AI for the Fortune 2000: How Tenex Builds an AI-Native SDLC — Alex Lieberman & CJ Hess

“Yeah, uh very simply, Tan X is an applied AI partner for some of the biggest companies in the world to help them go from AI apps into AI native.”

Claim: Tenex describes itself as an applied AI partner that helps large companies transition from using AI applications to becoming AI-native.

“Our belief is that AI is incredible, but it is raw intelligence and the hardest problem to solve is the final mile to actually diffuse it through the economy.”

Claim: Tenex believes AI is powerful raw intelligence, while the hardest challenge is diffusing it through the economy.

“And so our view is to... act as the player to help enterprises to figure all this can i curse on to figure all this shit out”

Claim: Tenex aims to help enterprises work out how to apply AI.

“um so i was actually the first hire i don't even know if If Alex and Arman had the name 10X quite yet,”

Claim: CJ says he was Tenex’s first hire and is unsure whether Alex and Arman had chosen the company’s name by then.

“but I have like a graveyard of startups I tried to start before 10X ultimately culminating in like my own kind of engineering consulting company and I spent some time there doing traditional engineering eventually moved over to doing some of this like AI consulting and just like building more and more with it.”

Claim: Before joining Tenex, CJ attempted several startups, ran an engineering consulting business, worked in traditional engineering, and later shifted to AI consulting and AI-enabled building.

“I'm a software engineer by trade and I just felt my output getting better and faster.”

Claim: CJ says that as he used AI, his software-engineering output improved and became faster.

“ultimately getting paid less like I was charging hourly to my clients and I really liked kind of the ethos of 10x that Alex had first posted when I saw it on Twitter about kind of looking to reward engineers for their output as opposed to rewarding them for their time I think you know fast forwarding two years you can see like how engineers can now parallelize a ton of work and just have a ton more output at the time with”

Claim: CJ says engineers can now parallelize substantially more work and produce substantially more output.

“ultimately getting paid less like I was charging hourly to my clients and I really liked kind of the ethos of 10x that Alex had first posted when I saw it on Twitter about kind of looking to reward engineers for their output as opposed to rewarding them for their time I think you know fast forwarding two years you can see like how engineers can now parallelize a ton of work and just have a ton more output at the time with”

Claim: CJ supports rewarding engineers according to their output rather than the time they work.

“But I basically took the bet that I think with my skill set and leaning more into AI, I could have much higher output and ultimately have like a higher earning capacity to do that.”

Claim: CJ believed that applying his skills more heavily to AI could substantially increase his output and earning capacity.

“And since then, I've just like fell in love with that model. I think that's like what the future of engineering really looks like. And ultimately.”

Claim: CJ believes that engineering’s future will resemble a model in which engineers use AI to increase their output.

“When AI is solving all these tech problems. companies are going to pay based on outcome and they're not going to want to pay based on the time that's required to build something.”

Claim: CJ predicts that as AI solves technology problems, companies will pay for outcomes instead of the time spent building solutions.

“Yeah, so for more detail on that people can refer back to our previous pod. We literally pay people by story points, and you just have to agree what the story points are. Uh but that also means that some 10x engineers can literally make millions of dollars a year and some others we may not make as much just because they're not as productive.”

Claim: Alex says Tenex compensates people according to story points that the parties agree on.

“and you just have to agree what the story points are. Uh but that also means that some 10x engineers can literally make millions of dollars a year and some others we may not make as much just because they're not as productive.”

Claim: At Tenex, some engineers may earn millions of dollars annually, while less productive engineers may earn less under its output-based compensation model.

“Yeah, exactly. Like the thesis is that AI is really just a leverage bet on a human's intellect and work ethic.”

Claim: Tenex views AI as leverage that amplifies an individual’s intellect and work ethic.

“And so you're just going to have more variance than ever before in earning capacity.”

Claim: Tenex predicts that AI will increase variation in individual earning capacity beyond previous levels.

“And our view is our model isn't for everyone, but for the person who basically wants to go along themselves, who has this incredible like foundation in software engineering, but also is constantly pushing themselves to leverage the frontier, to stretch their time as much as possible while having quality output, there's no better way to earn.”

Claim: Tenex believes its compensation model is especially rewarding for software engineers with strong foundations who continually leverage frontier technology, maximize their time, and maintain quality output, though the model is not for everyone.

“So so just to paint a picture, we started 10X 18 months ago when we started the business.”

Claim: Tenex says it began operating 18 months before this conversation.

“we started as like an AI powered dev shop.”

Claim: Tenex says it initially operated as an AI-powered software development shop.

“Really like the original idea was engineering is kind of like the canary in the coal mine for AI.”

Claim: Tenex’s original thesis was that software engineering would be an early indicator of AI’s effects on work.

“And part of the impetus was my co-founder Armand, who's technical by trade. was like starting to use coding agents a lot and then he had some people he knew that were still typing into their ID like not adopting the technology at all and so we're like why don't we just supercharged a company's engineering org who isn't adopting this so we started there then as we started moving up market from startups and like mid-market companies to now you know 50% of our customers are Fortune 2000 and then 50%”

Claim: Tenex says co-founder Armand began using coding agents extensively while some engineers he knew had not adopted coding-agent technology and continued typing code into their IDEs.

“was like starting to use coding agents a lot and then he had some people he knew that were still typing into their ID like not adopting the technology at all and so we're like why don't we just supercharged a company's engineering org who isn't adopting this so we started there then as we started moving up market from startups and like mid-market companies to now you know 50% of our customers are Fortune 2000 and then 50%”

Claim: At the time of the conversation, Tenex said 50% of its customers were Fortune 2000 companies and 50% were mid-market companies generating $100 million to $500 million in revenue.

“was like starting to use coding agents a lot and then he had some people he knew that were still typing into their ID like not adopting the technology at all and so we're like why don't we just supercharged a company's engineering org who isn't adopting this so we started there then as we started moving up market from startups and like mid-market companies to now you know 50% of our customers are Fortune 2000 and then 50%”

Claim: Tenex initially aimed to accelerate engineering teams at companies that had not adopted AI.

“than are mid-market doing 100 to 500 million in revenue they weren't just asking for basically staff aug and ai engineering they were asking for like ai transformation as a service today the business is at 75 people 90 of those employees are either engineers or ai strategists which are basically think of as like technical consultants and i think”

Claim: Tenex says larger customers sought AI transformation as a service, beyond staff augmentation and AI engineering.

“than are mid-market doing 100 to 500 million in revenue they weren't just asking for basically staff aug and ai engineering they were asking for like ai transformation as a service today the business is at 75 people 90 of those employees are either engineers or ai strategists which are basically think of as like technical consultants and i think”

Claim: At the time of the conversation, Tenex had 75 employees, including 90 people described as engineers or AI strategists.

“The thesis has proved out true, which is there's a very wide dispersion in earning capacity between the highest paid engineer at 10x and the lowest paid engineer at 10x. It's probably something around 4x difference.”

Claim: Tenex says the earning capacity of its highest-paid engineer is probably about four times that of its lowest-paid engineer.

“Not 10x difference.”

Claim: Tenex says the pay difference between its highest- and lowest-paid engineers is not 10x.

“There's been interesting learnings around our model of paying based on story points where there are limits to the model that we knew would be limits and how we think about that moving forward. But overall.”

Claim: Tenex says its story-point-based compensation model has limitations, including limitations the company had anticipated.

“I think the concerns that people had around this model of like paying people based on output, like the classic concerns are, well, story points are subjective.”

Claim: One concern about output-based compensation is that story-point assessments are subjective.

“People are going to either sandbag story points or they're just going to ship a bunch of bad code.”

Claim: Critics of output-based compensation worry that employees may inflate story-point estimates or produce poor-quality code.

“Actually, the biggest challenge has been how do you compensate engineers for all the time that's not spent coding, whether it's architecting,”

Claim: Tenex says its main compensation challenge is accounting for engineers’ non-coding work, including architecture.

“I was coding a lot, coding a lot. coding a little less and now we're kind of getting to the point where the models are more or less given the right context and can get you to 90% of the solution and I do think there is some opportunity here to think about like”

Claim: The speaker says that as models improved, they went from coding a lot to coding less, and models given the right context can now produce about 90% of a solution.

“coding a little less and now we're kind of getting to the point where the models are more or less given the right context and can get you to 90% of the solution and I do think there is some opportunity here to think about like Really, the value is sitting down with a customer, hearing everything that they can brain dump, and then just providing that to the model to actually get closer to the solution.”

Claim: The speaker sees an opportunity for engineers to gather customers’ requirements directly and provide them to a model so it can get closer to a solution.

“So I imagine, you know, in the future, we're going to start to look more at like true what this output is, how it works, how it looks, as opposed to, you know, kind of that traditional story point model, but still anchoring to those as the unit.”

Claim: The speaker expects the team to evaluate software work increasingly by its actual output and characteristics rather than relying on the traditional story-point model, while retaining story points as an anchor unit.

“So we basically started trying to kind of standardize a set of skills. I think most engineering orgs have done something similar by this point. And for us, that grew. into what is actually like a very structured set of documentation within a repo and then a CLI that's around it that Kind of the models are using to actually drive the work that's being done,”

Claim: At 10x, a standardized set of engineering skills developed into structured repository documentation and a CLI used by models to drive engineering work.

“whether it's creating a plan, a visualization, you know, creating some readout for a client, effectively like building all the pieces of the process around actually writing the code to just happen more automatically and standard across our team.”

Claim: At 10x, models can create plans, visualizations, and client readouts, while the surrounding software-development process is designed to make code-writing more automatic and consistent across the team.

“So we have a couple of things like one is this like big canonical plan document that has some similar structure, but. I'd say as the models have gotten better, we've actually reduced the enforcements there in the structure um kind of similarly across all of these. But once you have your plan, you actually build it out, and we have a very opinionated way on how agents will go and log their decisions and log what they're doing and just keep track of what they're up to,”

Claim: At 10x, improvements in models have led the team to reduce how strictly its canonical plan and related workflows enforce structure.

“I'd say as the models have gotten better, we've actually reduced the enforcements there in the structure um kind of similarly across all of these. But once you have your plan, you actually build it out, and we have a very opinionated way on how agents will go and log their decisions and log what they're doing and just keep track of what they're up to,”

Claim: 10x’s engineering workflow prescribes how agents record their decisions and ongoing activities.

“spit out. Um but all that gets checked in and then it serves as the surface where not only do you have the code as the source you know I think there's a debate on like do you even need docs like the code can actually be the memory of a code base. Um and I do think you do. You need to know why you made a decision, you need to know the context around the conversation that was had. Um but ultimately that plus the code base”

Claim: The speaker believes software repositories need documentation alongside code to preserve the reasons for decisions and the context behind them.

“If there's a question, if there's something we need to answer, if we need to look back on why we did something, we just always have that in some durable record and it's never like in someone's head. It's always in the repo.”

Claim: At 10x, information needed to answer questions or understand past decisions is kept in a durable repository record rather than relying on individual memory.

“So that just gives like our agents and our tools these superpowers to have all the context that I think a lot of engineering teams still just have in conversations or have in their heads.”

Claim: The speaker says 10x’s repository-based records give its agents and tools access to context that many engineering teams retain only in conversations or individual memory.

“And our goal is to kind of continue to push that forward so that. All the context is in one place and more and more of that work is just driven by the agent as opposed to us having to provide everything up front.”

Claim: 10x aims to centralize context and have agents perform an increasing share of the work without people supplying all information in advance.

“And one way we even measure like if any of these kind of like processes are working is are we seeing an increase in trend in average story points per month of every engineer in the company?”

Claim: 10x uses the trend in average monthly story points per engineer as one measure of whether its engineering processes are working.

“Um, I'd really say overall it's been like a two X increase this year in particular.”

Claim: The speaker estimates that overall engineering output at 10x has roughly doubled this year.

“I think basically if you think of December as the like Opus four five moment. where the models really started to take on these long running tasks to today we're back to Opus, but Opus 5.5. I mean, that's just a magnificent model that can really do a ton of work.”

Claim: The speaker says models began handling long-running tasks around December and describes Opus 5.5 as capable of doing a large amount of work.

“Over that period, I've seen like the actual output of engineers, I would say roughly double in terms of like what they were able to do to what they're able to cover, you know, in the same time span now.”

Claim: The speaker estimates that engineers’ output—what they can accomplish or cover in a given time—has roughly doubled over the period discussed.

“And know. that's roughly how we think about it. I would say at least once a quarter we kind of sit down and evaluate like what is our criteria for these story points? How are we actually estimating these projects?”

Claim: At least quarterly, 10x reviews its criteria for assigning story points and its project-estimation methods.

“There's a fine balance to be made between like still rewarding this output, but also As the models get better we do need to be able to kind of calibrate like the value that we're creating.”

Claim: The speaker believes 10x must balance rewarding engineering output with recalibrating how it measures the value created as models improve.

“So overall like I would say roughly that 2x has played a similar role in like how we do these estimations but I mean if I just look at some of the features and things people are building and pushing out now versus this time last year I mean it's it's insane what the difference has been.”

Claim: The speaker says the roughly twofold productivity increase has influenced 10x’s estimation practices, and describes the change in features built and shipped compared with a year earlier as enormous.

“Opus is so good, it's fast, it's cheap, it kinda hits all my check boxes. Um I would say I previously have had like two to three models depending on whether they're cheap and fast, whether they're really smart, like whether they can solve specific problems. And now Opus is kind of covering my my full range. Um I think a lot of the team still reaches for tools like codecs when we need to do something”

Claim: The speaker considers Opus very good, fast, and cheap, and says it now covers the full range of model needs for which they previously used two or three models.

“Opus is so good, it's fast, it's cheap, it kinda hits all my check boxes. Um I would say I previously have had like two to three models depending on whether they're cheap and fast, whether they're really smart, like whether they can solve specific problems. And now Opus is kind of covering my my full range. Um I think a lot of the team still reaches for tools like codecs when we need to do something to do something maybe more in the computer use world and really workflow-wise I think using both is still the best option,”

Claim: The speaker thinks using Opus and Codex together remains the best workflow, with Codex used for some computer-use tasks.

“whichever your driver is, which I think for most of our team is Opus 5.5.”

Claim: The speaker says Opus is the primary model driver for most of their team.

“Using a codex to review and having them actually work back and forth on your machine I think still produces the best results,”

Claim: The speaker believes that using Codex for review and having models work back and forth on their machine produces the best results.

“but right now we're pretty aligned on those two I don't think there's too much else to say.”

Claim: The speaker says their team is currently fairly aligned on using Opus and Codex.

“So yeah, they recently they recently launched like the ability to hook in your cloud code subscription.”

Claim: AMP recently launched a feature that lets users connect their Cloud Code subscription.

“So that's been a pretty big game changer for me.”

Claim: The speaker says AMP’s ability to connect a Cloud Code subscription has been a major game changer for them.

“Totally. And like, I mean, I love AMP because it takes everything off my computer.”

Claim: The speaker likes AMP because it moves their work off their computer.

“They have this like cloud setup, the orbs.”

Claim: AMP offers a cloud setup called orbs.

“it yeah but yeah that's my that's my interface of choice and”

Claim: The speaker says AMP is their preferred interface.

“Yeah, I think there's a piece of that I think on the amp side of them being very opinionated and not trying to build this general solution. Um tools like cloud code in the cloud and cloud tag I think provide value for teams that are really trying to like collaborate maybe on non-technical tasks but part of the reason I love AMP is that it has all these cross-model functionalities and I can do that Opus writes it”

Claim: The speaker likes AMP partly because it supports functionality across multiple models.

“provide value for teams that are really trying to like collaborate maybe on non-technical tasks but part of the reason I love AMP is that it has all these cross-model functionalities and I can do that Opus writes it you know, I can even hop to another orb and say, look at that orb and review it and all of those things still come together to basically replicate what I'm doing on my machine.”

Claim: The speaker says AMP lets them use Opus to write work and another orb to review it.

“you know, I can even hop to another orb and say, look at that orb and review it and all of those things still come together to basically replicate what I'm doing on my machine.”

Claim: The speaker says AMP’s combined workflow can replicate what they do on their own machine.

“And I just couldn't quite get there with, you know, a cloud code and the cloud, a cloud tag.”

Claim: The speaker says they could not quite reproduce their preferred workflow using Cloud Code in the cloud and Cloud Tag.

“I actually don't have a number off the top of my head. I'd say we're somewhere between, if we're talking like raw tokens, somewhere between one to one and maybe like two to one.”

Claim: The speaker estimates that their raw token spend is between one and two times their employee spend, while noting they do not have an exact figure.

“what we think about a lot is like, when to reach for the biggest model.”

Claim: The speaker says their team frequently considers when to use the largest model.

“I think every company is starting to think about that.”

Claim: The speaker believes companies broadly are beginning to consider when to use the largest model.

“But I will say recently our costs have been driven just way down from Opus 5.5,”

Claim: The speaker says their costs have recently fallen substantially.

“we Yeah. kind of had an era of like everyone was using Astra and Fable and they just gave the best results. But ultimately those models are expensive.”

Claim: The speaker says Astra and Fable once gave the best results, but were expensive.

“yeah i mean there's a few ways that i think about it is like we were for sure we were for sure worried about it just as like our headcount was growing and um i i think also like even though our team i would say like i don't feel like anyone on our team ever really ascribed to like token the token maxing zeitgeist because”

Claim: The speaker says no one on their team really embraced the token-maxing zeitgeist.

“yeah i mean there's a few ways that i think about it is like we were for sure we were for sure worried about it just as like our headcount was growing and um i i think also like even though our team i would say like i don't feel like anyone on our team ever really ascribed to like”

Claim: The speaker says their team was concerned about costs as its headcount increased.

“I think there's a couple ways like the future might play out there.”

Claim: The future of model use may develop in several different ways.

“If there's a world where there's a Astra 6-5, a Fable 5-5 and they're super expensive, but they get the best results, I could see us trending back that direction. But generally I agree that, you know, you look at something like Muse Wonder that they launched yesterday”

Claim: If Astra 6-5 and Fable 5-5 become very expensive models that deliver the best results, model usage could trend back toward those high-cost models.

“or today and it's on the frontier, it's dirt cheap, the competition is heating up and both an open source and closed source and”

Claim: Muse Wonder is a frontier model offered at very low cost, and competition among open-source and closed-source models is intensifying.

“I think that just will continue to drive costs down.”

Claim: Competition in AI models will continue to reduce costs.

“And when you really look at it, a lot of tasks that are economically valuable can be done with OpenSpy 5.”

Claim: Many economically valuable tasks can be performed using OpenSpy 5.

“So, you know, it's that cliff of like, would a 10 times smarter model really justify the cost of like moving some cells in a spreadsheet? Probably not.”

Claim: A model that is ten times more capable probably would not justify its additional cost for a simple task such as moving spreadsheet cells.

“So one way I think about this is I think about enterprises adopting AI as almost like these different like milestones in a race.”

Claim: Enterprise AI adoption can be understood as a sequence of milestones.

“first milestone in the enterprise AI race is just rolling out a general purpose LLM to your people. And I'd say of the, you know, let's call it 100 enterprises we're talking to today.”

Claim: The first milestone in enterprise AI adoption is deploying a general-purpose large language model for employees.

“90% have rolled out either Claude or Codex to their employees.”

Claim: Among roughly 100 enterprises the speaker's company is currently speaking with, about 90% have deployed Claude or Codex for employees.

“Then the next milestone is like actually teaching people how to use it, like training and enablement,”

Claim: After deploying a general-purpose LLM, the next enterprise AI adoption milestone is training and enabling employees to use it.

“which like never gets the love that it should get,”

Claim: Enterprise AI training and enablement receive less attention than they deserve.

“but is arguably more important or as important as the technology itself.”

Claim: Training and enablement for enterprise AI are arguably at least as important as the AI technology itself.

“I would say that's at like 60%.”

Claim: The speaker estimates that approximately 60% of companies have reached the enterprise AI training-and-enablement milestone.

“Yeah, like there's some group there of companies that are like, our people will just figure it out.”

Claim: Some companies expect their employees to learn to use AI tools without formal training.

“But then there's like a good chunk that is like, okay, we need to teach our people what skills are. We need to teach our people the features and functionality of these surfaces.”

Claim: Some companies believe they need to teach employees the skills and product functionality required to use AI tools.

“Then the next after training and enablement within a general purpose LLM is coding agents.”

Claim: After training and enabling employees to use general-purpose LLMs, the next enterprise AI adoption milestone is coding agents.

“I was asking a few people here like, what percentage of enterprises do you think have rolled out coding agents to the engine product org in their company? And most people are like, it has to be 70 or 80%. The number is closer to 50%. Like every other enterprise.”

Claim: Most people the speaker asked estimated that 70–80% of enterprises had deployed coding agents in their engineering product organizations; the speaker estimated the share at closer to 50%.

“that we speak to either hasn't rolled out a coding agent or what they are using is closer to not having a coding agent than having a coding agent then the next and this is where like the biggest I would say drop off in like is where like the biggest diffusion gap is from like raw intelligence to enterprise usage is agents in production for non-engineering work drops to like five percent”

Claim: The largest gap between AI capability and enterprise deployment is production use of agents for non-engineering work, which the speaker estimates occurs at about 5% of enterprises.

“that we speak to either hasn't rolled out a coding agent or what they are using is closer to not having a coding agent than having a coding agent then the next and this is where like the biggest I would say drop off in like is where like the biggest diffusion gap is from like raw intelligence to enterprise usage is agents in production for non-engineering work drops to like five percent”

Claim: Many enterprises the speaker's company speaks with either have not deployed a coding agent or use something that barely qualifies as one.

“Most enterprises that we talk to that have, let's just say, more than a thousand employees do not have an agent in production that is doing work that is not building software.”

Claim: Most enterprises with over 1,000 employees that the speaker's company speaks with have no production AI agent performing non-software-development work.

“One is there is because like. there's this this whole wave is driven by a combination of some small group of people understanding what is possible from this technology in a way larger group of people driven by FOMO and by the way FOMO isn't bad it creates some forward momentum but it also means that people will latch onto the things that are most in the zeitgeist I guess now for the last year and a half people have been saying it's the year of the agent or however long”

Claim: Enterprise AI adoption is driven by a small group that understands the technology's potential and a larger group motivated by fear of missing out.

“there's this this whole wave is driven by a combination of some small group of people understanding what is possible from this technology in a way larger group of people driven by FOMO and by the way FOMO isn't bad it creates some forward momentum but it also means that people will latch onto the things that are most in the zeitgeist I guess now for the last year and a half people have been saying it's the year of the agent or however long”

Claim: For approximately the last year and a half, people have repeatedly described the current year as the year of AI agents.

“there's this this whole wave is driven by a combination of some small group of people understanding what is possible from this technology in a way larger group of people driven by FOMO and by the way FOMO isn't bad it creates some forward momentum but it also means that people will latch onto the things that are most in the zeitgeist I guess now for the last year and a half people have been saying it's the year of the agent or however long”

Claim: Fear of missing out can create forward momentum in enterprise AI adoption and is not inherently bad.

“like enterprises are like Can you help build us an agent for some back office task within the organization?”

Claim: A common request from enterprises is assistance building an AI agent for a back-office task.

“That is one of the most common things and it's why now a lot of the work we're doing at 10X is deploying agents into production outside of the end org for companies for repeatable work that they're doing.”

Claim: The speaker's company is deploying production AI agents for repeatable work outside enterprise engineering organizations.

“The second is permissioning and like RBAC.”

Claim: Permissioning and role-based access control are significant challenges in enterprise AI adoption.

“Basically, as every employee in a company has Claude or they have codex and they need access to certain data, it is turned into an absolute nightmare for CIOs and CISOs. And they don't have basically a good gateway for easily doing almost like a like a control panel for data”

Claim: When employees broadly use Claude or Codex and require access to different data, access management becomes extremely difficult for CIOs and CISOs.

“it is turned into an absolute nightmare for CIOs and CISOs. And they don't have basically a good gateway for easily doing almost like a like a control panel for data permissioning.”

Claim: CIOs and CISOs lack an effective gateway or control panel for managing data permissions across enterprise AI tools.

“and that is a huge issue so that's like something we spend a lot of”

Claim: The speaker's company devotes substantial effort to enterprise data permissioning.

“we started by building bespoke for companies and now we actually productize that where we have our own gateway”

Claim: The company initially built bespoke solutions for companies and now offers a productized solution through its own gateway.

“particularly the agent deployments. I think one of the pains that just always gets discovered, I phrase it to the team as, like, you can't automate a process that does not exist.”

Claim: In agent deployments, a recurring problem is that a process cannot be automated if it has not been established.

“So when you have 40 people that are all roughly doing the same function, but they're all doing it a little bit of a different way, there's actually, it's way harder to...”

Claim: Building an agent to proactively perform work in the background is much harder when many people carry out the same function in different ways.

“build an agent that's kind of proactively doing that in the background you can totally sit down with them give them the tools walk them through how they could augment their own workflow but when you really start to get the leverage is when you have these agents that are surfacing work to you you're reviewing it you're throwing it back and you become more of this like you know everyone's been talking about becoming an agent manager and I think we're on the cusp of that”

Claim: Agents provide greater leverage when they proactively surface work for people to review and send back than when employees merely use tools to augment their own workflows.

“build an agent that's kind of proactively doing that in the background you can totally sit down with them give them the tools walk them through how they could augment their own workflow but when you really start to get the leverage is when you have these agents that are surfacing work to you you're reviewing it you're throwing it back and you become more of this like you know everyone's been talking about becoming an agent manager and I think we're on the cusp of that”

Claim: The speaker thinks people are close to taking on roles as managers of AI agents.

“But when you zoom out, oftentimes our recommendation to companies is like, hey, first, we need to build a process around this organization within your company. And then here are the steps for us to actually hand off that work to an agent in the background.”

Claim: Companies should establish an organizational process for a task before defining how an AI agent will take over that work in the background.

“But if we don't have some form of standardization, it's going to be impossible for the agent to know whether it's doing it right for us to build evals around this system and actually deliver the value that an agent that is doing this function would have.”

Claim: Without some standardization of a task, an AI agent cannot reliably determine whether it is performing that task correctly.

“it's going to be impossible for the agent to know whether it's doing it right for us to build evals around this system and actually deliver the value that an agent that is doing this function would have.”

Claim: Standardization is necessary to build evaluations for an agent system and deliver the value of automating the function.

“like uh cj will be okay with me sharing this is like the we were talking about like vertical specific ai solutions and like what does the future look like for these solutions and we we were generally expression expressing an overall bearishness with vertical specific ai solutions in different functions or industries and part of it is just from our vantage point of spending all day in like the guts of enterprises there is so much nuance and so much”

Claim: The speaker and their colleagues were generally bearish about vertical-specific AI solutions across functions and industries.

“like uh cj will be okay with me sharing this is like the we were talking about like vertical specific ai solutions and like what does the future look like for these solutions and we we were generally expression expressing an overall bearishness with vertical specific ai solutions in different functions or industries and part of it is just from our vantage point of spending all day in like the guts of enterprises there is so much nuance and so much such inertia and so like so many systems you have to wrangle and people you have to wrangle and”

Claim: Enterprise environments involve substantial nuance and inertia, as well as many systems and people that must be coordinated.

“also when a buyer out of a vertical solution is expecting truly a one of one experience like we're in this we are in an age where people are expecting one of one software solutions and so if you're expecting that if there's there's”

Claim: Buyers of vertical-specific software solutions expect highly customized, one-of-one experiences.

“It's no wonder that every software company is basically building up a team of FTEs because there's just this convergence happening of services and software when every company wants all this nuance that was not traditionally expected from software.”

Claim: The speaker says software companies are increasingly building teams of full-time employees because customer demand for customization is driving a convergence of software and services.

“And so I think it's really interesting to think about, yeah, like four vertical specific solutions, either one. How will they be able to do this at software margins if everyone expects a one-of-one solution?”

Claim: The speaker questions whether vertical-specific software can provide one-of-one solutions while maintaining software-like margins.

“And also, why should we not expect frontier models to just gobble them up?”

Claim: The speaker raises the possibility that frontier models could absorb or displace vertical-specific AI solutions.

“I would say generally bearish on them, correct.”

Claim: The speaker is generally bearish about vertical-specific AI solutions.

“and uh, you know, I think the the the the answer very simply is like access to data that ChatGP is not gonna have, which is like very common. That's uh that that is a mo”

Claim: Access to data unavailable to ChatGPT is commonly cited as an advantage for vertical AI companies.

“Um obviously and topic and opening are also acquiring those same data sources, but they're just not as dedicated. Uh you know, as as uh, you these guys are gonna be. Um they do have innovations uh if you look if you've like tried”

Claim: Thomson Reuters and OpenAI are acquiring the same data sources as vertical AI companies, though they are less dedicated to those sources.

“Harvey, um they have like the tabular view, like Legora also have them, we we have the C_T_O on the podcast. Um where they they have affordances and U_I_s and all that that like is just customised because they they only serve the them.”

Claim: Harvey and Legora offer specialized interfaces and features tailored to the users they serve.

“The pitch is focus basically.”

Claim: A central selling point of vertical-specific software is its focus.

“in a way that you care about the last mile for your customers, they care about their last mile f in terms of the software. Um and as long as people would train in that software like it never goes away. Like as a banker uh or uh you know in my finance career I could always access the same exact information from the NASDAQ, from the New York Stock Exchange. But why do I pay the fifty thousand dollars a year for Bloomberg? Well, Well, it's just like that's that's what everyone all my peers do and it's all consolidated for me and it's all specialised for me so I don't make mistakes or whatever. Uh there are like it's tied into my trade uh trading system on my chat or whatever, my new system.”

Claim: The speaker says traders may pay $50,000 annually for Bloomberg despite being able to access the same market information elsewhere, because Bloomberg consolidates and specializes information, helps prevent mistakes, and integrates with trading systems.

“in a way that you care about the last mile for your customers, they care about their last mile f in terms of the software. Um and as long as people would train in that software like it never goes away. Like as a banker uh or uh you know in my finance career I could always access the same exact information from the NASDAQ, from the New York Stock Exchange. But why do I pay the fifty thousand dollars a year for Bloomberg? Well,”

Claim: Software tailored to customers' last-mile needs can remain entrenched once users are trained on it.

“No, I think people play for Bloomberg because the social network. Think people love the D_M_s in Bloomberg.”

Claim: The speaker thinks Bloomberg's social network, particularly its direct messaging, is a reason people pay for the service.

“and tools. And like I think ultimately it's hard to be bearish on any of them, mostly because the the demand for software, demand for tooling to do your job is just huge. And some of it yes is gonna be fulfilled by horizontal generalize”

Claim: The speaker argues that strong demand for software and job-specific tools makes it difficult to be bearish on vertical-specific software.

“and tools. And like I think ultimately it's hard to be bearish on any of them, mostly because the the demand for software, demand for tooling to do your job is just huge. And some of it yes is gonna be fulfilled by horizontal generalize tools like the clause and chat G_P_T_s and some of it you're just gonna want something better.”

Claim: The speaker expects horizontal AI tools such as Claude and ChatGPT to satisfy some demand for work software, while other users will seek more capable or specialized tools.

“Um and uh you know, it's just I I was thinking about uh maybe the success of superhuman uh as a as a email client. Everyone has Gmail, but here here's just premium Gmail. with like more stuff for people who live in email all day long. And like th that's a market, right? Yeah, it w maybe wasn't the largest startup in the world, but uh I think uh reasonable people can disagree how large uh each thing is.”

Claim: The speaker describes Superhuman as a premium email client with additional features for people who work in email extensively.

“I think you I think what's not at at uh to be disagreed with is that the the total market's gonna grow in all rising tide is gonna lift all boats. Uh and then then you're just talking percentage purchase of market share”

Claim: The speaker expects the overall market to grow broadly, benefiting companies while leaving them to compete over their share of that market.

“But when I think about like, I guess like products where they are really pushing some agent as the product that is doing work in some specific vertical. cool unless there is some very strategic compliance data whatever it might be mode I think those are going to struggle going forward as we start to see like quite literally the ability to either spin up your own really quickly”

Claim: Standalone products that sell an agent to perform work in a specific vertical will struggle unless they have a strategic advantage, such as compliance capabilities or proprietary data.

“cool unless there is some very strategic compliance data whatever it might be mode I think those are going to struggle going forward as we start to see like quite literally the ability to either spin up your own really quickly continue to compound as the models get better or just you know those tasks start to fall into the the world of the models just generally um but anything that has this like app and tool built around it kind of to my previous point changes the process and has an agent on top of it and i think like”

Claim: As models improve, companies will increasingly be able to build their own solutions quickly, or the relevant tasks will be absorbed into general-purpose models.

“continue to compound as the models get better or just you know those tasks start to fall into the the world of the models just generally um but anything that has this like app and tool built around it kind of to my previous point changes the process and has an agent on top of it and i think like That kind of builds on the same philosophy and I think that can still be super helpful and those will be successful.”

Claim: Products that combine an app and tools that change a workflow with an agent layer can still be helpful and successful.

“uh that overall agent. Which I which I agree with. Um then it's it's it's interesting that Salesforce you know being uh one of the system of record companies that the biggest probably th or the most well-known um is embracing that path as as like uh like you wouldn't You will not use our agents, like we will just provide APIs and M_C_P_s for you to plug us into your”

Claim: Salesforce is embracing an approach in which it provides APIs and MCPs for customers to connect Salesforce systems to their own agents, rather than using Salesforce's agents.

“Um it's uh we were quoted forty thousand dollars a year, we spent a thousand dollars to replace it. It's it's gonna it's going to production next uh for the next conference. So uh almost want like we are that customer that says like I want the one-on-one tool and I'm expecting all these things that we are building a domain specific or vertical specific thing for our business. And like so far it's going okay. Tell me where it's going wrong. Also I'm wondering how many sorry points that that is.”

Claim: The speaker says their team replaced a mid-tier SaaS product quoted at $40,000 annually with a solution built for $1,000, which was scheduled to go into production for the next conference.

“when it's on it um i mean my general view is i think the the ripping out of my my general view is like the more you move up marketing business the harder the proposition of ripping out and building from scratch gets i agree with you like part of the reason when as people have said like”

Claim: The higher a software product sits in the marketing and business stack, the harder it is to replace it and rebuild it from scratch.

“A very small group have said Salesforce is really at risk, SAP is really at risk. Don't agree with that at all.”

Claim: The speaker strongly disagrees with the claim that Salesforce and SAP are seriously at risk.

“It's just those businesses in my mind just accrue value in this moment in time. And part of the reason they accrue value is there's so much money at stake where if a company gets it wrong ripping them out, you're talking about potentially hundreds of millions or billions of dollars.”

Claim: The speaker believes that large system-of-record businesses accrue value because replacing them can put hundreds of millions or billions of dollars at risk, which makes companies risk-averse.

“Yeah, exactly. I think for mid-tier. It really, it actually comes down to what is the size of the opportunity? And is that ever going to be like the worthiest opportunity to rebuild inside of a company?”

Claim: For mid-tier software, whether an enterprise should rebuild internally depends on the opportunity's size and whether it merits company resources.

“I'm trying to even think like in an enterprise, let's just say there's some mid-tier SaaS that is a few million bucks a year. I think the first question is, is like, From a percentage of revenue or cost perspective, is that going to ever move the needle enough to basically allocate engineer resource to?”

Claim: When considering replacing a mid-tier SaaS product that costs a few million dollars annually, an enterprise should assess whether the savings or benefits are large enough relative to revenue or costs to justify engineering resources.

“I think it's going to be case by case. I think for some enterprises, the answer may be yes. For other enterprises, the answer may be no.”

Claim: The case for rebuilding software internally is enterprise-specific: it may be worthwhile for some enterprises and not for others.

“And one thing I think we realized, and every company realizes this, is one of the beauties of software is not just the utility that you get from it. It is all of the headaches you don't have to think about in maintaining it and improving it when it's not in your core. So I think.”

Claim: Software provides value both through its utility and by sparing a company the ongoing work of maintaining and improving non-core systems.

“It basically comes down to, is the size of the prize big enough to justify? The cognitive load that is going to take from some amount of resource in your company to nurture something that is not core to your business.”

Claim: A company should build or maintain a non-core system internally only if the expected benefit is large enough to justify the cognitive load and resources required to nurture it.

“So there's two pieces to it. One is like not just we're being overcharged for a product, but it's like the product has been mediocre for a long period of time.”

Claim: The speaker says the replaced product had been mediocre for a long time, in addition to being perceived as overpriced.

“What would the 10x better version of this product look like?”

Claim: The speaker frames the product redesign question as identifying what a version ten times better would look like.

“because we are the customers and we can iterate with like minutes and not months.”

Claim: Because the team is also the customer, it can iterate on its product in minutes rather than months.

“I think I actually think what's interesting is I think startups and like even like high growth midsize companies can be nimble in that way.”

Claim: Startups and high-growth midsize companies can often iterate nimbly on solutions to their own needs.

“I think for enterprises where like they're so entrenched in a. in a in a certain uh capability rarely is that going to be the the thing that they're rarely are they going to say like oh we can create a 10x better version of this sass we use and we're going to spin it out potentially as its own thing or whatever it is like i just don't think enterprises function that way”

Claim: Enterprises rarely build a 10x-better version of software they use and spin it out as a separate business, because the speaker does not think enterprises generally function that way.

“Yeah, I mean I think there's there's two pieces here. I I really deeply believe in the like AI will produce good for humanity,”

Claim: The speaker deeply believes that AI will produce good for humanity.

“it's going to augment people's jobs,”

Claim: AI will augment people's jobs.

“it will absolutely change and displace, but I think it's more of a movement as opposed to just like cutting off work.”

Claim: AI will change and displace work, but the speaker views the shift as a broader movement rather than simply eliminating work.

“I think A16Z posted something about how like call centers are truly now on the decline and AI is starting to handle more and more of that work and ultimately if that is a skill you've built up your whole career that is devastating to have something come in and take that function over.”

Claim: According to something A16Z posted, call centers are declining as AI begins to handle more of their work.

“I think A16Z posted something about how like call centers are truly now on the decline and AI is starting to handle more and more of that work and ultimately if that is a skill you've built up your whole career that is devastating to have something come in and take that function over.”

Claim: It can be devastating when AI takes over a function that someone has spent their career developing as a skill.

“That being said, part of what we do when we're talking to customers about, hey, an agent is now going to do this work is actually around reallocation.”

Claim: The speaker's company frames the introduction of agents to perform work for customers partly as a matter of reallocating work.

“I don't think any business would claim like once we have this solved, there's nothing for this group of people to work on and try to solve.”

Claim: The speaker does not think businesses would claim that solving a problem leaves the people who worked on it with nothing else to do.

“So our approach really is about. What can we replace and then what are the next biggest problems to look at?”

Claim: The speaker's approach is to determine what work can be replaced and then identify the next biggest problems to tackle.

“And I'd make an argument in most large companies that that cycle kind of continues forever.”

Claim: The speaker argues that in most large companies, solving one problem leads to another, so the cycle continues indefinitely.

“I think about how we've grown, there's always been another problem and actually taking those people that may have this like super deep subject matter expert and bringing them above water to...”

Claim: As the speaker's company has grown, there has consistently been another problem to address.

“always been another problem and actually taking those people that may have this like super deep subject matter expert and bringing them above water to... Start to maybe work more across teams or like think about parts of the business less so on like the level of a cell in a spreadsheet and more about how these teams are working together.”

Claim: The speaker describes shifting deep subject-matter experts toward cross-team work and a broader view of how teams collaborate, rather than focusing narrowly on individual spreadsheet cells.

“Ultimately, in the long run, I think that keeps most people there and allows those businesses to become more nimble and actually tackle like the big meaty problems and not the like drudge of most people's day to day when you, you know, when you're thinking about someone that's in like a finance division.”

Claim: The speaker thinks that, in the long run, reallocating people toward broader and more consequential work can keep most people employed while making businesses more nimble and better able to tackle major problems.

“They're mostly in spreadsheets, but how can we elevate them to think strategically about the business and where money is allocated and things like that.”

Claim: The speaker says finance-division employees mostly work in spreadsheets and argues they could be elevated to think more strategically about the business and allocation of money.

“And I also feel like when we have these conversations around like job displacement or job reallocation, we talk about it as this thing like happening to people. And one of the reasons I don't like that is I feel like. It doesn't give enough credit to people to like have agency in their own lives,”

Claim: The speaker objects to framing job displacement or reallocation as something that merely happens to people, because it underestimates their agency.

“meaning I think, people really should be thinking like when I think about like enablement or people reskilling, I think it is the job of an employer because they want to drive growth in their business, but it is also the job of an employee. And I think when people say that AI is going to replace all of the work that people don't enjoy doing.”

Claim: The speaker believes both employers seeking business growth and employees have responsibility for reskilling.

“people really should be thinking like when I think about like enablement or people reskilling, I think it is the job of an employer because they want to drive growth in their business, but it is also the job of an employee. And I think when people say that AI is going to replace all of the work that people don't enjoy doing. I think that's a little bit of like cope for the reality of just people are going to be put out of jobs because a lot of people don't have the luxury of liking or not liking a job. They do a job to make money.”

Claim: The speaker thinks the claim that AI will replace only unpleasant work is partly a coping response to the reality that some people will lose their jobs.

“I think that's a little bit of like cope for the reality of just people are going to be put out of jobs because a lot of people don't have the luxury of liking or not liking a job. They do a job to make money.”

Claim: Many people work to earn money and cannot afford to choose jobs based on whether they enjoy them.

“And also I think people to the extent that it's possible should have agency in saying to themselves, OK, what is the stability or replaceability of my job over the next few years?”

Claim: People should, where possible, evaluate the stability and replaceability of their jobs over the next few years.

“And if I am to try to skill up in a way that I'm better protected, what does that look like? look like people have agency to say that to themselves and actually take action yeah”

Claim: People seeking greater protection from job replacement should identify ways to build skills and take action where possible.

“Yeah. I mean what I'd say is I feel like basically I would just say there's a bubble in New York that is a smaller bubble than exists in S_F_ If S_F_ is like a big bubble”

Claim: The speaker thinks New York has an AI-focused bubble that is smaller than San Francisco's.

“I think we at 10x feel like you know we are in a what I would say is like a microcosm of San Francisco because of the people that self-select to work at the company and then also the companies that we're partnered with and like the people that we engage with in New York but I would for sure say also it is true that it is a smaller universe than like in SF when you go to SF it feels all-encompassing where it's like everyone everywhere you look”

Claim: The speaker views 10x as a microcosm of San Francisco, reflecting its self-selected employees, partner companies, and New York contacts.

“I think we at 10x feel like you know we are in a what I would say is like a microcosm of San Francisco because of the people that self-select to work at the company and then also the companies that we're partnered with and like the people that we engage with in New York but I would for sure say also it is true that it is a smaller universe than like in SF when you go to SF it feels all-encompassing where it's like everyone everywhere you look”

Claim: The speaker says San Francisco's AI-focused community feels all-encompassing, as though such people are everywhere.

“people go on Twitter or X because they feel like they're on edge city they're living in the future like AIE is like that on steroids like I feel like SF is similar to AI I mean it's like I probably like the the overlap of AIE generally and and people who have lived at some point in their life in SF is probably a very”

Claim: The speaker says Twitter/X can make people feel as though they are living on the edge of the future, and describes AIE as an intensified version of that experience.

“We're roughly 1% of the city now.”

Claim: The speaker says their group is roughly 1% of the city now.

“actually for Wolfsburg yeah that's crazy yeah but like to that point I just think like it is in the culture DNA of SF where a very big population of people live in the future.”

Claim: San Francisco has a culture in which a very large population of people live in the future, according to the speaker.

“I think you can find that group in New York and I think we at 10X have found that people and part of our goal is to drive like AI as a hub or in New York as a hub for AI in in the US because there is so much commerce here that like it is going to be a place where diffusion of the technology actually has such a meaningful impact on the GDP of this country.”

Claim: 10X aims to help establish New York as a hub for AI in the United States.

“I think you can find that group in New York and I think we at 10X have found that people and part of our goal is to drive like AI as a hub or in New York as a hub for AI in in the US because there is so much commerce here that like it is going to be a place where diffusion of the technology actually has such a meaningful impact on the GDP of this country.”

Claim: The speaker expects diffusion of AI in New York to have a meaningful impact on United States GDP because of the city's substantial commerce.

“I think the bubble is smaller now and we take it almost as like a charge of ours to increase our bubble.”

Claim: The speaker believes New York's AI community is smaller now and regards expanding it as one of their responsibilities.

“possible maybe I don't know that it's possible to get to SF size, but I do think we can have a meaningful impact on getting more people to live in the future here.”

Claim: The speaker is uncertain whether New York can match San Francisco's size, but believes they can meaningfully increase the number of New Yorkers who live in the future.

“it's slowly creeping in. graffiti on that. That was I think that was like a little bit of an intentional rage bait. But now you have the unironic A_I_ billboards and the people are like oh I guess that's part of life now.”

Claim: The speaker says AI is gradually becoming normalized in everyday life, as reflected by AI billboards and public acceptance of them.

“I think there's an opportunity to bring kind of like the edge and like the just like the living in the future nature that SF has created because of just like the economic climate that exists or because of venture, because of tech companies, then because of the people that are attracted there.”

Claim: The speaker attributes San Francisco's future-oriented culture to its economic climate, venture activity, tech companies, and the people they attract.

“I think there's an opportunity to bring kind of like the edge and like the just like the living in the future nature that SF has created because of just like the economic climate that exists or because of venture, because of tech companies, then because of the people that are attracted there.”

Claim: The speaker believes New York has an opportunity to develop the edge and future-oriented culture associated with San Francisco.

“of us is like we're in a it's weird we're in a smaller pond for ai but i would argue we're in as big of a pond if not a bigger pond for like uh commercial value and”

Claim: The speaker considers New York a smaller ecosystem for AI than the comparison implied, but says it is as large or larger in commercial value.

“The way I put it is like, the reason I do AI in New York is I bring all the people from San Francisco, come here and meet their customers and also like get a reality check.”

Claim: The speaker brings people from San Francisco to New York to meet customers and gain a reality check.

“totally and i think we've actually think there's a really interesting arbitrage there yeah yep”

Claim: The speaker sees an interesting arbitrage in connecting San Francisco-based people with New York customers and commercial realities.

“yep we're hiring uh For deployed engineers at 10X, just go to 10X.co.”

Claim: 10X is hiring deployed engineers.

“We're also always hiring AI strategists, which are basically technical consultants who are leading client engagements.”

Claim: 10X is always hiring AI strategists, defined as technical consultants who lead client engagements.

“And the team is 75 strong in New York City.”

Claim: The speaker's team has 75 people in New York City.

“We're continuing to make New York City our home.”

Claim: The speaker's team is continuing to base itself in New York City.

“And then if you are an enterprise who is looking to actually get AI native and you've tried it, it's not working, hit us up.”

Claim: 10X invites enterprises that have tried but failed to become AI-native to contact the company.

“I promise we'll do better work than, not going to name them, but a lot of other traditional consultancies.”

Claim: The speaker promises that 10X will do better work than many traditional consultancies.

“a special project involves anything outside of kind of our client work. So I think about my brain right now, it is building that meta harness and system internally that”

Claim: The speaker is building an internal meta harness and system.

“a special project involves anything outside of kind of our client work. So I think about my brain right now, it is building that meta harness and system internally that”

Claim: A special project is work outside the company's client engagements, according to the speaker.

“Basically, it's like find the bleeding edge of what the models can do and give them tools to augment how well they can do it. Um and then also internally working on the platform that Alex mentioned, um big goal there, diffuse these background agents and actually get them out to enterprises in a way that's truly truly helpful to them.”

Claim: The speaker's approach is to identify the limits of current model capabilities and provide tools that augment those capabilities.

“Basically, it's like find the bleeding edge of what the models can do and give them tools to augment how well they can do it. Um and then also internally working on the platform that Alex mentioned, um big goal there, diffuse these background agents and actually get them out to enterprises in a way that's truly truly helpful to them.”

Claim: A major goal is to deploy background agents in enterprises in a way that is genuinely helpful to them.

“And solve a lot of those like messy IT problems that that we hit on earlier.”

Claim: The speaker intends the background agents to solve many of the messy IT problems discussed earlier.