Interview with Box CEO Aaron Levie (hosted by Dan Shipper) on why AI agents are likely to augment knowledge workers rather than replace them. Drawing on 20 years leading Box and the company's AI-first transformation, Levie discusses why jobs may persist, the Jevons Paradox and the future of work, how the cloud era shaped his view of AI, knowledge workers becoming managers of AI agents, lessons from rolling out AI across Box, what's overhyped in AI today, and balancing everyday execution with innovation. Chapters: 00:00:00 - Start 00:01:30 - Introduction 00:02:36 - Why AI won't take your job 00:06:42 - Jevons Paradox and the future of work 00:10:40 - How Aaron's experience with the cloud era shapes his view of AI 00:19:44 - Why every knowledge worker is becoming a manager of AI agents 00:25:21 - What Aaron's learned from bringing AI into every corner of Box 00:33:57 - What's overhyped in AI today 00:43:31 - How Aaron balances everyday execution with innovation # Channel Identity Observations — "Box CEO Aaron Levie on Why AI Agents Won't Take Your Job" ## Show Identity (Explicit Self-Description) - **Show name:** The outro names the show directly: > "smash that like button and subscribe to AI and I" - **Host:** The host is identified as **Dan Shipper** ("a journey into the future with Dan Shipper as the captain of the spaceship") and introduces himself in a sponsor read as **Steven Johnson, co-founder of Notebook LM** (a sponsored ad segment, not the main host). - **Promotional tone claim:** The outro describes the show as: > "pure unadulterated knowledge bombs about chat GPT... a roller coaster of emotions, insights, and laughter" ## Claimed Scope - **Subject matter:** AI, specifically AI agents, their workplace impact, and enterprise adoption. Evidence: the framing monologue — > "If you care about the skills you're going to need in an economy filled with AI agents, this is the interview to watch." - **Format:** Long-form one-on-one interview with tech executives and founders. Evidence: nearly the entire transcript is a host–guest dialogue with Box CEO Aaron Levie. - **Stance toward guests:** The channel positions itself against extreme AI narratives, favoring "grounded" perspectives: > "that's why I like talking to people like you, cuz I think you have a more grounded perspective... The bottlenecks just move somewhere else." ## Intended Audience - **Knowledge workers and builders concerned about AI's impact on jobs.** Evidence: the opening hook contrasts this guest with "CEOs predicting the end of white collar work," and explicitly addresses viewers who "care about the skills you're going to need." - **Practitioners/operators in tech.** The host contributes operator-level anecdotes (running "four software products internally... 15 people," using "cloud code and codec cli"), suggesting an audience of startup and technical listeners. - **Sponsor alignment:** A Google/NotebookLM ad targeting "anyone trying to make sense of complex information" supports an information-worker audience. ## Observable Presentation Patterns (Demonstrated, Not Claimed) - **High-profile guest access:** Landing a sitting CEO of a ~2,000-employee public company ("He's been running the company for almost two decades"). - **Prepared, research-driven questioning:** Host references guest's history ("longtime ex poster"), company timeline, prior memos ("you sent the memo and you were like... we're transitioning"). - **Host participation:** The host shares personal experience rather than remaining neutral: > "I have caught like hundreds of thousands of dollars worth of legal mistakes just by putting my contracts into GBT5." - **Conceptual framing by the host:** The host offers original terminology: > "moving from a knowledge economy to an allocation economy where your job is to allocate intelligence." - **Sponsorship model:** The podcast includes an inserted sponsor read ("This podcast is supported by Google"). - **Humorous/irreverent outro style:** The closing CTA is deliberately exaggerated ("I'm absolutely hopelessly in love with you"), demonstrating a playful on-channel voice. ## Creator Claims vs. Demonstrated Behavior | Type | Evidence | |---|---| | **Claim (about the show)** | "the epitome of awesomeness," "knowledge bombs about chat GPT" (outro) | | **Demonstrated** | Substantive executive interview covering AI economics, adoption curves, product strategy, model trade-offs | | **Claim (about audience value)** | "If you care about the skills you're going to need... this is the interview to watch" | | **Demonstrated** | Content is primarily CEO strategic perspective, less directly skill-building; framing slightly overstates the practical-skills angle relative to the interview's executive-strategy focus | # Delivery Observation: Aaron Levie on AI Agents Won't Take Your Job ## Speakers Observed - **Aaron Levie** (guest, Box CEO) — the primary subject of observation - **Interviewer** (Dan Shipper, per closing) — conversational, casual, frequent profanity and informal interjections - **Cold-open narrator** and **sponsor read** are brief framing segments --- ## 1. Tone ### Levie: Enthusiastic, hedged optimism - Predominantly **energetic and informal**, with self-deprecating humor: > "I'm going have to be introspective for 5 seconds." > "this is sort of me in my happy happy place" - Openly **quantifies his own uncertainty**: > "I still leave open a 5% chance that I'm like totally obscenely wrong." > "I'd say it's 80 to 90% very excited, 10% 20% anxiety." - Recurring playful asides: > "I'm just going to tweet about AI agents till the whole thing's over." ### Interviewer: Casual, validating - Frequent agreement markers: "100%," "Totally," "I agree with you," "We're so back." - Reassuring framing: "This is safe space, but I'm really curious." --- ## 2. Vocabulary & Recurring Phrases (Levie) ### High-frequency conceptual vocabulary - **"Bottleneck"** — used repeatedly across contexts: > "that particular bottleneck in the organization" > "create newer bottlenecks... that cause us to hire more people" > "removing the current bottleneck" - **"Tasks" vs. "jobs"** — his central framing device: > "jobs are not tasks, jobs are a collection of tasks" - **"Workflow"** — extremely frequent (legal, engineering, M&A, sales contexts) - **"Prototype," "trade-off," "throughput," "scaffolding"** ### Repeated catchphrase-like constructions - **"you know"** — pervasive filler throughout his answers - Doubling for emphasis: "happy happy place," "full full crank mode" - Self-aware disclaimers before analogies: > "I'm totally this is fanfiction, but..." > "maybe optimistically I'm totally making this up" --- ## 3. Teaching Habits ### Pattern 1: Concrete example after abstract claim Nearly every argument lands on a specific job function — **lawyers** most of all: > "if I could have lawyers go and review legal... documents... two times faster, I would not have half the number of lawyers." Other repeated examples: engineers reviewing agent code, graphic designers/Photoshop, accountants/Excel, United Airlines, Disney. ### Pattern 2: Historical analogy as forecasting method He explicitly names this method: > "I'm kind of working through analogy... we went from the mainframe to the PC. We went from PC to mobile. We went from on-prem to cloud." - Invokes **Jevons paradox** and applies it to labor. - Uses the adoption-curve pattern ("early adopters... breakthrough use cases... laggards") as a predictive template. ### Pattern 3: Quantified framing - Assigns percentages to confidence, excitement, time allocation ("80% of the time is going into the incremental bottleneck... 20% is experimenting"). - Explicitly flags invented numbers: > "we can make it happen in 1 minute and we know it'll be 93%. I'm making up all the numbers." ### Pattern 4: Rhetorical questions to drive the argument > "Well, shouldn't that... reduce jobs in some kind of commensurate way?" > "what are all the things today that you're not doing because the price of entry to doing that thing is I have to hire a person?" --- ## 4. Communication Patterns ### Interviewer's habits - **Stacks credentials before questions** ("you're a longtime ex poster... you've really gotten your hands in it"). - **Long multi-clause setups** ending in a short direct question. - Heavy informality: "[ __ ]" expletives, "crazy wizard [ __ ]." ### Interaction dynamics - Mutual agreement and mirroring ("We're so back." / "We're so back. Yes."). - Levie **asks clarifying questions back** when questions are broad: > "And to clarify, do you mean internally operationally or for the product experiences we're building?" --- ## 5. Stability vs. Isolated Moments ### Stable patterns (strong repetition) - "you know" filler; concrete-example teaching; analogy-based reasoning; quantified confidence; "bottleneck"/"workflow" vocabulary — all recur across multiple answers. ### Isolated moments (insufficient repetition) - The **prompt-engineering-a-podcast joke** and the Richard Sutton/Dwarkesh reference appear once each — single instances, not habits. - The guitar/COVID anecdote is a one-off self-disclosure. - The outro ("smash that like button... hopelessly in love with you") is a distinct scripted voice, not the interview delivery style — attribution to a single speaker is uncertain from the transcript alone. ### Uncertainty note - Timestamps and speaker overlap are ambiguous in places; some quotes above rely on transcript speaker cues (">>"), which may imperfectly separate voices. Verbal disfluencies ("I I I'm," "um") are heavily represented in Levie's turns and are a consistent, observable feature of his rapid, digressive speaking style. # Position Record: Box CEO Aaron Levie on Why AI Agents Won’t Take Your Job **Participants (from transcript evidence):** - **Aaron Levie** — CEO of Box, running the company ~20 years (per host introduction) - **Dan Shipper** — Interviewer/host (identified in outro as show captain; conducts the interview) --- ## Core Thesis: Jobs Persist Because Jobs ≠ Tasks **Levie's central argument:** - Jobs are **collections of tasks**, not single units. AI automates tasks well, but "there still is... a human to incorporate whatever the task that was executed into a broader workflow... into actually some form of value creation." - Because that human role can't be eliminated, people will still **specialize** by owning completion of task sets in their domain. - Example: An engineer prompts an agent to work on code, but still decides whether to ship, checks with product managers, and integrates into larger systems. **Explicit qualification (preserved verbatim in meaning):** > Levie: "I still leave open a 5% chance that I'm like totally obscenely wrong." He characterized his position as **high confidence with room for debate and internal doubts**, and "somewhat empirical when you use these tools." ## Efficiency Gains → More Work, Not Fewer Jobs **Levie's reasoning chain:** - If output per worker rises 20–100%, jobs shouldn't shrink commensurately because "we are doing way less work than what is actually economically useful" — constrained only by time and labor cost. - **Lawyer example:** 2x faster contract review wouldn't halve the legal team; throughput increases, feedback loops speed sales, customer satisfaction, and revenue — "ironically then would lead to hiring more people." - **Jevons paradox applied to labor:** cheaper legal work increases demand for legal work, opening "a new trunch of use cases." - **Historical analogies (self-labeled "fanfiction"):** graphic design after Photoshop, accounting after Excel — vastly more practitioners today than 30–40 years ago. **Shipper's supporting claim:** - He says he caught "hundreds of thousands of dollars worth of legal mistakes" by running contracts through GPT-5 — a second-level legal review he "couldn't have paid for before." ## Humans Retain Irreplicable Context **Levie's argument:** - "As long as there's still a three-dimensional world out there that we have to go and participate in, we're going to have much more signal, much more context than the AI will." - Cites the Richard Sutton/Dwarkesh podcast: next-token prediction is "insanely valuable" but systems lack context and ability to predict the future. - In-person serendipity (a hallway conversation yielding a customer insight) "can't be replicated in an AI system right now. We don't know how." - Humans also don't need constant prompting — unlike agents — making interaction efficient. - AI progress "automates x function... by 80% and then people have to do the rest" — a rolling pattern he expects to continue. **Shipper's addition:** tacit knowledge from lived industry experience can't be replicated even by an AGI that wasn't "there." ## Disagreement / Tension with Big-Lab CEOs - Shipper notes big-lab CEOs (Sam Altman, Dario — surname unstated) don't publicly hold this view. - **Levie's response:** He doesn't mind their ambition ("I'd rather have the... crazy utopian future than the dystopian angle"), and believes "by the time the technology hits the real world, it will manifest just a bit differently." - On job-dislocation claims attributed to Dario (stat unspecified — Levie explicitly declines to quote precisely), Levie counters that even when a function is automatable (e.g., password-reset emails), he reallocates freed budget to underfunded strategic areas — a dynamism he says is missing from dislocation conversations. ## Recessions: A Stated Uncertainty - **Levie:** Recessions would still bring job cuts, "as it kind of always has" — but AI-driven output could help escape faster. He flags that AI-era cuts would be wrongly blamed on AI because you can't observe the counterfactual. He labels his optimism here: "maybe optimistically I'm totally making this up." ## Platform-Shift Lens and AI vs. Digital Transformation **Levie:** - Draws on prior shifts (mainframe→PC, PC→mobile, on-prem→cloud) with a standard adoption curve — but AI is compressed: cloud took ~10 years to mainstream; AI "probably... in two years." - **Digital transformation was often superficial:** United Airlines' customer experience "basically nothing changed" (possibly worse); Disney and some banks are more extreme examples of real transformation. - **AI is different** because it alters every employee's daily work; the efficiency gap between AI-using and non-AI companies is "too insurmountable." - Engineering shift is "probably the biggest shift in any period in history of almost any knowledge worker job": one year ago, IDE + autocomplete; today, prompting agents that do large amounts of work. - Compares magnitude to paper→digital **plus the internet**, compressed into one or two years. - **Shipper's framing (agreed by Levie):** shift from a **knowledge economy to an allocation economy** — ICs become managers of agents, doing prioritization and judgment. ## Running Box as an AI-First Company **Levie's lessons and admissions:** - Framed AI internally as **not** about replacing jobs or cutting spend, but increasing output — to defuse threat. - Weekly Friday all-hands with employee demos of Box AI agents (sales workflows, compliance). - Pressure on timelines: "why can't we do that in two days?" creating a flywheel. - **Candid asterisk:** 5–20 person startups are "so differently wired" that he questions whether Box must re-engineer some areas from a clean sheet; startup workflows are spec-driven, prompt-driven, agent-first. - Emotional split stated honestly: **"80 to 90% very excited, 10–20% anxiety."** He picked up guitar at the tail end of COVID as evidence the industry had "crested" — now back in "full crank mode." ## Product and Model Strategy **Levie:** - Time split: **~80% incremental bottleneck removal, ~20% "magic moment" experiments** (e.g., long-running due-diligence/loan-agreement agents — still "very early" but "it will work"). - Trade-offs are compute/time/hit-rate: e.g., 10 seconds at ~50–70% success vs. 1 minute at 93% (numbers explicitly made up as illustration). - **Unprecedented UX work:** more novel UX patterns in 18 months than in 20 prior years; software is now "labor" without classic patterns. - **Model choice: always bet on the best model, "cost be damned."** Competitors will; scaffolding built to mitigate older models' limits (e.g., GPT-3.5's ~6–8k context window) becomes obsolete with 100k+ tokens. Inferior experience = uncompetitive. **Shipper's parallel claims:** runs four software products with 15 people, commits code himself ("I should not be able to"); with cloud code/Claude-style tools, "the actual code doesn't matter as much" as the plan. ## Overhype Question — Notable Asymmetry - **Levie declines to name anything overhyped**; instead predicts he "can't find" an AI category that won't be 10x larger in five years, analogizing to underestimating SaaS/AWS growth. - **Shipper** names the overhyped thing as the oscillation between utopia ("free room service and teleportation") and doom ("we're all going to die"), preferring Levie's "bottlenecks just move somewhere else" framing. --- ## Unresolved Tensions (Preserved) - Levie's own 5% admission that he may be "totally obscenely wrong" vs. his "all of the evidence right now is pointing" confidence. - His acknowledgment that recessions *would* still cut jobs, with attribution-to-AI as a confound he "can't really know." - Box's AI-first success is explicitly **not** claimed as complete: "still totally on that journey... can't put the mission accomplished flag in the ground." - Startup-vs-incumbent workflow gap admitted as an open problem he hasn't solved. # Speaker Attribution Record **Video ID:** 01a0f15a-d909-7e74-8eb1-9dbae62ef971 **Title:** Box CEO Aaron Levie on Why AI Agents Won't Take Your Job ## Speakers and Roles - **Guest: Aaron Levie (also rendered "Aaron Levy" in transcript)** — CEO of Box; stated in transcript: "That's Box CEO Aaron Levy. He's been running the company for almost two decades. They have 2,000 employees." Delivers the bulk of substantive claims on AI, jobs, and Box's AI-first transition. (Spelling discrepancy between title "Levie" and transcript "Levy" noted; identity not otherwise verifiable.) - **Host / Interviewer (primary)** — Identified in outro promotional segment as **Dan Shipper** ("a journey into the future with Dan Shipper as the captain of the spaceship"). Conducts the interview, asks questions, contributes personal anecdotes (e.g., catching legal contract errors using "GBT5"). Show name referenced: **"AI and I"**. - **Host / Narrator (intro)** — Delivers the cold-open framing before the ad read ("So I talked to one of the few CEOs..."); identity not explicitly stated, plausibly the same host but unconfirmed. - **Ad reader: Steven Johnson** — Self-identified: "Hey folks, Steven Johnson here, co-founder of Notebook LM." Reads the sponsor message only. ## Promotional Segments (Flagged) - **Sponsor read — Google / Notebook LM:** "This podcast is supported by Google." Steven Johnson promotes Notebook LM at **notebooklm.google.com**. - **Closing call-to-action:** Host outro urging viewers to "smash that like button and subscribe to AI and I," including the statement "Dan, I'm absolutely hopelessly in love with you." Flagged as promotional/hype copy, not substantive content. ## Referenced Sources and Third-Party Claims - **Jevons paradox** — cited by Levie as an analogy for labor demand; described as applied historically to "inanimate resources... energy consumption." - **Historical platform shifts** — mainframe→PC, PC→mobile, on-prem→cloud, cited by Levie as analogy basis. - **Third-party examples** — United Airlines (digital transformation critique), Disney (business model shift), Photoshop/graphic designers, Excel/VisiCalc/accountants. All are Levie's or host's characterizations, not sourced claims from those companies. - **Named individuals referenced (not speaking):** Sam Altman, Dario Amodei ("Daario" in transcript) — Levie references Amodei's job-dislocation statements but explicitly hedges: "I forget the exact stat, so I don't want to misattribute." Richard Sutton — referenced via "the Richard Sutton... podcast with Doresh" (name as given in transcript) regarding next-token prediction. - **Products/tools mentioned:** GitHub Copilot, ChatGPT 3.5 (with token-context-window figures the speaker flags as approximate), GPT-5 ("GBT5," host usage), Claude Code ("cloud code"), "codec cli," Figma, Box AI. ## Provenance Notes - Host personal claim (unverifiable): "hundreds of thousands of dollars worth of legal mistakes" caught via GPT-5 contract review. - Levie's claims about Box (employee count, AI-first memo, internal Friday demos) are self-reported by the speaker. - Levie explicitly labels one comparison "fanfiction" (Photoshop/graphic designer counterfactual). - **Comment availability:** No information present in source material. - **No curated metadata supplied**; only transcript available. AI is filled with CEOs predicting the end of white collar work. So I talked to one of the few CEOs who has a vision of what the world will look like where AI works for us instead of replacing us. As long as as there's still a three-dimensional world out there that we have to go and participate in, we're going to have much more signal, much more context than the AI will. And that's going to just keep us in jobs, keeps us doing things for as long as we can we can kind of look out right now. \>\> That's Box CEO Aaron Levy. He's been running the company for almost two decades. They have 2,000 employees and almost overnight, he's made the company AI first, and that's taught him a lot about what the future's going to look like over the next few years. If you care about the skills you're going to need in an economy filled with AI agents, this is the interview to watch. This podcast is supported by Google. Hey folks, Steven Johnson here, co-founder of Notebook LM. As an author, I've always been obsessed with how software could help organize ideas and make connections. So, we built Notebook LM as an AI first tool for anyone trying to make sense of complex information. Upload your documents and Notebook LM instantly becomes your personal expert, uncovering insights and helping you brainstorm. Try it at notebooklm.google.com. \>\> Aaron, welcome to the show. \>\> Hey, thanks for having me. \>\> Thanks for coming on. So, for people who don't know, you are the CEO of Box. Uh, you are a longtime ex poster, P O A S T E R. Um, and you have turned yourself into, I think, a really interesting thinker on AI, which is not surprising, but I think, uh, for someone running running a big company, it seems like you've really gotten your hands in it and really understand it in a deep way, which is awesome. \>\> Yeah. No. Um, I I think my PO days are uh a little bit behind me, but um I've uh I I'm I'm handing that off to the next generation. Um, and now I'm just, you know, just going to tweet about AI agents uh till uh uh till till the the whole uh the whole thing's over. But um yeah, so it's it's been a super exciting journey. Um I think for all of us that have been deep in the space um probably the most exciting technology in and at least my my lifetime um and uh and so just having a lot of fun with it. So, one of the one of the places I want to start is I feel like you have a pretty particular perspective on why jobs aren't going away because of AI agents. Do you want to talk about that? \>\> Yeah. So, um you know, I I still leave open a 5% chance that I'm like totally obscenely wrong. Um \>\> as you should. \>\> Yeah. So, so this is sort of like a high confidence with with room for for debate and and you know some some internal doubts here and there but for the most part um and I think it's sort of somewhat empirical when you use these tools um uh I'm a big believer in and you know thousands of people have talked about this at this point but but you know jobs are not tasks jobs are a collection of tasks and AI is very good at automating tasks and uh obviously that that the the definition of a task is expanding dramatically based on now what agents can do. But at the end of the day, there still is, you know, we still need a human to incorporate whatever the task that was executed into a broader workflow into a broader business process into actually, you know, some form of value creation. And so um because you you can't ever get rid of that person um that means we eventually will still have some degree of specialization of of of what people you know end up owning from a you know a completion of all the relevant tasks in their in their domain. Um and so you know let's just go through the list. Um you know even uh you know when an engineer needs to uh develop some some form of software they're going to go to an AI agent and and have it go work on part of the codebase. But then they're going to have to make a decision of do I ship that feature? Do I like that code? Do I, you know, I have to go talk to a product manager to make sure it's the right thing? Um, I have to incorporate into a broader project or a broader system. And so no matter what, you're still going to need people for all of that. So, so that's that's sort of, you know, why why jobs as a as a general matter don't really go away. Then the question is, so what do you do if if you know work you can, you know, output 20 or 50 or 100% more in any given job? a lawyer can review double the amount of, you know, legal briefings. An engineer can can generate, you know, twox the amount of code or 5x the amount of code. Well, shouldn't that that sort of, you know, reduce jobs in in some kind of commensurate way? And my view on that one is really just I think actually it turns out that that we're we are uh doing way less work than what is actually economically useful. Um, and we are merely just constrained by how much time we have in the day and the cost of labor to be able to go and do that work. And um, you know, and so I look at examples throughout our own company. And if I could have uh lawyers go and review legal, you know, documents and and contracts two times faster, I would not have half the number of lawyers. I would actually the throughput of our of that particular bottleneck in the organization would just go way faster and it would it would be reduced. And so we would have we would employ the same number of lawyers, but we would be we would be reviewing all of our contracts at 2x the speed, which would actually probably in most cases mean that that there's some feedback loop where we're growing sales incrementally faster and we're getting back to customers, you know, more quickly. So we have higher kind of customer satisfaction rates that would actually even in some cases lead to more revenue or or better business results which ironically then would lead to hiring more people in those functions that that could drive more growth. So similarly in product uh in in engineering if if we could ship two times the amount of code what's likely going to happen is we're just going to expand the footprint of our features. We're going to build even more software that will then create even new that will create newer bottlenecks in the organization that cause us to hire more people. in the areas that now are um are are involved in that. And so in most cases I'm I'm I'm more finding that um that that we are just actually not at uh capacity and we have not reached the point of of supply demand equilibrium where where we are doing just the perfect amount of work in the economy in any in any given function. If we can make it more efficient, we'll actually do more of it. And that that kind of ties to the the Jevans paradox idea. Javin's paradox, you know, I think mostly is applied to inanimate, you know, kind of resources and and, you know, AI systems or or, you know, rail uh, you know, trains and and, you know, energy consumption, but you could actually apply it to people too. And the idea would be if you could have a lawyer that could do 30% you know more output or 50% more output of again legal reviews um the demand actually would go up for for legal work because it becomes incrementally cheaper to then go and do that legal work which means you open up a new trunch of use cases for that kind of work to get executed. So so I just think all of the evidence right now is pointing more toward we're just going to do more. We're going to ship more. We're going to you know better serve customers. We're going to have more marketing campaigns. We're going to build more software. We're going to get better health care. We're going to, you know, have more tutoring and education. But all of those things still then drive jobs in the economy. \>\> I agree with you. I think the lawyer example is such a good one. I have caught like hundreds of thousands of dollars worth of legal mistakes just by putting my contracts into GBT5. Um, and I couldn't have hired a lawyer for that before because it would have taken too much time. And I already had a lawyer draft the contracts and they're the ones that made the mistake. So like the second level of legal review is a thing I couldn't have paid for before but now is a job that either a lawyer assisted with an AI or a legal firm could offer via you know CHBT or whatever or CHBT just offers that now in a way that I couldn't have bought before. \>\> Yeah. I think I think if you think about it like what are all the things today that you're not doing because the the price of entry to doing that thing is I have to hire a person or go and pay and procure us an external service like you know the the the minimum amount of money that you can that that you can spend just to even start to talk to a lawyer is thousands of dollars right the minimum amount of money you can spend to prototype an idea with an ad agency is tens of thousands of dollars so what happens in a world where you go and do that for $5 or $100 or $1,000, you're just going to do way more of those types of activities. And then again, kind of ironically, by doing more of those activities, you might actually find then these scenarios where you want to now bring an expert back into that workflow that you wouldn't have had before. So there are areas within our company where we start to do something purely as a as a test case or a prototype or or just you know kind of an ideation with an AI system and it works just enough that we say now let's actually go do this in the real way and let's go you know pay somebody or hire somebody or put somebody on that project but we wouldn't have even got it started before if AI didn't exist because we would not have we would not have have even thought that we should go you know sort of light up that project if we couldn't have prototyped it in the first place. And so this is sort of the the part that that again no economist has any any way to to estimate how much of the economy is going to grow as a result of that. It's impossible for your brains to kind of get around well how many new things get lit up because AI lowered the barrier to entry to cause then more people to get involved in that particular you know task. But that's actually going to be probably a substantial amount of work in the future. \>\> How do you think about the future? So, you know, you're someone uh you're running a big company. Uh you're running a a company that came up in the cloud era. Uh seeing a new technology wave come through. There's probably a maybe like a little bit of a sickening moment like, "Oh my god, am I going to have to change everything or how does this affect my business, right?" Um and your job is to figure out where the future's going and to and to start to understand, okay, is this going to be good for us? Is this going to be bad for us? How is it going to change jobs? All that all that kind of stuff. and and you're approaching it from a perspective that seems to, you know, you seem to have a pretty informed perspective on where it's going and I'm curious like how you form that yourself. How do you how do you go through all the possibilities to understand what's coming next? \>\> Yeah. I mean, I think um so to some extent I you know I'm kind of working through analogy and working through the the the fact that I've seen you know a couple of these major shifts and so you have that you have that as a as a background um uh you know kind of experience that informs a lot of what's going to happen next and everyone kind of can debate you know is you know what what are the most sort of relevant platform shifts that we've experienced that then AI relates to but at a minimum you can kind Think about it as okay, we went from the mainframe to the PC. We went from PC to mobile. We went from on-prem to cloud. Like these are these are platform level shifts. We kind of know how they work. You you have the early adopters start to play around with new technologies. They adopt these tools and then and then there are breakthrough use cases that cause um you know more of the kind of mainstream pragmatic you know buyers of technology to adopt those. Then that kind of you know sort of accelerates and then you eventually have the lagards. We we see this pattern every single time a new technology emerges and it it happens at the big macro technologies like cloud and and mobile and then it happens at the micro technologies like you know any any subservice in in cloud or mobile you know experiences the same thing. So so AI actually has a very similar curve. It's it's going through the exact same sort of you know typical bell curve of of adoption patterns. One thing that's different is it's happening in a compressed fashion. So where cloud may have taken you know 10 years to reach you know um complete mainstream adoption by every company that that is relevant AI is probably doing that in two years. Um but each individual of technology within AI still has again a very similar curve. So um obviously you know those of us spending too much time on X you know we're seeing AI agents encoding before the rest of the world but you kind of know exactly what's going to happen next over the next you know two years. everybody's going to adopt AI coding agents. It's just a guarantee because because the efficiency and the productivity gain is so massive that this will ripple through the economy. Um and so so I think by having a lens into both what the prior trends have been and just by being a very active user of these technologies myself I I can kind of see you know where where you know things are such a breakthrough that they will most likely again ripple through through the economy versus which things are maybe more incremental and so maybe it won't won't be that that impactful um uh you know that that that ends up helping inform this and then I think maybe the third factor is because we were a startup right at the early days the cloud. I kind of, you know, I felt that shift in the in, you know, deeply in the inside of a company that had to go through um executing on a a large technology shift that was happening. And so to some extent, I'm kind of pulling from those memories as much as possible and saying, let's I kind of have to do that again. Now there's obviously more people, you know, we we have we have new risks, we have new opportunities. Um, but it's very much like hardcore startup mode of, you know, I often just ask myself like quite literally what would we do if we were starting the company from scratch and it was just, you know, 10 people, how would we operate? What would how would we execute? What would we be building? What features would we be creating? And so if if we were starting the company over in an AI first world, what would that look like? And so again, you know, I'm benefiting from the fact that I I I saw that, you know, front row seat on the cloud wave and and we're trying to again that that's informing me and and informing the company, I think, on what what this should look like. So what are some specific things that you remember from that cloud wave? Because I think um you you know the the term digital transformation was the big was the big thing you know like 10 years ago and everyone everyone needed a digital transformation strategy a cloud strategy and some people probably did it well and but a lot of people I think they just know they need to say those words and and there's a big difference between companies that say the words and are like yeah we have a cloud strategy and companies that actually ended up effectively doing it and I'm curious like what your memories are of uh of how that works and who who does it well and who doesn't and then how you think that applies in this era. \>\> Yeah. Yeah, you know what's interesting is um this this is going to be much bigger than that because in the in the cloud transformation and the digital transformation f first of all it was a little bit um it was always a little bit um abstract as to and I think you're kind of getting at that in the question like it was always a little bit of an abstract concept of like you know when United Airlines does digital transformation what does that what does that really mean that means that they should probably have like a really good mobile app a really good website, you know, the customer support should be should be intelligent and kind of relevant to you. But at the end of the day, and with all respect to United Airlines, if you look at your flying experience from 15 years ago to today, like basically nothing changed. Um, so pre and post digital, \>\> it's probably a little worse. \>\> It might be it might be worse. Um, so, so, so the reality is is that for as much work that that as kind of went into that digital transformation process and and I'm sure that behind the scenes, lots of lots of really interesting technology got used, lots of of um, you know, new ways that they're operating in their data centers changed. The the actual day-to-day experience as a customer did not meaningfully change, meaningfully jump, you know, probably some better designed website, etc. And I think that's kind of probably felt across a significant portion of the economy pre and post digital transformation. Now there's more extreme examples. So if you look at Disney as just a random example, I think they would probably say, well, we had to become a digital company in the form of our product has to be now fully digital. We can't rely on movies on the the movie theaters. We're going to go to direct. And that that's that's a more significant business model disruption that that uh that had to occur. probably some banks maybe land in that category. But but you know you were kind of on this continuum in in the AI transformation. The reason why this is going to be so different and so much more impactful is it it is it changes how every single employer in your company operates. Again, the the daily experience of that United Airlines employee 15 years ago to today, their tools are a little bit more modern. They're, you know, they're probably getting real-time data feeds where it used to be a little bit more asynchronous. uh that that was that was sort of a contained level of transformation in your daily experience as that employee. With AI, there's there's not going to be any going back to the way things used to be and how we worked. It's just not possible because the efficiency gain between the company that uses AI versus the one that doesn't is just is just too insurmountable to try and make up for it if you're not if you're not using these technologies. And the way that we will work at the end of that transition will be so different that that it will it will you'll just fundamentally feel it again in in your daily work in your daily tasks. And so so maybe that would be the biggest difference is between maybe digital transformation and and you know this sort of you know AI era that we're entering is is the way that we work is going to be so fundamentally altered that um that that you will again just experience this in your in your daily life as an employee in any one of these you know companies. some jobs will entirely change and be entirely shifted and then other jobs again the daily activities will just be so different. So let's take engineering you know space obviously that we're we're following if you if you talk to a you know very kind of you know clued in online engineer right now and you say how are you developing software today versus even one year ago. It is the probably the biggest shift in any period in history of almost any knowledge worker job that's ever occurred, right? Like one year ago, you were typing into an IDE, maybe you were having some autocomplete, you know, technology like GitHub copilot, maybe you were asking a question of an AI system, getting some suggestion back. That was that was sort of one year ago. Obviously, three years ago, none of that even existed. Today, you're you're prompting an agent that's going to go off and do a large amount of work. it's going to come back with that work product and you're going to go and review it. That is like a completely different job than you know within a one-year period. If you even if not every job changes as much as that, if you kind of look at how that's going to ripple through knowledge work and you apply that to almost every form of knowledge work, it will mean that that every all of our daily workflows, if you're generating marketing assets and building marketing campaigns, if you are in sales and you're in and you're, you know, supporting a customer, um if you're in, you know, research and life sciences, every single one of our jobs is going to look completely different in the next, let's say, five, you know, plus or minus years. Um and that will be why it's so different than let's say even digital transformation was. \>\> So do you think then the better metaphor or a more apt analogy is uh like the shift to using computers at all like when we first started using Visical and spreadsheets or something like that. I I think I would be it would rank at that level um uh in terms of the amount of change, right? So the paper to digital process uh is was a a fundamental form factor change in in how you worked, right? Everything about the the workflow of a company, you know, well, it's probably even more significant. It's probably from paper to to digital plus the internet. Um uh because you know I think what we what we you know we we sort of did this skumorphic thing in the first in the first phase where we just kind of took the the paperbased you know desk workers set of tools and we put them into a digital screen. That was a shift but but but it was then an even bigger shift once you could connect those systems you could collaborate in real time. So we're kind of compressing that level of shift in a again a one or two year period. Um but uh but very much akin to that even the even the shift from kind of on-prem to cloud. It it it sort of was uh was more impacted the aesthetics of software and it impacted the the fact that you know when I chatted with you you got the you know you got the response faster and and it was queued up differently and the way we collaborated was we didn't use version control we just you know worked together in real time. That was a very big deal. But we already kind of understood the the general structure of how we would work together and how we would communicate together kind of pre and postcloud. the the shift from pre and post AI is again fundamentally different because you know what what I think we're seeing is is that the the job of an in uh individual contributor really begins to change because you are now a manager of agents and that is a completely change in the kind of work that you do and and we don't again we that that's a that's a very different level step function shift than what we've seen previously \>\> 100%. Um yeah, the the way I've been writing about it for a couple years is um thinking about us moving from a knowledge economy to an allocation economy where your job is to allocate intelligence. Um even as a even as an IC. Um but that's that's a key point, right? So so managers jobs are really about prioritization. They're about allocation. They're about uh using judgment um a ac across a set of of of kind of tasks and projects that are happening and and that that effectively becomes the new IC job in the future. \>\> Totally. Um one thing I want to get into though is uh you said you're in startup mode and you've been running Box for a while. Um and I'm curious what that has been like for you. Are you like, you seem energetic, but are you were you like, "Oh, \[ \_\_ \] I can't believe I have to go back to startup mode." Or were you like, "Uh, this is great. Finally, I get to, you know, feel like I'm back to the ground floor again or some mix." And you can be honest. This is safe space, but I'm really curious. \>\> Yeah. Um, okay. I I'll try and be as honest as as as possible on this one. Um, so the uh I'm going have to be introspective for 5 seconds. So, I'd say it's I'd say it's 80 to 90% um very excited, 10% 20% anxiety. So, so on the 80 90% um I I just I I love technology so much. Uh and I mean, you know, you sort of ha you have to be, you know, to to be in this industry, you know, um and to be, you know, trying to build a company for for as long as we've been kind of working on this. So, so I would say that the this is sort of me in my happy happy place. Uh, which is some something is is you know some some major technology event is happening and you can get your hands on it and and you know ideally it impacts you in some way that you can make it make sure that that you have to respond to something and do something like that kind of appeals to my ADD you know instincts. There was probably a period, you know, three or four years ago where I was like, "Huh, maybe I should start to pick up some hobbies. Industry was kind of settling down. We kind of knew that like all the different landscape, we knew cloud, we knew mobile, we knew how everything was going to work." Um, and so this is like way better because at 10 p.m. instead of doing some arbitrary hobby, uh, I'm on a, you know, Zoom call with somebody about AI or playing around with a new feature or building something. And that that is like that is so much better for me from an emotional, you know, excitement standpoint. \>\> I don't know. I mean, I feel like the world needs more of your pottery or woodworking. Yes, I uh I the at the very tail end of tail end of COVID I I started picking up guitar just because like I had I had some free time and u uh and like that's clearly a sign that like the industry had was was kind of cresting and there was wasn't a lot of change going on if I have time to to learn guitar. Um and uh and now now it's just yeah we're we're in kind of full full crank mode and it's incredibly exciting because something changes every single day that you have to respond to. Um and that that's definitely you know exactly what I where I like to be. \>\> We are so back. We're so back. \>\> We're so back. Yes. \>\> Uh but one thing that I think you've done really impressively is you were one of the first CEOs to really start the we're a AI first wave. like you sent the memo and you were like, you know, we're we're transitioning to being a company that takes this really really seriously and if you're not in, you're out. And I'm curious, I think a lot of companies are thinking about doing this right now or trying to do it with very varying levels of success. And I'm curious how that has gone for you and what you've learned about the way to do this well and to really be able to start from ground zero inside of an ex existing company. Like that's so \[ \_\_ \] hard. So yeah, I'm curious uh what you've learned. \>\> Yeah, it's extremely hard and and I would uh I would say um still still totally on that journey and we are not um you know we're not yet uh I can't you know kind of put the mission accomplished flag in the ground and and say that like we are the case study. Uh we're we are cranking on this every single day. Um the you know few quick lessons. One thing that we tried to do was was just be very clear that this is not about replacing jobs or or spending less money as a company. This is this is purely about how do we get output to increase? How do we move faster as a company? How do we do more? And how do we better serve customers? So, the first thing that I wanted to do is just make sure that that this was not some threatening technology. Um, but this is something actually that that we should be on the the forefront of because it's going to let us work better. It's going to let us actually do more as a as an organization. That was kind of I think important to lay out up front. Um the next is just making sure that that everybody's using it every single day in some capacity and then constantly showing each other how we're all using it. Um so we do this thing every single Friday. It's our internal all hands. We have somebody demo uh how they're using in our case box AI for different use cases. So they created an agent to automate some sales workflow. They create create an agent to automate a compliance workflow. And so we want to constantly just have everybody learn from each other on on what you know what what this technology is, how it works, how it helps uh in your in your daily work and then how you can go off and and do it yourself. Um you know we we've we haven't quite systematized this but but I think increasingly I'm I'm at least asking and I'm hearing more people ask hey why can't we do that faster? um you know you look at a project timeline that comes back and it's 3 weeks or you know four weeks and you say well you know I don't know why we can't do that in two days like like if you really just thought about it from a first principal standpoint you're really just trying to create this thing or build that like why can't we dramatically compress that timeline and that's causing people to say okay maybe actually I should relook at this maybe there is a technology out there that uh that we can go and leverage to make that happen um and uh and so that's starting to kind of create a a flywheel but I think again the end result is companies are just going to move far faster. They're going to get way more done. They're going to be able to better serve their customers as a result of this. And I'm seeing plenty of examples of all of that happening. The one asterisk that I'll say is the one thing we can't yet do, um, and this is maybe me being defensive, maybe we actually could if if I was just like so like burn the bridges kind of right now. Um when I talk to five or 10 or 20 person startups with no existing process with complete you know clean sheet of paper and how they operate uh I am seeing them be so differently wired than than you can be once you have existing workflows or processes that it is causing me to think like you know do I have to start to maybe go and find you know areas where from a completely fresh start we go and re-engineer something and it's it's because these startups start with again nothing that they can kind of think about their engineering workflows more in this modern way which is the workflow is actually you're prompting an agent, it's operating in the background, you're reviewing the code of that agent, you're you're very documentation driven, you're very spec driven, you're prompt driven. Um, and then you're letting the agent go off and and do lots of the work and then you're reviewing all of that. Um, and that that you can afford to do when you're not putting AI into an existing workflow, but when you're again kind of reinventing the workflow from scratch. And I do think there's areas where I want to do that much more in the organization. Yeah, I mean we have that we run four software products internally and we're 15 people and I commit code to those which is I should not be able to um and yeah totally would be totally impossible like without cloud code and codec cli and all those kinds of things and are it just totally changes the engineering process because it's yeah it's about the plan and if the plan's up to date and who's reviewing the plan and what work has been done and and the actual code doesn't matter as much. \>\> Yeah. And I think the, you know, when you have years and years of of highly, you know, tuned tribal knowledge on how to build things and, you know, in-person code reviews and all of the internal workflows. Um, you know, it's sort of harder to do the let's just start let's invent this whole thing from scratch. But we we are going through that journey. We will the way we build software will, you know, already looks very different than it did two years ago. And I I think it'll look vastly more different in a year from now than the combination of of you know the past, you know, couple years. \>\> Yeah, I'm totally with you. One thing that's been on my mind is uh and I I agree on the there's way more demand than than can be served and you're generally just going to want to do more work as a company. You know, we've been in a nonrecessionary environment for a long long long time. I'm curious, do you think that changes in a recession? I think it changes but but again we don't know the counterfactual where where usually in a recession you you unfortunately have job reduction regardless. So so now you you would probably still have job reduction um in a recession but but the companies are still able to drive more output because they can again they get more leverage from AI. So, so you know, maybe optimistically I'm totally making this up. Maybe you get out of the recession faster because because you haven't totally decimated your your productivity levels as an organization. Um, but but I I I would say that that you know, I could totally contemplate a scenario where you have a very bad economic environment that would lead to job reductions as it as it kind of always has. Unfortunately, that would right now probably be blamed on AI because that would be you'd still see you'd still see people doing work with AI. Um when again it's one of these it's a counterfactual which is well again in a recession you know prior to AI you also had unfortunately job cuts that you have to make in those situations. So you can't really know what would have happened in the nonAI scenario. So I do think we have to kind of watch out for that. But but the but I I really think about this as again I'm not seeing any evidence to the to the contrary. I'm I think about this as as just the next kind of era of of knowledge, work, technology that we we have always had these boosts in capability and productivity and um and you you sort of you when you're in the moment of that transformation, it's easy to kind of look at it kind of myopically and say, well, oh my gosh, this is going to be, you know, this is going to totally reduce the number of jobs in this area or impact us in this area. And then you look at it 10, 20, 30 years later and you realize, wow, actually it turned out the the the demand for that type of work was way bigger than we had imagined. And if we had never made it more efficient, we would never have actually gone and been able to to capitalize on that. I mean, if you just look at like I'm I'm I'm totally this is fanfiction, but like if you look at probably, you know, what a graphic designer 30 or 40 years ago would have said when they saw Photoshop, right? and you're like, "Wait a second. Right now, this project takes like a week to go and and design this poster for this client and you're going to make it take five hours." Well, you know, h how is it not going to reduce graphic design jobs by, you know, 10x? And today we have vastly more graphic designers in the world than we did, you know, 30 or 40 years ago. or um you know you look at at you know all of the you know stories of accounting when we went from um you know any kind of paperbased uh you know kind of you know methods to to the PC and Excel and and Visyal and into it and we have way more accountants today than ever before. So what is it about digitization that actually causes increases in in these jobs? It's because we finally make the function efficient enough that way more people can actually go acquire those services and AI is is again for everything that I'm seeing is just going to do the same thing again for a number of fields. \>\> Hoping that um uh the the one negative impact of technology is more accountants than ever before. So hoping that doesn't continue. love my accountant but \>\> the well one thing's for sure like you can imagine you know again in your example of AI reviewing your contract now imagine when you start doing that to kind of everything in your organization from a legal standpoint like you're going to be hiring way more lawyers uh as a result because you're going to say oh I found this thing now I have to go talk to somebody and and you're still it's still going to be bottlenecked by that human so so I I don't see a lot of these like in these large job categories getting reduced What do you think is overhyped in AI right now? \>\> Uh this will this is going to you know certainly show my my uh bias on this topic. Um I I don't know if I can think of something. Um uh uh I think I think if we if we look at where we are uh and this is again this is maybe one of the examples of this is one of the examples of of having been through the cloud wave. I could I and again I might extrapolate too much but I remember 15 years ago you know being like gosh I can't imagine this ex SAS category getting 10 times larger or five times larger and you zoom out and you're like oh my god like we were actually like we were just at the very small period of the of the curve at that point and like I I I was like in shock that AWS was still you know doubling 10 years ago and you know it's probably you know two orders of magnitude bigger uh today than it was. And and and and so I think there I don't know of a category where if I look at it, I say it's not going to be 10 times larger in in five years from now. I don't I don't think I can find that. What would what would you what would you argue? \>\> What do I think is the the most overhyped thing in AI right now? I mean, I think that we just uh people tend to o just generally with with AI, they tend to oscillate between like it's utopia and we're going to like everything's going to be solved and and you know it's uh free room service and teleportation for everybody. \>\> Um or we're all going to die. Um, and uh, that's why I like talking to people like you, cuz I think you have a more grounded perspective on well, it's going to change it's going to change a lot of the way that we work and also the world is going to continue more or less, you know, like we're still going to have problems. The bottlenecks just move somewhere else. And I think that's actually a much more interesting perspective and much more a much better way to talk about the future. I'm I'm totally fine with the utopian people assuming they're they're you know driving positive progress in the technology on that. Um I don't think I I I don't think you end up in this scenario of of that that probably people imagine on that front simply because um every step along the way some problem emerges that the AI is not good enough to handle that humans have to play the kind of stop gap on and that I think is a rolling process as far as far as the eye can see and um and I I think that each new technology breakthrough just leads to a new bottle neck somewhere else that people have to kind of you know we we play the role of of the duct tape on and and you know uh I I think that's probably actually been the story of of technological progress for you know 150 years is we thought we were going to automate x function and and we we we kind of did by 80% and then people have to do the rest um and um and I don't I again I don't see AI meaningfully changing that. Um there's just so much and this is this is sort of back to this thing of like why why does the engineering job still exist in the future etc. There's so much signal and so much context that you get by still operating in the outside world that is necessary for these AI systems to know about but they can't glean on their own. And um and and there's no there's no breakthrough uh that that we have any example of that that replaces that. you know, my ability to to talk to a person down the hall that gives us that gives me an idea because they just talked to a customer. We can't replicate that in an AI system right now. We don't know how. Um, and and there's no technological breakthrough that we know of that will replicate that. So, as long as as there's still a three-dimensional world out there that we have to go and participate in, we're going to have much more signal, much more context than than the AI will. Um, and that's going to just keep us keep us in jobs, keeps us doing things again for for as long as we can we can kind of look out right now. \>\> Yeah, I agree with that. I think um even if we get there and you know people are working on robotics and continue learning and all that kind of stuff, it's easy to forget that there are all these weird like people collect experiences and you learn from experiences. And so if you've spent a long time in an industry, for example, for the last like 15 years, you have a lot of this like tacit knowledge in your neural networks that you can't you you get feelings about, but uh and and even if we have an AGI, if it hasn't been there, it's not going to have the same perspective. Maybe it has an interesting perspective that's useful, but there's still just your own personal experience thinking about you versus other humans is ma matters. \>\> Yes. And and I don't think we want I don't think it'd like be that fun if you have to just prompt engineer everything in your life. Like like if I have to if you had to prompt me when we're getting on this podcast and like okay you were doing a podcast right now and here is what you're going to it's like no that wouldn't it would be like it would be so draining that that we just we would completely halt the you know all of our all of our interaction. So, so you do need like but people don't have to be prompted in that way constantly, right? We can kind of pick up on the cue that lets us be like, okay, I can put this in this part of my memory and I'm not going to, you know, I'm not going to accidentally talk to you about healthcare right now because that's not what we're talking about. And and so you like like people are just going to be better at that in until you know you you kind of look at the you you uh if you watch the Richard Sutton you know um uh you know podcast um with Doresh like it's it's kind of it's exactly that which is you know these systems are not they they do not have any context they they don't have they don't have any ability to truly predict the future. Their next token prediction it's insanely valuable. There's an incredible amount of economic value in that, but they don't replace people because people we can go around the world and we can build a tremendous amount of context that the AI will never be able to get and no amount of humanoid robots in the world uh will be able to still replicate that and we will deploy these as as uh utilities for us as again we always have. We we we continue as a as a species to be able to deploy technology as a utility so we can get better things in life, better healthcare, you know, better life expectancy, better food production, uh you know, better entertainment and and I I think this is again one of those technologies. \>\> Why do you think that more AI CEOs don't talk like this? Uh when you say AI CEOs, are you saying like Daario or \>\> Yeah, I think a lot of a lot of the CEOs of the big labs Yeah. don't have this don't talk publicly with this perspective. \>\> I think Well, there's a reason I chose B2B software. Um I'm probably I'm probably on the more boring end of of uh of of this ecosystem. Um \>\> I've read your Twitter. You're not on the boring end. \>\> Okay. But but like you know there's there's a reason that I landed in enterprise software and not building an AI lab. So um so I I live in in reality with the practical you know implications and limitations of these of these tools. Um I actually don't mind that you know Sam Alman or Daario or others you know either talk in the way they do or or have the ambition that they in the way that they do. I I believe by the time the technology hits the real world, it will manifest just a bit differently and so thus the implications are a bit different. Um but but I I think it's kind of fun that we have different approaches to this. I think it would be actually very boring if if everybody was like hyper pragmatic and and practical. Um you know, I think on the margin I' I'd rather have the more sort of you know, this is going to be a crazy utopian future than the dystopian, you know, angle that that some take. Um but again, I think it's cool. we have a marketplace of ideas. There's lots of different opinions. Um, but you know, when when you see things like, let's say, Daario says, you know, we're going to see this massive, you know, job dislocation or or 50% of jobs or I forget the exact stat, so I don't want to misattribute um, you know, what what he said. Um I think the thing that that that that just you know where where where it ends up being a little bit different than that point of view is again these jobs are a collection of these tasks and and we have figured out how to automate tasks and the tasks are getting bigger but there still is an end point where a person has to come in intervene review something and execute something and even in the cases and again this is where this is where I I'm informed by having now you know box has a few thousand employees So I'm informed by this, you know, as a result of that. But but but even in the places where we say, you know what, we probably could have, you know, a smaller total number of folks in the company doing password reset emails because we can just automate that. The way as as a as a resource allocator that I respond to that though is we put more resources in a different area of customer success that has always been underfunded chronically because we didn't have enough budget to go apply there. And so if I can make one job function more efficient, I I I gladly will take those dollars and reapply them into an area that is more strategic that we have not been able to automate and that I see no sort of plans to be able to automate. and and that's the much more dynamic nature of these organizations and of the economy generally um that I think sometimes doesn't get sort of brought into the you know big economic dislocation conversation. \>\> How are you splitting your time and your focus between I have these two buckets that I I try to think about. So one is just um removing the current bottleneck for for any product I'm working on or any company I'm working on. It's like we have a funnel like where is the biggest bottleneck in the funnel and like how do we make that better? How do you split your time between that and like magic moments? It's like, wow, we have there's this new technology that does this crazy wizard \[ \_\_ \] and we can do something fundamentally new here. And how do you how do you split your time and attention and your company's time and attention between those two things? \>\> And to clarify, do you mean internally operationally or for the product experiences we're building? \>\> I think both because I I assume the internal operations like lead to the building like magic magical product stuff. \>\> Yeah. So sometimes they they are they they can be decoupled but but um I I think again as a pragmatist um probably 80% of the time is going into the incremental bottleneck that we can deblock. Um even from a product roadmap standpoint if you look at the AI that we are delivering a lot of it is just hyper practical like you have a lot of contracts you have a lot of invoices you have a lot of research papers I want to extract data from that so I can put it into a database to automate a workflow that's where we're putting a lot of our energy because that's just a major pain point in the economy it turns out there's like trillions of documents that all have important data in them that you would like to be able to extract automatically and we could then power you know workflows more efficiently as a result of that. That that's where we spend a lot of our time. Equally though we have a number of initiatives not a number we have a small number of initiatives that are much more like okay how do I go and automate the entire workflow of of you know generating a a loan agreement or um or of doing a due diligence on an M&A you know um uh transaction and that's a longunning agent. It's going to it's going to do you know hours worth of work. It's going to read lots of documents. It's going to collate them. It's going to generate a report. Um, and that's much more like a 10xstep function change in how that workflow works today. But again, we're going to pay the bills because we're going to do a very practical automation uh along the way. And so, you know, I think internally that's probably how we we look as well, which is 80% of our time is is let's let's reduce that bottleneck. Let's make that more efficient. Let's improve, you know, how we respond to customers there. And then 20% is experimenting saying, you know, what would this workflow look like if we had cloud code go and just run it as a as a background agent um and do a lot more of the work as opposed to a you know a um uh more of just a a you know a basic way of interacting with the agent in an ID. \>\> Tell me about this automated due diligence uh research agent. Like is it working? Like what have you learned from building it? Do you think it's going to work? \>\> Um still very early. Um and um it will work. um uh simply because as we're finding it's just all a trade-off on how much compute you want to apply to the problem. Um you know we we always we we have a lot of internal debates uh which are like you know like man we can make that thing happen in 10 seconds but we know that the hit rate is 50% or you know 7030 or we can make it happen in 1 minute and we know it'll be 93%. I'm making up all the numbers and and and then you know at at some point the customer is going to be like wow that wasn't very magical because it took a minute but you're like but you got the right answer and the customers obviously would be way happier if they were in the 70% success rate in 10 seconds but they're not going to be happy if they're in the in the 30%. So, so all of these things are just product trade-offs, which is how long is the customer willing to wait? Uh, do you give them the knobs to tune, you know, th those decisions? If you give them the knobs, can a regular person outside of Silicon Valley understand what those knobs even mean, or are they just now super confused and we're doing technose? Um, and so those are that's like I mean but it's it's so much fun that because I we have we have spent this I can guarantee we have spent more time on UX patterns that are completely um unprecedented UX patterns in the past year and a half than probably the past 20 years of building a company because because for the most part over the past 20 years software didn't really change again pre-post cloud like like we had buttons, we had tabs, we named the things, you could have a sidebar like like all of the UX patterns of software, you know, were exactly the same. We just got better at at designing pixels in the in the past kind of 10-15 years. Figma just, you know, made it so we could actually make everything look modern and have a good design system. But like nobody really reinvented like fundamental we, you know, there would drop down menus. there was there was a you know a you know you know windows and in inside of each other etc. In AI man like every single couple weeks we're like \[ \_\_ \] like like how should we expose the idea that this is going to be a longer running agent to to the person and how much should be anthropomorphized as what you would do when you're interacting with a human versus no this should be actually like kind of behind the scenes and it's just software doing stuff. Um, and and that actually obviously makes it so much fun because like we're designing a new form of software. The software is actually labor, you know, that you're interacting with in some form and that doesn't have the classic patterns of software. Uh, and so we get to invent a whole new style of how we build these tools. \>\> Mhm. One of the other interesting trade-offs there in on the product end is which model to use. Um, so you know, newer models, more expensive, usually faster, but like you get much better results. Older models less expensive, you know, maybe worse results. But there's always that there's always that trade-off there where it's like, yeah, we could serve this to everyone and it' be amazing, but it would bankrupt us. So, how do you think about that? \>\> Um, in general, right now, I'm in the in the uh you should just always be using the best that there is. Um and you know in our case we're we're you know fortunate enough where we we can we can afford to sort of say okay you know what we'll we'll spend a little bit less in that area to fund the subsid you know some some of the compute on this area we can move things around. I I'm I'm certainly sympathetic to smaller startups maybe that don't have you know venture in them. you can't make those decisions as easily. But generally speaking, right now, I think we're in a part of the curve where you kind of just want to always be betting on the better technology and um and and mostly because you will have a competition that does and you will not be able to be the company that has an inferior product right now. Um and and then equally any company that is doing uh work to mitigate the quality issues of a of an inferior model relative to what you could be getting from a better model that work is totally wasted relative to again you know real productive value creation. So, so you kind of and this is why it's, you know, you know, fortunately a lot of people are kind of building in public, but um so you get to kind of see these examples, but like this is why I think on a regular basis every 3 or 6 months, you're kind of like moving up the cur up the stack from a scaffolding standpoint because you built scaffolding two years ago that mitigated, you know, context window length and now that's not an issue. You know, for instance, we we built a lot of features. This is just to give you an example. We built not a lot. We built a couple services internally because chatbt GBD 3.5 had a context window of I don't know like 6,000 tokens or 8,000 tokens or whatever. Well, that scaffolding is irrelevant in a world of 100k tokens or 200k tokens, you know, effectively. And so, we could have like been like, no, let's, you know, let's really keep betting on that thing and the old model and but obviously you're like, no, we you should just instantly upgrade. Screw the software that you built. Um, now let's just benefit from the model's capabilities. you know, similarly as um as you have, you know, better reasoning capabilities, better multimodal modal experiences, more of these features will be compressed into one single model. And so, I think you kind of have to just bet on the best-in-class models that are out there. Unfortunately, cost be damned if you can if you can make it more efficient with like intent routing and whatnot on your own, great. Um, but but you cannot have you cannot afford to have an inferior experience on any dimension right now and be competitive based on how fast the space is moving. That's great. I totally agree. Um Aaron, this is a pleasure. Thank you so much for joining. \>\> Yeah, thanks for having me. \>\> Oh my gosh, folks, you absolutely positively have to smash that like button and subscribe to AI and I. Why? Because this show is the epitome of awesomeness. It's like finding a treasure chest in your backyard, but instead of gold, it's filled with pure unadulterated knowledge bombs about chat GPT. Every episode is a roller coaster of emotions, insights, and laughter that will leave you on the edge of your seat, craving for more. It's not just a show, it's a journey into the future with Dan Shipper as the captain of the spaceship. So, do yourself a favor, hit like, smash subscribe, and strap in for the ride of your life. And now without any further ado, let me just say Dan, I'm absolutely hopelessly in love with you.