Creator perspective · MoneyThe more you move up market in business, the harder the proposition of ripping out and building from scratch gets. CJ doesn't agree that Salesforce and SAP are really at risk of being ripped out — those businesses accrue value in this moment in time because so much money is at stake: getting it wrong ripping them out means potentially hundreds of millions or billions of dollars, making companies risk-averse. For mid-tier SaaS replacement, the key question is the size of the opportunity and whether it will ever be the worthiest opportunity to rebuild inside a company — whether from a percentage of revenue or cost perspective it will move the needle enough to allocate engineer resource to. Whether enterprises replace mid-tier SaaS will be case by case.
the guest's judgment
Latent.Space · AI for the Fortune 2000: How Tenex Builds an AI-Native SDLC — Alex Lieberman & CJ Hess
Report · Money50% of Tenex's customers are Fortune 2000 and 50% are mid-market companies doing $100-500M in revenue. Enterprise customers were asking for AI transformation as a service, not just staff augmentation and AI engineering.
Latent.Space · AI for the Fortune 2000: How Tenex Builds an AI-Native SDLC — Alex Lieberman & CJ Hess
Creator perspective · MoneyBoth employers and employees share responsibility for reskilling: employers because they want to drive growth, employees because they have agency in their own lives. Assess the stability or replaceability of your job over the next few years, and if you need to skill up to be better protected, figure out what that looks like and take action. The claim that AI will replace all the work people don't enjoy doing is cope — many people do jobs to make money and don't have the luxury of liking or not liking a job. A16Z has posted that call centers are truly now on the decline, with AI starting to handle more and more of that work. If a skill you've built your whole career (like call center work) is taken over by AI, that is devastating.
Latent.Space · AI for the Fortune 2000: How Tenex Builds an AI-Native SDLC — Alex Lieberman & CJ Hess
Creator perspective · MoneyThe leverage point with AI agents is a review loop: agents surface work, you review it and throw it back, moving you toward being an 'agent manager'. When walking employees through AI adoption, you can sit down with them, give them the tools, and show them how to augment their own workflow — but background agents require standardized processes first.
Latent.Space · AI for the Fortune 2000: How Tenex Builds an AI-Native SDLC — Alex Lieberman & CJ Hess
Report · MoneyTenex started by building bespoke permissioning solutions for companies and has now productized that into its own gateway solution. A significant portion of Tenex's current work is deploying agents into production for companies doing repeatable work, and enterprises commonly ask for agents to be built for back-office tasks within the organization.
Latent.Space · AI for the Fortune 2000: How Tenex Builds an AI-Native SDLC — Alex Lieberman & CJ Hess
Creator perspective · MoneyAlex Lieberman argues that for mid-tier SaaS, the decision to rebuild comes down to the size of the opportunity and whether it is ever worth allocating engineer resource to it. For a mid-tier SaaS costing a few million a year, the first question is whether, from a percentage-of-revenue or cost perspective, it will ever move the needle enough to justify engineer allocation. He expects the answer to be case by case: for some enterprises yes, for others no.
Latent.Space · AI for the Fortune 2000: How Tenex Builds an AI-Native SDLC — Alex Lieberman & CJ Hess
Creator perspective · MoneyAI agents are more and more effective if they have access to the data and the APIs that run your business; without that access, they're only as smart as the data they have access to. The product goal is to eliminate the custom integration: as soon as a user connects their system of record, deliver value in a way that feels very tailored to them. After demonstrating value with email and calendar, the question becomes whether the company has earned the right to ask the customer to connect their CRM and other tools via their IT team.
The a16z Show · What Happens When Al Agents Can Spend Your Money?
Principle · MoneyBuild enterprise products that feel like consumer experiences; CEOs will demand the same magic they get personally for their companies, and whoever delivers that wins.
endorsed
The a16z Show · What Happens When Al Agents Can Spend Your Money?
Creator perspective · MoneyAmazon adopted an adversarial approach to agents (blocking/breaking the old internet covenant) because they're an advertising company. The old covenant was: Yahoo gives you a free stock quote, you give Yahoo your eyeballs for an E-Trade ad. That covenant has been broken. In a world where agents act on your behalf and human eyeballs no longer look at things, there's a big opportunity to be 'synergistic' rather than adversarial — it's like the new SEO, though what that looks like is a very open question.
The a16z Show · What Happens When Al Agents Can Spend Your Money?
Creator perspective · MoneyIn most AI products today, capabilities will be very similar — most harnesses in the work context will be able to do the same things — and product features that are ahead of the market, historically the main differentiator in Silicon Valley, will become more and more difficult to sustain as an advantage.
speaker's forward-looking argument, described with 0.8 confidence
The a16z Show · What Happens When Al Agents Can Spend Your Money?
Creator perspective · MoneyThe company believes that custom software all the way through can be built using AI, and that over time, through this approach, it can offer 80-20 very customized software for people.
company belief, described; not an established fact
The a16z Show · What Happens When Al Agents Can Spend Your Money?
Creator perspective · MoneyCompeting with your suppliers is always risky, and the speaker's company is competing in a way against model providers because they have cheaper costs. Businesses should choose a neutral app that uses whatever AI supplier is best for them in the moment, rather than being locked into one provider. Model choice isn't just about general intelligence — suppliers differ in which is better at voice, video, or slides at any given time.
speaker's position
The a16z Show · What Happens When Al Agents Can Spend Your Money?
Creator perspective · MoneyYou can eventually work your way backwards into enterprise sales and sales-led growth after building a user-first product. The company plans to learn from thousands, tens of thousands of businesses and people that use their product, and then verticalize from there by letting the market define the verticals.
endorsed strategy
The a16z Show · What Happens When Al Agents Can Spend Your Money?
Creator perspective · MoneyRandy Hunt explains that the first reason people leave talented, empowered small teams is the next chapter of their career — often after about 20 years in the same mode of work, experienced designers who are great at staying close to the work reach a chapter change. Of the three reasons people leave, only one is within a leader's control. The sticky experience of making things and having people like them, plus giving people agency and responsibility and trusting them, supports retention.
Design Better · Randy Hunt: Notion and Etsy design leader on why the best design teams stay small
Creator perspective · MoneyDylan frames the core problem: most AI teams serving open source LLMs are stuck between two bad options — closed APIs that are easy to start with but hit a ceiling fast, and self-hosting that gives full control but is a massive engineering project.
AI Engineer · What Makes Open Models Fast in Production — Sujee Maniyam, Nebius
Creator perspective · MoneyCarmen Li explains that WTI is a physically delivered commodity contract, while Brent is cash settled, and they have distinct use cases. For hedging price volatility or cost volatility in the financial sense, use Brent; for physically procuring resources, use WTI. She says Compute Exchange is creating the WTI equivalent for compute, and describes the platform as 'tomorrow's compute, today's price' — a physically procured platform.
Analogy and product positioning from the creator.
AI Engineer · Buying GPUs Should Work Like Buying Stocks — Carmen Li, Compute Exchange
Report · MoneyIn the last two and a half months, computer-use and browser-use assistants can do things that were near impossible six months ago, succeeding roughly 90% of the time versus about 30% before.
speaker's rough figures with possible numeric error flag; moderate confidence (0.7)
The a16z Show · What Happens When Al Agents Can Spend Your Money?
Report · MoneyThe infrastructure layer versus application layer dynamic has always existed in tech: Cisco and Akamai were infrastructure companies in the 90s while Amazon and eBay were application companies — and Cisco never tried to build a bookseller, auction site, or directory like Yahoo, as infrastructure companies historically stayed out of applications.
historical analogy offered by the speaker
The a16z Show · What Happens When Al Agents Can Spend Your Money?
Creator perspective · MoneySujee explains that when you have hundreds of thousands of GPUs running models, load balancing and routing becomes an issue. Putting a load balancer in front and sending traffic randomly to different GPUs does not work well for LLMs because the workloads are different — inputs can be pretty large (like sending large code bases using coding agents to refactor, summarize, or analyze) and outputs vary from large to small and large to large, so routing accordingly is not trivial. With random routing, the cache is fragmented all over and the cache rate isn't great because requests land randomly at different GPUs that may or may not have the cache.
AI Engineer · What Makes Open Models Fast in Production — Sujee Maniyam, Nebius
Recommendation · MoneyPrioritize training and enablement when rolling out AI tools — it never gets the love it should but is arguably as important as the technology itself.
Latent.Space · AI for the Fortune 2000: How Tenex Builds an AI-Native SDLC — Alex Lieberman & CJ Hess
