Layoffs, the two levers, AI

• Companies are justifying layoffs with AI restructuring now. Many are seeing strong financials but are still laying their workers off. Recently ClickUp's CEO explicitly tweeted that he would use the salaries of the laid off workers to increase current AI leveraged employees to million dollar salary bands. To me it seemed a little out of taste but I get it. It is wartime now. Wartime CEOs will succeed while CEOs who don't innovate will be eaten.

• I have been studying game theory through twitter/ChatGPT and right now companies are all exhibiting game theory fundamentals. Due to Nash equilibrium all companies must announce layoffs due to AI even if it causes instability and mistrust if they do not see AI revenue gains. Because the alternative is bloat, wasting dollars, or making the executive leadership team unhappy. If AGI is here (some say it is) then those who do not rebuild and restructure will be left behind and eaten up by those who do. The dominant strategy becomes restructure first, those who are smaller, faster, flatter, AI-native will be rewarded by the market.

Nash equilibrium is when each player is choosing their best move given what everyone else is doing, so no one can improve by changing their strategy alone.

• This is similar to what people have been saying about the industrial revolution and electricity. Productivity did not get realized until the factories were rebuilt from the ground up with electricity in mind. In a recent interview by Ali Ghodsi (Databricks CEO) he mentioned how in the beginning people tried to just replace steam engines with electrical components. This did not do anything to productivity. They had to rewire and rebuild the factory infrastructure, machines, workflows, managerial methods, etc in order to see the productivity gains. This was also mentioned in the Stripe sessions talks where I wrote about it in my Stripe Sessions reflection.

• The insight with the landscape of current tech companies now is to leverage AI as much as possible or be a part of a AI-driven company. Your company either has to create AI products or utilize AI products within. The first part is easy (if the company is an AI company already) the second part is not. Executives have no idea how to leverage AI. This is another reason why layoffs have become the de facto method in being "AI-forward". The company can't prove the productivity gains, because there is no way to prove it, so what they do instead is lower costs. There's two levers, the revenue lever and the cost lever. If your revenue is not growing due to AI then your costs must be lowering. That's what the investors want to see which inevitably push the executives who then have to decide *how do we lower costs*.

• And that is what we are seeing now, the AI product companies are seeing explosive growth in AI products. You have the huge market players like Anthropic who claim to see annualized revenue increase by 80x in Q1. You then have small AI one-man or small team startups who are working 996, deploying agents 24/7, and are prime for taking over old enterprise market share. It is kill or be killed right now which describes what we are seeing as the SaaSpocalypse.

• Some layoff stats:
  • Coinbase - 700 jobs 14% of workforce
    • smaller teams, fewer pure managers, more AI-enabled productivity
  • Block/Square - 4,000 jobs 40% of workforce
    • smaller teams, smaller company structure = better AI implementation
  • ClickUp - 22% of workforce
    • focusing on 100x org, states revenue is doing well but they must restructure and reward the employees who become AI empowered
  • Meta - 8,000 jobs 10% of workforce
    • lower bureaucracy, smaller teams, more AI-native workflows

• You can see that these layoffs are not AI companies, they are in different industries i.e. crypto, fintech, SaaS. They are fearing for their lives because if they do not innovate then they will be killed. For example you see Notion as a huge AI player now. They pivoted quickly and made sure they integrated AI into their products ASAP. Now people are using their platform for agent workflows, that is the correct way to act in this market. They are being rewarded and see AI revenue gains, this is an example of pulling the revenue lever vs the cost lever. Coinbase, Block/Square, ClickUp are pulling the cost lever.

• Notion is doing everything right in my opinion. They recently launched a developer API which is essentially perfect for agent workflows. Companies that follow this path and open up their platform with APIs, CLIs, and MCPs will win. In my Stripe Sessions reflection I mentioned how agentic commerce is the new thing. Instead of humans interacting with shops and stores you will see agents doing everything. This is the future, you need to set up your platform/product so that agents can interact with it.

• Pricing will need to be reimagined as well. In the SaaS era, people had to pay for seats. This is because companies built UIs for humans who were trained specifically to use their product. We used to see SFDC training certificates, Oracle training certificates, Quickbooks etc., each user would have a seat purchased for them as part of an enterprise plan. In the agentic era this will be gone. Companies will need to open up their infrastructure for AI agents. The new moat will be having the data and allowing agents to interact with the data. Pricing will be usage based, pay as you go, metered billing, etc. We are already seeing this with Claude Code. This is also something I attended a talk about in Stripe Sessions reflection.

The three layers to succeed in an agentic era

• Companies need to restructure two ways:
  1. They need to rebuild processes and workflows from the ground up
  2. They need to restructure their products for AI Agents

• Restructuring processes and workflows will require the 3 pillars I've stated before
  • the quantitative layer
  • the qualitative layer
  • and the execution layer
• I'll expand on each

• The quantitative layer is the system of record many companies used to pay huge enterprises for. For example in finance and accounting it would be the NetSuite GL Ledger. This holds the quantitative tabular information you needed to see information such as revenue, cash, expenses, etc. With this layer you can grab things like flux analysis, income statements, financial reporting. In the agentic era, this layer is very important because it is the quantitative context agents will use to answer source of truth questions. This layer is not just in financial accounting it is also in other types of departments such as HR/people, game analytics for gaming companies, retail store purchase data, webstore data, app store data etc. Although many of these databases do end up being used by finance and accounting the quantitative layer exists for all departments and industries.

• The next layer is the qualitative layer this is the data that provides the why in the quantitative dimensions. For example in finance and accounting you can easily grab an automated flux MoM or QoQ number from pulling the financial statements. The agent can tell you how much changed but it cannot tell you why. This is why we need the qualitative layer. This may include unstructured data or it may include fields that you can grab from system of records. This layer, I would say, is more difficult than the quantitative layer and people/companies usually do this part wrong or completely miss it which is why they do not see gains in their AI productivity. Context is king, and getting these two layers is how you grab context for your AGI systems/agents.

• The last layer is the execution layer. This is the AI agents, the codebase, the pipelines, the apps. Without the first two layers this is useless. Many people and companies are skipping the first two layers and going straight to the execution layer. This is when we see executives push for employees to use AI but all they do is send them ChatGPT or Claude. They are not investing in the first two layers. It is impossible to rebuild, be AI-driven, and realize productivity without all three layers.

Conclusion

• The next few years will be very interesting. It took ~30 years for the industrial revolution to see productivity gains in from electricity. I assume the AI revolution will be a lot faster. In the end we will see huge amounts of innovation and hopefully expand human capabilities and resources to things we cannot even fathom. Something I always think back to is, imagine explaining to a hunter-gatherer what a twitch streamer is. They would first have to understand all the revolutions we went through; agricultural, industrial, tech, and more, and then they would have to understand that there is no more scarcity in resources such as food. A successful twitch streamer can play games all day and talk to people around the world and make millions of dollars. This is something incredibly hard to describe or even imagine 50 years ago. In the next 10-50 years we will see things that are 1000x that.