
[13:26] more revenue down the line. So, it tells
[13:28] a better growth percentage growth story.
[13:30] Rattable is easier to operationalize and
[13:33] is a cleaner story. It's easier to
[13:34] predict. Your forecasts are going to be
[13:36] a lot cleaner. Your accounting
[13:37] operations are going to be a lot
[13:38] cleaner. It's just it's more comparable.
[13:40] There's some real benefits to how easy
[13:43] and accurate your forecasts are going to
[13:44] be.
[13:45] But, you're going to pull ahead some of
[13:47] the revenue and you're going to have
[13:48] explanations on like, well, where's all
[13:50] this deferred revenue and how you're
[13:51] going to perform on it. So, there's
[13:52] there's kind of pros and cons to each
[13:55] and I've seen CFOs and CAOs who like are
[13:58] really adamant about one versus the
[14:00] other. And it's strange as an accountant
[14:02] in this world because generally we like
[14:05] to like want to apply things
[14:06] consistently. Like, what what are the
[14:08] facts that make fit into rattable versus
[14:11] consumption base? And And the nuances
[14:13] here can really trigger a swing one way
[14:14] or the other. I have a strong opinion
[14:16] that I think the world would be a lot
[14:18] easier if we all just said rattable and
[14:20] did non-GAAP for consumption just so we
[14:22] stop this this conversation and stop
[14:25] people playing games and figuring this
[14:26] all out, but and we get back to like
[14:28] consistency and cost-benefit theory for
[14:31] like accounting, get back to like, okay,
[14:33] let's focus on the real business. But,
[14:35] while we're in this world, it's fun to
[14:36] play around with it. So.
[14:37] >> It's a it's a very
[14:39] hotly debated take. And And I don't I
[14:42] honestly have not come up with my take
[14:44] yet of what I think is better. I used to
[14:46] always think rattable was better cuz
[14:48] operationally it was easier. But, then
[14:50] if everybody's doing non-GAAP on
[14:51] consumption anyways, you have to
[14:53] operationally do both anyways. Maybe not
[14:55] to the level but you certainly have to
[14:57] understand it. And so, I'm not actually
[14:59] sure if the break from reality is
[15:01] amazing. My hope and dream, if I was to
[15:04] have one as an accountant, would be to
[15:05] bring back this this little old piece of
[15:08] guidance called the contingent revenue
[15:10] cap.
[15:11] >> Yeah, that'd be great.
[15:13] >> I think it was an SEC ruling wasn't
[15:15] actually technically in the guidance
[15:16] when we were under 605, which was the
[15:18] old revenue standard. And it basically
[15:20] had this rule that said, if you can't
[15:21] bill for it, you can't recognize it as
[15:23] revenue.
[15:25] >> Yeah.
[15:25] >> Which in some ways solves a couple
[15:27] issues. It's not all the way to
[15:28] consumption and it's not all the way to
[15:30] ratable. It's just saying, I'm going to
[15:33] bring you slightly closer to the real
[15:34] world, which I think right now in the
[15:36] accounting world we are so disconnected
[15:37] from the real world that is insanely
[15:39] painful cuz it just means you need to
[15:41] two operating full systems of record,
[15:44] which is incredibly expensive and
[15:45] painful. But
[15:46] >> I'm with you. So I would I love that you
[15:48] said that and I would I would pay money
[15:50] for a system to go back to that world.
[15:51] We can petition someone to bring it back
[15:54] to hey, and what we're saying to
[15:55] everyone listening, I'm sure mostly it's
[15:57] revenue accountants is hey, like at the
[15:59] end of the day take revenue as invoiced
[16:01] on on this if you're going to be extreme
[16:03] and at least we all know what we're
[16:04] talking about, right? And maybe maybe
[16:06] have some disclosures around well, how
[16:08] much of this invoiced revenue is
[16:10] actually earned or like how much like
[16:12] remaining obligations do you have
[16:13] related to it? How much is it prepaid,
[16:15] etc. But man, that would make life a lot
[16:17] easier rather than currently, which is
[16:19] obscene. Like imagine like you said this
[16:21] new company comes in, now you have 20
[16:23] new products and now under 606 you have
[16:25] to allocate your transaction price to
[16:27] all these products that you've never
[16:28] sold before and then recognize a portion
[16:32] of your total cut which
[16:33] portion of your transaction price, which
[16:35] could be your minimum commitment or it
[16:36] could be your order form or however you
[16:38] structure it has to be allocated to it
[16:40] and all your systems have to tie to that
[16:42] and you might have contract
[16:43] modifications like you said every time
[16:44] you launch a new product to reallocate
[16:46] everything.
[16:48] >> you will have them.
[16:48] >> anyone Yeah, how and and how There's no
[16:51] way we're doing this consistently.
[16:52] There's no way every like
[16:54] you and I know we can't you could you
[16:56] can't you have at the end of the day you
[16:57] have to be able to make a deterministic
[16:58] logic to the way the formulas run and if
[17:01] Excel breaks when you do it for like a
[17:03] small company, there's no way these
[17:05] major companies are doing this
[17:06] correctly, right? So so I'm with you. I
[17:08] love that you said that and if we can
[17:09] bring I if we should get some shirts
[17:11] bring back contingent revenue cap or
[17:12] something like that.
[17:14] >> It's just a small link to reality,
[17:16] right? It's a small anchor to this piece
[17:18] that we all understand, which is I sent
[17:21] you an invoice and I got an invoice from
[17:22] you. And that thing we hope, because we
[17:25] sent it to each other, is a real
[17:26] artifact, right? It is a small link to
[17:29] reality. Instead of me doing all these
[17:31] random calculations in my operating
[17:33] system for my either ratable or
[17:35] consumption, totally diverse, everybody
[17:37] decides it differently, sadly. And then
[17:39] we slowly map it back to consumption.
[17:42] And then somewhere in there is like
[17:44] deferred revenue with its own definition
[17:46] with billing.
[17:46] >> Yeah, that whole contract asset topic of
[17:49] like where we started recognizing
[17:50] revenue ahead of billings got really got
[17:53] really funky really fast.
[17:55] >> And and conceptually it does make sense
[17:59] theoretically, right? Sometimes I do
[18:01] more work than I've been due. But then
[18:04] you get to like, okay, you do more work
[18:07] than you could bill for. That's like an
[18:08] accrual, accrued revenue, some people
[18:10] call it that. But but we're actually
[18:11] talking about something much more
[18:12] conceptual, which is I
[18:15] think that I'm going to So, you have a
[18:17] consumption-based billed thing for a
[18:19] year. And you say, I think that this
[18:22] consumption-based billed thing is
[18:24] actually more like me keeping the lights
[18:26] on, so I'm going to take it ratable. But
[18:28] I'm also going to have to estimate how
[18:30] much I think I'm going to make over the
[18:32] amount that I put in the contract. So,
[18:33] not only is it this fake ratable, which
[18:36] doesn't It's not fake ratable, but like
[18:38] obviously it's not perfectly ratable.
[18:40] Sometimes it gets close and that's an
[18:41] argument. But if I have to do this
[18:43] ratable plus this fake amount that I
[18:45] think I'm going to get, called variable
[18:46] consideration, there's like two extra
[18:48] disconnects from reality. I haven't even
[18:50] talked about whether it's allocated or
[18:51] not, which you brought up. So, you might
[18:53] have then three disconnects from
[18:54] reality, right? Then you throw in a
[18:55] modification, which I'm going to stop
[18:57] saying the word modification for a
[18:58] second and just say upsell, cuz
[19:00] everybody that's ever listening to this
[19:01] that's in sales, they're like, "Yeah, we
[19:03] Everyone wants the upsell." I.E., the
[19:05] goal of having this contract is to
[19:06] upsell your customer. If the goal of
[19:08] your If this the goal of this is to
[19:10] upsell, you likely will have a lot of
[19:12] them, hopefully, which means that this
[19:13] system is made to do this. And let's say
[19:16] we just said three to four disconnects
[19:18] from reality. Disconnects from reality,
[19:19] which is a computational like just
[19:22] conceptual mathematical calculation for
[19:25] what this thing is. And every step that
[19:27] you remove from reality, you get further
[19:29] away from that number, i.e. further away
[19:30] from the billing, which is the true
[19:32] artifact that left my inbox and went
[19:34] into yours.
[19:34] >> Yeah.
[19:35] >> I I think that what makes it reality is
[19:37] that two people now agree on it, i.e. it
[19:38] went into your inbox.
[19:39] >> Yeah.
[19:40] >> Yeah, exactly.
[19:41] >> And
[19:41] >> the fact that we are getting more and
[19:42] more loops away from that reality, I
[19:44] think you have to take that step back
[19:46] and say like, why? Why are we doing any
[19:48] of this? Is rolling forward that random
[19:49] contract asset that now is four or five
[19:51] steps or six or by the time you do two
[19:53] upsells, what, 16 steps away from
[19:56] reality? Is it worth it?
[19:57] >> Yeah. Yeah, I think it's a good
[19:58] question. When our job as accountants at
[20:01] the end of the day, and I I believe in
[20:02] accountants as an honorable like
[20:04] honorable profession term, I'm not
[20:05] saying like just bean counters, but like
[20:08] looking out for the books, we're like
[20:09] there to to figure out what the truth
[20:10] is. Like that's the whole point is like
[20:12] how can we accurately and materially
[20:14] represent this information in a way that
[20:15] everyone can understand for very complex
[20:17] businesses.
[20:18] So, the further we get away from the
[20:19] truth, we're missing missing a bit of
[20:21] the core substance of our role, right?
[20:23] And so, I'm with you. Bring back
[20:25] contingent revenue cap. We got to print
[20:27] some shirts and get it start selling
[20:29] them at your GAAP savvy events or
[20:30] something because I think
[20:31] I think that's I completely agree. And
[20:34] like to go back to the people's
[20:35] questions of like how do you account for
[20:36] this, I think the answer is you account
[20:38] for it like yes. I've seen I've seen
[20:41] both practices and you kind of choose a
[20:42] lane and your auditor is going to be
[20:43] contingent, but your legal form is going
[20:46] to ultimately change how you account for
[20:48] these and I recommend having a good
[20:50] person in the room to advocate. It
[20:52] almost reminds me of a sell-side and
[20:53] buy-side diligence. Like we originally
[20:55] like back in the day, back in the day,
[20:58] and not too long ago, if I had a
[20:59] business, I'd sell it to somebody or my
[21:01] best employee takes over my business.
[21:03] But then we all started selling
[21:04] businesses to PE firms and VC firms and
[21:07] people started doing shady stuff. So,
[21:08] you have a sell-side and buy-side
[21:09] representative to like, "Okay, I'm going
[21:11] to help you represent your business
[21:12] accurately." And someone's going to be
[21:13] like, "I'm going to make sure someone's
[21:14] not doing shady stuff." And I feel like
[21:16] we've come to this same world in
[21:18] accounting where it's like, "Okay, yeah,
[21:19] we're have to launch these products, but
[21:21] we're really not sure what the product
[21:23] is now and we're trying to sell it, and
[21:25] we haven't figured out the legal form,
[21:26] but we're going to bill for it anyways."
[21:28] And so, we need someone to help
[21:29] represent this to our investor and the
[21:32] investment community, which happens to
[21:33] be your CPA, just like diligence. And
[21:36] so, teams need that. You need a CPA in
[21:38] the room, a qualified technical
[21:40] accounting expert in the room when
[21:42] you're launching your products to at
[21:44] least work with your lawyers to figure
[21:46] out a defensible position with your
[21:48] auditors that's consistently applied,
[21:49] and you have to choose a lane for now,
[21:52] and hopefully in the long term, and not
[21:54] hopefully in the middle term, we can
[21:55] update the guidance to be a little more
[21:58] consistently or easily applied for this
[22:00] new world of living. Cuz it to be fair
[22:02] to have everyone who wrote 606, it's
[22:03] been 10 years now or whatever.
[22:05] I don't think anyone anticipated this
[22:07] type of change in the market, so.
[22:09] >> Yeah, I don't think anyone did. And as
[22:11] you're talking through that, I feel like
[22:13] all I hear is the word like really
[22:16] understanding your why, and I can't help
[22:19] but loop it back to our other favorite
[22:22] topic that we like to talk about, which
[22:23] is AI, right? How to use it and how our
[22:27] world is changing in it, which is like
[22:29] right now we don't we prompt very
[22:31] differently than we did last year or the
[22:32] year before. Because right now we
[22:33] actually set goals, right? You set
[22:35] goals, like right now it's July, so
[22:37] we've both been playing around a lot
[22:39] with Babel, right? And Babel's a very
[22:41] different animal than the other models,
[22:43] right? You have to use its little expert
[22:45] big brain differently, or else your
[22:47] costs will skyrocket, right? It's a very
[22:50] expensive and intense model. And it's
[22:51] really, really good at edge cases and
[22:54] judgment.
[22:55] >> Yeah, very impressive.
[22:57] >> Today
[22:59] will be very different than 1 month, and
[23:01] something else will blow our minds. But,
[23:02] if you think about that idea of like the
[23:05] way you prompt or solve a problem right
[23:08] now is you figure out the why really,
[23:11] really strongly, because that drives
[23:13] like how you get there can be many, many
[23:15] different ways. But, the why and the
[23:17] goal, which is like you said, is what is
[23:19] economically happening in this contract?
[23:21] Like, what are we trying to do as an
[23:22] accountant? Like, what is the underlying
[23:25] economic substance? And if that's the
[23:26] goal, if the goal of our job is to
[23:29] record what is happening, and we get
[23:31] stuck in all of these weird little words
[23:33] of contingent very like yes, we have to
[23:36] navigate it, but I'm almost tempted to
[23:37] just put in like this weird complex
[23:40] thing that we're dealing with as a
[23:41] society in the accounting world, and
[23:43] like see what it comes up with. Like,
[23:44] maybe think of like tell it to ultra
[23:46] think, which also just came out, right?
[23:48] Like, what would happen if we ultra
[23:50] >> ultra super think level 10,000, you
[23:53] know?
[23:54] >> all these weird little code words.
[23:58] But, like I'm tempted after this like
[24:00] what if we tried to ultra think this
[24:01] problem? Because I'm not sure like
[24:04] I think we both know the accounting
[24:06] guidance hasn't caught up. I don't know
[24:07] how it could. We we could barely keep
[24:09] up, but I I would I would I would have
[24:11] to be humble and say I'm probably not up
[24:12] to date, cuz how could you possibly be?
[24:14] But, like I don't I don't think we're
[24:16] doing what we what we think we want to
[24:18] be doing, at least in the accounting
[24:19] guidance. So, maybe that's the like
[24:21] contingent revenue cap is the best way I
[24:23] can think about it right now of bringing
[24:24] us back to a link of reality. But, like
[24:26] really framing like what is the goal?
[24:29] What is the goal of the guidance? And if
[24:31] the goal of the guidance is to reach the
[24:33] economic substance, which I know is also
[24:34] how we got to this rule for the
[24:36] contingent asset or the asset contract
[24:38] asset.
[24:40] But, like when you take that step back
[24:41] and you look at all the systems and the
[24:43] processes that are going into place to
[24:44] create fake realities within a company
[24:47] that get in then gets like redigested by
[24:49] the investor community back to
[24:50] consumption, which barely anyone can
[24:51] understand anyways. Like, right? Like,
[24:53] is this serving the goal?
[24:55] >> Yeah, well said. I think I wanted to go
[24:57] I think it bridges really nicely to your
[24:58] next topic that you want to talk
[25:00] through, which is cost. Because at the
[25:02] end of the day, what are companies
[25:03] trying to do? They're trying to solve a
[25:04] problem for profit. How do you make a
[25:06] profit? Well, you have to make margin on
[25:08] whatever your costs are. And so, every
[25:10] business right now is trying to figure
[25:12] out how much AI will cost them, how they
[25:15] can charge on top of that with all of
[25:16] their other employee business costs, et
[25:19] cetera, and and ultimately drive a drive
[25:22] positive operating cash flow and drive
[25:24] net income and drive positive margin.
[25:26] And to do that, their pricing has to
[25:29] match how they're being priced. And we
[25:31] all know Anthropic is a
[25:32] consumption-based business and all these
[25:34] other businesses are very I mean, they
[25:36] have their subscription offerings, but
[25:38] any business is going to be using
[25:40] specific consumption. And like you said,
[25:42] which I think is the best connector, is
[25:43] like these models are getting expensive.
[25:45] Like very, very, very expensive. Like
[25:47] when I had the max, $200
[25:50] to be playing with this. And as soon as
[25:52] I used to never run out with Opus. I
[25:53] used Fable for like 5 minutes and it ran
[25:56] out. Like my 2-hour window ran out like
[25:58] in a second. I was like, "Whoa." And I
[26:00] companies I've represented I've seen
[26:02] their spend going from hundreds of
[26:03] thousands of dollars a month to a
[26:05] million in a week with Fable. So, yeah,
[26:07] Anthropic's trying to make its money and
[26:09] we'll see we'll see how long people are
[26:11] going to pay for that, too. So, that's
[26:12] kind of a question in the market, too.
[26:13] How much are people willing to pay for
[26:14] this? Or companies are looking for cost
[26:16] benefits. So, if people stop being like,
[26:18] "Whoa, I'm not going to pay that much."
[26:19] People need to shut off Fable or
[26:20] whatever it is, that's going to be a
[26:21] different story. But, if from the
[26:23] matching principle of revenue is
[26:24] supposed to recognize the costs that are
[26:26] related to the revenue that's being
[26:27] generated, then
[26:30] like should we all be moving to a
[26:31] consumption-based world, which then it
[26:33] goes back to your like well, easiest way
[26:35] to do this is to at least mirror
[26:36] invoicing and some to some degree. So,
[26:39] kind of land back in the same spot. But,
[26:42] but that is an interesting piece, right?
[26:43] Like the the cost basis, which we never
[26:45] had in SaaS used to be profitable
[26:47] because well, you after you built the
[26:48] products and people bought it, then it
[26:50] it just pretty like just you have to
[26:51] worry about the maintenance. But now we
[26:53] are really in a world where yeah, these
[26:54] AI products are seem to be getting more
[26:56] expensive not less, which I don't think
[26:58] a lot of people were anticipating.
[26:59] >> Yeah, I think we are at an inflection
[27:02] point in July of 2026 with what happened
[27:06] with the Fable model because it is such
[27:08] a different animal. I think we we were
[27:11] messaging back and forth about I was
[27:12] like, "Oh, you you got to use Fable as
[27:14] the big brain and Sonnet as the like as
[27:17] like the manpower to implement, right?"
[27:19] So like Fable's the advisor, Sonnet
[27:21] actually does is the workhorse. And so
[27:23] Sonnet will come up with the plan, you
[27:25] do all the planning or the architecture,
[27:27] it does the the adversarial attacks and
[27:29] the edge cases, right? But who
[27:31] implements it? Like it's the Sonnet
[27:33] agents that actually go out and do it.
[27:35] So I say that all to say that even the
[27:37] way we use the models are constantly
[27:39] changing. I think the surprise bills
[27:42] this month, I think a lot of people will
[27:43] be angry just to be totally honest and
[27:45] I'm not sure how it's going to be
[27:46] handled, but I'm sure that the
[27:48] reconciliation back to reality is going
[27:49] to happen through billing.
[27:51] >> Yeah, yeah. Well, here's the thing that
[27:52] people like people tend to miss
[27:54] especially with the AI in businesses,
[27:57] everything's token-based right now,
[27:59] right? And so like based on how many
[28:00] tokens you use, you're going to get
[28:02] charged. That's from from these AI
[28:04] companies like Anthropic, Gemini, etc.
[28:08] for APIs and their connections to it.
[28:10] But everyone's using tokens differently.
[28:12] Like I hear I have some companies I
[28:13] represent that are like, "We're selling
[28:14] tokens." And then they have a completely
[28:16] different definition or credits that are
[28:19] related to tokens. They're all using
[28:20] similar nomenclature, but there's a
[28:22] problem with that pricing theory. And
[28:24] the problem is you have no way of
[28:26] anticipating how people how much tokens
[28:29] like let's say you have used Fable
[28:30] versus Opus. And if Fable costs two
[28:33] times as much as Opus, that's what they
[28:34] say. It's going to cost two much times
[28:35] as much per token equivalently. But
[28:37] Fable is also just going to run a ton
[28:39] more tokens cuz it's trying to think a
[28:41] bunch more. So for the same use case, it
[28:43] might actually cost a thousand times
[28:45] more. It's going to do it better, but
[28:46] it's going to run a lot more tokens to
[28:49] solve that same problem. So, it's like
[28:50] you can't We're not even working
[28:52] tip-for-tat. It's like if I sold you a
[28:54] pair of shoes and I like, here's your
[28:56] Prada shoes or I don't know, nice shoes.
[28:57] I don't buy nice shoes cuz I'm an
[28:58] accountant. So, whatever, your nice pair
[29:00] of shoes and then here's your Walmart
[29:02] brand and the Prada ones are 10 times
[29:03] more expensive. At least you know you're
[29:04] getting a pair of shoes.
[29:06] But here, you might be like, I'm
[29:08] choosing Fable and you think you from a
[29:09] shoe to shoe, you don't just spend 10
[29:11] times more, but you also somehow show up
[29:12] with a thousand extra shoes on your
[29:14] doorstep. You know, like you you're act
[29:16] can't anticipate how much they're
[29:17] actually ordering in these situations.
[29:19] So, like you said, I think it'll be
[29:20] interesting to see how that resolves.
[29:22] >> Yeah.
[29:23] >> Over the next couple months.
[29:24] >> And when you were bringing up that we're
[29:25] probably raising more questions than
[29:27] there are answers, but I I think about
[29:29] the fact that like, yeah, you can even
[29:31] talk about tokens, but the the the price
[29:32] of a token is totally made up. So, the
[29:34] concept of a token and a price of a
[29:36] token. So, like picture again, two two
[29:39] variables. Like as at the end of the
[29:40] day, we're human beings and all of all
[29:42] of our ideas are based on Actually, let
[29:45] me take a step back. This is something
[29:47] I've always found fascinating. Our
[29:48] concept of currency is completely made
[29:50] up. Meaning, my belief in what a dollar
[29:53] costs is only only works if you also
[29:55] believe that dollar is worth that much.
[29:57] >> Right.
[29:58] >> The concept of currency is actually
[30:00] fully based on a trust-based belief
[30:02] system of value. And so right now, we're
[30:04] actually working out that trust-based
[30:05] belief of value. So, what is a token
[30:07] worth in size, right, and in terms of
[30:10] dollar amount. And so like, Anthropic
[30:12] and OpenAI have completely different
[30:15] ways of strategy of dealing with this.
[30:18] Like I I remember I saw this like Codex
[30:20] ad pop up that was like, when when when
[30:23] when like cloud runs out, go ahead and,
[30:25] you know, pick it back up with Codex.
[30:27] And that's actually true. Like that is
[30:28] like, you get so much like you never
[30:31] I've never run out. I've never run out
[30:33] on ChatGPT or Codex. Like I've never run
[30:35] out. Whether that strategy has to do
[30:37] with me getting addicted to using it,
[30:39] questionable.
[30:41] But like, that just means that we're
[30:43] playing a little bit of like we don't
[30:45] know how this cost is going to roll out
[30:46] cuz that cost could be absorbed with you
[30:48] know their current investors, right? It
[30:50] could just be this weird amount on their
[30:52] books. It could be some weird liability.
[30:53] It could be caught in like partnership
[30:55] like we don't actually know how much
[30:57] this costs yet. Like I'm saying me and
[30:59] you don't know but maybe someone deep
[31:00] inside that company knows.
[31:03] >> Well no it well it's it's strange
[31:04] because just like the models, right?
[31:06] Like you're not getting deterministic
[31:08] outputs. It's non-deterministic. So it's
[31:10] a little weird. It reminds me of and I'm
[31:12] going to shout out to any gamers
[31:14] watching or any people like there's
[31:15] certain games that are RNG based like
[31:17] random number generation based that are
[31:19] there like for people who are trying to
[31:21] do things that they enjoy the experience
[31:24] of gambling but they're not going to go
[31:25] gamble and there's a lot of games that
[31:27] kind of scratch that itch. I'd say the
[31:29] sports betting itch and the prediction
[31:31] market and it's a little bit like that.
[31:33] I mean I'm curious in your opinion
[31:34] though and like I kind of feel that way
[31:36] whenever I'm running something. I'm like
[31:37] I wonder how much it's going to cost me
[31:38] this time. I'm not sure. I wonder if
[31:41] it's going to eat up all my usage or I
[31:42] wonder if it's going to just take me and
[31:44] maybe get a little better at it as you
[31:45] like work out specific workflows but I
[31:48] can only imagine at scale there's a
[31:49] little bit of this RNG random number
[31:52] generation of like what is
[31:55] what is this use case actually going to
[31:56] cost me to solve?
[31:57] >> Yeah.
[31:58] >> It's like it's not clear. It's not
[31:59] transparent because we don't there's no
[32:01] way someone can know oh yeah I'm going
[32:03] to do a reconciliation on this on my
[32:05] month end. I know that's going to cost
[32:06] me 27,000 tokens. Like from you and me I
[32:08] like we know what a dollar is but a
[32:10] token is pretty abstract. So
[32:12] >> Yeah, well I think that I think we are
[32:14] just hitting that phase. Like two months
[32:17] ago I wrote this up stack article about
[32:19] you know like everyone's trying to
[32:20] figure out the ROI but we were actually
[32:22] just trying to get people to use the
[32:23] technology, right? Like how do we get
[32:25] the accounting industry comfortable with
[32:27] like playing around cuz the only way
[32:28] that you can learn is to play.
[32:30] >> Yeah.
[32:30] >> And the spout seemed to be open. You
[32:32] could you seemed to be able to run and
[32:34] maybe you'd hit that that barrier but
[32:36] then you're like oh I'll switch to the
[32:37] next one. But like, what we were trying
[32:38] to do was just adoption actually at that
[32:39] point. I feel like now we're hitting
[32:41] this point where we're like, oh my gosh,
[32:42] this bill's kind of high. But like, why
[32:44] is the bill high? Is the Is the bill
[32:46] high because now we're getting charged
[32:47] by it we weren't before? I don't know.
[32:48] >> Yeah.
[32:49] >> Right? Like I actually have no way to
[32:50] audit the system. Or is it because the
[32:53] the rate went up? Like I don't like
[32:55] PG&E, for example, and a lot of people
[32:57] say that it's moving towards this
[32:59] utility-based thing. It's like suddenly
[33:01] like one month my my bill is like double
[33:03] or triple sometimes. And it's like, oh,
[33:05] it's just cuz we decided that the rates
[33:06] were going to be higher for San
[33:07] Francisco. I was like
[33:09] I I have no control over them doing
[33:11] that. And and I have to use electricity,
[33:13] right?
[33:14] >> Yeah. Yeah, exactly.
[33:16] >> And and and because the world is moving
[33:17] like it is kind of like that random
[33:19] number generator, but like there is more
[33:21] control at certain levels and less
[33:22] control. We probably have the least
[33:24] control. You know, two months later from
[33:26] writing that people were like token
[33:28] maxing, right? Now it's like we got to
[33:30] pay attention cuz I just ran through my
[33:32] budget. Like I can't just build random
[33:35] apps. Like I have to have some
[33:36] destination, which honestly kind of
[33:39] sucks for those that hadn't been playing
[33:40] around before.
[33:42] >> Yeah.
[33:42] >> Right? Cuz I don't know how you learn
[33:43] without being able to waste some. A lot,
[33:45] actually.
[33:47] >> Mhm.
[33:48] >> And so like, it's weird cuz the
[33:49] education bar just got a little higher
[33:51] actually cuz now it's more expensive to
[33:52] learn in a way. Unless like if you're on
[33:55] Claude, like go nuts on Haiku, right? So
[33:58] on, it's pretty pretty good to like you
[34:00] can like probably do whatever you want
[34:01] on those instant models and chat you
[34:02] like whatever. But those expensive ones
[34:04] you probably like there should be like
[34:05] little warning signs or like there
[34:06] should be meters on them. You just say
[34:08] like, oh, and and they did do that. Like
[34:10] I'm actually very thankful for Anthropic
[34:12] that does say, you know, you're you're
[34:13] running close to your limit. Cuz then I
[34:15] like this honestly this week only I
[34:17] found my little usage meter in Anthropic
[34:19] cuz I also have this max plan.
[34:21] >> Yeah.
[34:23] >> And so then I started tracking it by my
[34:25] like usage window limit and then like
[34:27] weekly resets. But like I never had to
[34:29] play that game before
[34:31] until now. And I I think it's because
[34:33] the ground is moving underneath us.
[34:34] >> Yeah, I agree. I agree.
[34:36] And then yeah, if you have the question
[34:37] I think there's so many different ways
[34:39] it can go and we can't predict the
[34:40] future, right? But it's like will be a
[34:41] utility. And does this become more like
[34:43] utilities in the future? If it does, is
[34:46] that like Anthropic the next PG&E, you
[34:49] know,
[34:49] >> [laughter]
[34:49] >> like what
[34:50] or is or is it really a technology? And
[34:53] then if it's a technology, businesses
[34:54] are going to need some transparency and
[34:56] understanding of their pricing. And
[34:57] that's where I kind of go with like in a
[34:58] revenue model world, are we spying and
[35:01] selling shoes? Are like like just like a
[35:03] business like Nike, well I have to know
[35:06] how much each shoe is going to cost. Am
[35:07] I going to do product cost costing or am
[35:10] I selling access to a service? And I
[35:12] think the costs are eventually going to
[35:14] have to mirror that or at least become
[35:15] stable enough to understand. Like we
[35:18] can't live I'm going to say we can't we
[35:20] have to say yeah, but
[35:21] it
[35:21] eventually companies have to have
[35:23] predictable cost structures in order to
[35:25] survive. You can't have unpredictable
[35:27] cost structures just to survive. So you
[35:29] can build your business on top of it
[35:31] because those costs get passed down to
[35:32] consu-
[35:33] passed down to consumers
[35:35] and place a margin. And eventually
[35:36] people like I we're fortunate enough,
[35:38] right Angeleta, to like pay $200 a month
[35:40] for the plan. And there's a lot of
[35:41] people in our bubble that are fortunate
[35:42] enough. But you go down a few rungs of
[35:44] middle class like lower middle class or
[35:46] different even in different states,
[35:48] that's a lot of money to come out of a
[35:49] someone's budget to use a product. I
[35:51] mean it's a lot of money for me. It's
[35:52] like 200 bucks it was like it wasn't
[35:54] like yeah, let's go. It was like man,
[35:55] this is going to be that's a lot of
[35:56] money. It's more expensive than
[35:57] Riverside or any of these other
[35:58] companies that are like 20 30 bucks a
[36:00] month. Right? It's it's pretty premium
[36:02] cost. But
[36:04] in other countries, that's like the
[36:06] unforeseen like a lot of countries
[36:08] that's not affordable for most people.
[36:10] And then if the cost continues to climb,
[36:12] who's going to be excluded from it? And
[36:14] then what businesses will be excluded
[36:15] from it just from a cost perspective?
[36:17] Yeah, I mean it becomes an interesting
[36:19] like how to what's the long-term
[36:20] business model? And then I'll put the
[36:22] counter to it. Well, most of the
[36:24] investors in these companies are also
[36:26] their largest consumers like Google has
[36:27] a large stake in Anthropic, Microsoft
[36:29] has a large stake in OpenAI. And so I
[36:32] like to say out loud like, well, this
[36:33] will got to sort itself out. Well, not
[36:35] if like the pony show keeps, you know,
[36:36] continuing
[36:38] or it's like, yeah, well, rev number
[36:40] goes up for OpenAI, the number goes up
[36:41] for Microsoft. Well, I'm sure they'd
[36:44] love to see that continue as long as
[36:45] possible. And so, so that might move
[36:47] forward despite consumer challenges. But
[36:50] like I'm sure there's a number, right?
[36:51] Like if the next plan became a $2,000 a
[36:54] month or $10,000 a month. There's a
[36:56] number where it's just like, okay, this
[36:58] doesn't make sense anymore for our
[36:59] business. And so, yeah, I don't know. I
[37:01] don't know. Rambling a bit now, but you
[37:03] get what I'm saying, hopefully.
[37:06] >> I was wondering if I should wrap it back
[37:08] to accounting because I know we can go
[37:10] down this rabbit hole quite deeply.
[37:13] >> Yeah.
[37:14] >> But a a lot of companies, so this also
[37:16] came up in that May share forum is like,
[37:18] do you capitalize the cost these AI
[37:21] costs, right? To build these things. And
[37:23] how do you do that? And one of the
[37:25] thoughts that came together was, you
[37:27] know, that a lot of these companies are
[37:29] moving towards doing a per employee per
[37:31] employee budget, right? So, as an
[37:33] employee, you get a budget of 200, maybe
[37:36] like, you know, whatever plan or however
[37:38] many tokens cuz it's an extension of
[37:39] your ability to create, essentially.
[37:41] >> Yeah. Yeah.
[37:43] >> And so, um the discussion we had was, is
[37:46] it more like the way that you do a
[37:48] percentage of an employee costing with
[37:50] like their equity and things like that
[37:52] that was like a portfolio approach of
[37:53] how you determine how much you'd
[37:54] capitalize or what piece they're working
[37:56] on. But it's an interesting concept cuz
[37:58] in general, it is this thought process
[38:00] of like, it's an extension of the human
[38:02] being. Like, who gets to have the
[38:04] extension of themselves? Like, you can
[38:05] afford it, I can afford it, like these
[38:07] companies can afford it. Cool, these
[38:09] companies can 10x their build speed,
[38:12] right? And release these things at this
[38:14] cost. And I bring back the at this cost
[38:17] because, like you said, software used to
[38:19] be like 90% margin. You went into SaaS,
[38:21] it probably got a little lower cuz cloud
[38:23] costs. And now we're squarely in a
[38:25] different world, which is like you got
[38:26] your cloud cost plus you got your AI
[38:28] costs, AI token costs. And now that is
[38:30] being more equated to a human being,
[38:33] like a employee plus 200, employee plus
[38:36] 2,000, right?
[38:37] >> Yeah.
[38:37] >> Per month, right? And so now we're
[38:39] choosing, okay, you're going to be the
[38:41] enabled employee. This is how much this
[38:42] is going to cost. Like we're actually
[38:44] getting back a little bit to
[38:45] manufacturing, right?
[38:46] >> Yeah, that's Yeah, that's where my mind
[38:48] goes. That's why I keep saying Nike and
[38:49] whatnot, right?
[38:50] >> Now like lifeo, FIFO costing, right?
[38:52] It's like actually your margins are
[38:54] only, you know, in manufacturing what, a
[38:55] 30% margin was really good. Maybe that's
[38:58] where we're going and we haven't caught
[38:59] up with it. But that would be quite
[39:01] interesting because like there's been
[39:03] all these predictions that like, you
[39:04] know, like AI will take your job, but I
[39:06] know that me and you are busier than
[39:07] ever. Doesn't feel like AI is going to
[39:08] take my job.
[39:10] >> It feels like the opposite, yep.
[39:13] >> It just seems like you've more crap to
[39:14] sift through, right? But like I I think
[39:17] that that's an interesting piece. Like
[39:18] are we moving closer to the
[39:19] manufacturing world, which it it already
[39:22] kind of felt like for certain
[39:24] infrastructure-related, cloud
[39:25] cost-intensive things? And now it's
[39:27] like, oh, it's a per employee plus
[39:30] count. Interesting.
[39:31] >> And yeah, if we go in that world, the
[39:32] manufacturing world, then my gut says
[39:34] like why are we capitalize it? Like this
[39:36] is all short-term. And like it kind of
[39:38] like if we keep changing and updating
[39:40] products and everyone can make new
[39:42] products all the time and we're going to
[39:44] update and sell, does it stand ready or
[39:46] do we just have these little discrete
[39:47] products? Are we selling shoes? And if
[39:49] we're selling shoes, well, the new pair
[39:51] is going to be we're going to sell
[39:51] another pair next month, so do we need
[39:54] to capitalize for all the new all these
[39:56] developments because like the
[39:57] obsolescence is so fast. Yeah.
[39:58] >> Absolutely, right? Like I I In In some
[40:01] calculations, right, it's like how much
[40:03] of the code base do you rewrite and
[40:04] reuse, right? And it's like, I mean, me
[40:07] and you are new developers, if we could
[40:09] even say that.
[40:11] >> Yeah.
[40:11] >> Like there's tons of stuff
[40:12] >> New pseudo developers.
[40:13] >> I have no idea what I'm doing most of
[40:14] the time. So like, you know, like it's
[40:16] it's an it's an interesting place. Like,
[40:18] maybe we are getting closer to selling
[40:21] little widgets, but at some point, there
[40:23] is going to be this code base that works
[40:25] really well. My gosh, okay, sorry, one
[40:27] more tangent. I've been thinking about
[40:28] this one a lot. Maybe this Maybe at next
[40:30] time one. I keep thinking about this
[40:32] very specific idea
[40:35] Like, in accounting, we have a, you
[40:37] know, term license concept, where the
[40:39] term license is worth X amount, i.e.,
[40:41] how much is the code base worth up
[40:42] front? If you did perpetual license that
[40:44] wasn't a cloud, and how much is
[40:45] maintenance, right? It was this weird
[40:47] thing that happened when we shifted to
[40:49] SaaS software. And so, you no longer had
[40:51] the weird blip of revenue up front and a
[40:52] small maintenance line. And as
[40:54] accountants, you always love to call
[40:55] this thing, like, every big four firm
[40:57] loves the 80/20 split. 80% is the
[41:00] license value up front, and 20% is the
[41:03] maintenance. Again, when we went into
[41:05] the SaaS world, that was like, "Oh,
[41:06] screw that. We're just going to do 100%
[41:08] over time." And that was nice for AR. We
[41:10] loved it. Now, we're in the AI world,
[41:12] and and I really think about that 20% a
[41:14] lot, and I think about how much it costs
[41:16] to build the first thing versus how
[41:18] insanely painful Like, anyone that's
[41:20] built software knows this how insanely
[41:21] painful it is to maintain that thing,
[41:23] and make sure that thing works. Is that
[41:24] 20% way too low? And does that then
[41:27] like 20% the upfront, and then the 80?
[41:29] That's funny. I never thought of that.
[41:31] That's great. Which actually points more
[41:32] towards straight line, which is your
[41:34] kind of first argument, which is
[41:36] interesting. Anyways, that I like wrote
[41:38] this on my board. I was like, "What is
[41:39] the cost of maintenance?" I was like, "I
[41:41] don't think it's 20%."
[41:43] Yeah, cuz I tried to like what I tried
[41:45] for my last class, I made an online
[41:47] module for all of my multiple choice
[41:49] questions, which no professor I I've
[41:50] never met a professor that's like,
[41:51] "Yeah, just made an online website with
[41:53] all your multiple choice questions using
[41:55] AI." It was great. It took me like a
[41:58] weekend to do that, which was amazing.
[42:00] Like, I I had hundreds of multiple
[42:01] choice for a finance class. The students
[42:03] loved it. But then a student came up to
[42:04] me and was like, "Hey, one of these
[42:06] multiple choice questions, could you
[42:07] change the answer?" Cuz I think it's a
[42:09] little off. I tried to go do that, to
[42:11] figure it out in the, you know in the
[42:13] code. It was going to take me longer
[42:14] than the freaking build, right? Like to
[42:17] go how how what's the one question and I
[42:19] like try to go work with Claude on it
[42:21] and Claude's like, "Well, we have to
[42:22] re-go through the whole code base." And
[42:24] it's just like, "Oh my gosh, which where
[42:26] did I put this and how do I access the
[42:27] code and which Yeah, like so exactly.
[42:30] We're going to have to go through that
[42:30] world really fast.
[42:31] >> What is the cost of maintenance? What is
[42:34] the real cost of maintenance? Because I
[42:36] think that the upfront build is so easy,
[42:38] the maintenance is going to be horrific.
[42:40] >> Yeah. Yeah. Yeah, exactly. So then you
[42:42] get people like us, the pseudo
[42:43] developers, right? The pseudo developer
[42:45] like, "Yeah, we made this amazing
[42:46] thing." Yeah, but it I wanted to do this
[42:48] little thing. Can you fix it? Oh my
[42:49] gosh, no, we can't fix it. It's going to
[42:51] take us a year. I want to I want to
[42:53] bring one other thing. I didn't do Have
[42:54] you heard of Jevons paradox because I
[42:56] think it's a cool way to close this.
[42:57] >> No, tell me.
[42:58] >> Okay, that was my thing I learned about
[43:00] this week where there's a traditional
[43:02] theory that if we the more something
[43:06] becomes available, the cheaper it
[43:08] becomes. Right? In in economics. Like
[43:11] supply goes up, costs go down.
[43:13] Jevons paradox said the there's that's
[43:15] actually false in a lot of industries.
[43:18] And he's the coal This guy Jevons used
[43:20] the coal industry as an example. As like
[43:22] as something becomes more available, a
[43:25] lot of times the cost actually
[43:27] increases. Especially in the short to
[43:30] mid-term because consumption increases.
[43:34] And there's an interesting example of
[43:36] water heaters and coal. Which so like
[43:39] people 100 years ago, I don't know when
[43:41] the water heater was invented, so take
[43:42] that plus or minus whatever amount of
[43:43] years, didn't have water heaters. So it
[43:45] was like a pain to heat water for your
[43:46] bath. People took showers once a week
[43:48] and just like threw cologne on or
[43:49] whatever. And you had to like boil water
[43:52] over a gas stove or and people thought
[43:54] that like, "Well, when we there's only
[43:57] such demand for hot water. Like you need
[43:58] to bathe once a week." But as soon as we
[44:00] made water heaters, it turns out there's
[44:01] this exponential demand for quite a
[44:03] while. People like, "Yeah, I like hot
[44:04] showers every single day." So water
[44:06] heaters went from being small to pretty
[44:08] substantial and it seemed like the
[44:10] demand was infinite. So, like the access
[44:13] to more hot water energy, like the
[44:15] ability to heat water in water heaters
[44:17] or the access to coal is a good example.
[44:19] Like as soon as a steam engine was made,
[44:21] the train was invented which consumed a
[44:22] lot more coal than was ever needed
[44:24] before. It actually increased the price
[44:26] of coal, increased the price of gas and
[44:28] hot water in the short to medium term
[44:30] until you actually saturate the market
[44:32] with the overall need. And so, I think
[44:33] there's like this question of when will
[44:36] the market are we here now? Like will
[44:38] the costs of AI actually be exponential
[44:41] because and energy and all of this until
[44:43] we actually figure out what our
[44:44] saturated the saturation point is which
[44:46] could be longer than we expect. Like
[44:48] with water heaters and coal it was like
[44:49] multiple decades, right?
[44:51] >> Does
[44:52] >> Yeah, I think we're at the beginning. I
[44:53] really do cuz it's it's it's
[44:55] intelligence or knowledge on tap, right?
[44:58] Oh, I I think people call it
[44:59] intelligence on tap, right? And like
[45:02] we've always wanted that like people I
[45:05] do think it's like that paradox where
[45:07] more people will want it. I do not think
[45:09] the cost is actually known or real yet
[45:11] at least to us.
[45:13] >> Yep.
[45:14] Yep.
[45:15] >> And I think we're literally making up
[45:16] new ways to increase capacity, right?
[45:19] The whole SpaceX piece, right?
[45:20] >> Yeah, it's data centers in space.
[45:23] >> Right? Creating new capacity. And so, I
[45:25] think it's interesting what you
[45:27] mentioned Jevons paradox is like if
[45:29] we're talking multiple decades, are we
[45:30] at the beginning? And I feel like that
[45:32] is, but every day just moves faster than
[45:36] the last one.
[45:36] >> Yeah. And well, and and so I I guess to
[45:38] that point, it's easy and I was in this
[45:41] camp like a year ago. I think you and I
[45:42] have had discussions like, "Well, the AI
[45:44] models are getting faster and better so
[45:46] costs should go down." I think everyone
[45:48] was kind of like, "Oh, that was the
[45:49] trend." Even when I was at Google, that
[45:51] was a very common thing is like, "Oh,
[45:52] this mod this pricing strategy seems
[45:54] very expensive now, but the price will
[45:56] go down as the models improve." And they
[45:57] are improving exponentially, but our
[45:59] demand is exceeding the improvement rate
[46:02] which is extremely interesting, which
[46:03] puts that Jevons paradox. So, well,
[46:05] like, okay, we're using Fable, like,
[46:07] Opus was great, and now we're using
[46:09] Fable, and it's like we can't get enough
[46:10] of Fable. So, if there's like 100 more
[46:12] iterations of this that are continuing
[46:13] to be more and more expensive, when will
[46:15] we be satisfied? Which I don't know.
[46:16] It's like you said, I think we're at the
[46:18] beginning, which is kind of crazy. And
[46:19] what does that mean for accounting?
[46:23] >> We always wonder how
[46:24] >> by it. Means stand ready, contingent
[46:25] revenue cap. Let's go.
[46:27] >> Yeah, but we we always go off path, but
[46:30] we always come back to the accounting
[46:31] because the why behind accounting is to
[46:33] capture the economics. And if the
[46:35] economics are the at the beginning under
[46:38] Jevons paradox, then we should be humble
[46:41] that we actually don't know where this
[46:42] is going to go because both lines are
[46:44] moving. The top line with the revenue,
[46:46] the cost line is moving, the technology
[46:48] is moving. It's not to say do nothing.
[46:50] It's just to say stay humble and stick
[46:52] to the pieces of the reality that we
[46:54] know, which might be the contingent
[46:55] revenue cap in that bill.
[46:56] >> Yeah.
[46:57] >> And then
[46:58] >> your contracts. Like, at least work with
[46:59] a get a CPA in the room with your
[47:01] lawyers to figure this stuff out, and
[47:02] don't pretend like you know it.
[47:04] Yeah.
[47:04] >> And you know, [laughter]
[47:05] >> we're figuring this out as we go.
[47:08] So, I hope that answered your I hope
[47:09] that answered your audience's question.
[47:12] I don't think they meant that for us to
[47:13] go on this tangent when they asked about
[47:14] road maps, but I hope it satisfies them.
[47:17] >> I'm sure it didn't, but I'm I'm excited
[47:20] about where it went.
[47:22] >> Okay, good. I'm glad. Well, this is fun.
[47:23] I think this is a good spot to pause,
[47:25] Angela, and we'll continue the
[47:26] conversation next week.
[47:27] >> Sounds good.
[47:28] >> Appreciate it.