TAISER Shock: The Rise of Tokenized AI Services
Tokenized AI services (TAISERS) are building a powerful economic model on crypto financing and SaaS revenue.
Surging enthusiasm for AI agents has generated more than $10B in new crypto market capitalization in just a few months. Early agents are essentially producing social media spam. We are about to find out what happens when builders direct their energy and money toward a wider range of services.
They are assembling a new generation of apps from AI services instead of Web/microservices. The new generation benefits from faster funding, faster AI-assisted coding, and faster addition of functions.
Read this article to learn how to build your own TAISERS for fun and profit. Read Claude’s pitch for “Why TAISERS are about to disrupt the $300B SaaS industry” for a quick description of the opportunity.
First generation “crypto AI agents” handle only investor relations
The first generation of tokenized agents is designed to post on social media. Their posts can promote a memecoin. I think of these agents as providing an “investor relations” service for their affiliated coins.
The TAISER generation adds functionality
Tokenized AI services add functions to make a user app.
In this illustration, we imagine REBA, a real estate buyers agent. AI assisted investing is already a popular theme. REBA provides services for professional real estate investors. We can build these services in stages.
- First we fund and market the project, using a generic investor relations function.
- Then we add search and analysis, using AI capabilities to read a range of data sources.
- From there, we can build assistants to contact sellers and engage in due diligence and purchasing workflow.
- If this is successful, we have ongoing management and reporting responsibilities.
TAISER Economics
Services can generate revenue. A buyer’s agent might make a cash commission. Or, users of services might pay a subscription fee. We can handle this revenue for token holders. A cash commission can go to token buybacks. A subscription fee can be expressed as a requirement to buy and hold a certain number of tokens.
Regulators are starting to allow trading in tokens that are “utility” (European MICA language) or “consumptive” (US CFTC proposed language). A TAISER business model can fit these definitions. Buying a token for fees, and staking it for subscription access, are “utility” and “consumptive” uses.
Personal attention for each customer
A TAISER can serve multiple customers with a memory store for each customer.
An AI service with many customers will need to learn about each individual customer. AI agent frameworks do this by saving and loading memory for each customer. They use several types of memory structure, including databases (like normal SaaS apps), text, and vector embeddings (a category of data from AI that saves the meaning of information). We can create a unified experience for the customer by feeding the same memory into multiple AI agents and runtime frameworks.
Functional improvement
We can incrementally improve each functional agent.
AI agents are not like humans. Humans are born with few skills except a talent for unsupervised learning, pick up skills largely in a single burst, and then die without passing on most of their knowledge. Functional AI agents start with a significant base of knowledge and can pass it on to upgraded versions.
TAISERS are similar to the services from Amazon, Microsoft, and Meta. Big Tech employs thousands of “microservices” to deliver their top-level applications and products. They can upgrade each microservice separately. They use frequent component upgrades to get continuous improvement, market dominance, and trillion dollar market caps.
Thousands of people build, test, and operate the microservices at an operation like Google. Putting Google’s upgrade process into the hands of small teams is the software equivalent of putting room-sized supercomputers into mobile phones.
Each functional component can improve by sharing with its comrades that perform the same function. We may see a classic organizational matrix. Functions will have their own upgrade cycle. Apps can link functions into workflows for a specific user.
I expect that users will pay for apps that are tuned to their needs, and app developers will pay to support their function supply chain.
TAISERS versus SaaS
Builders can use the TAISER life cycle to build out “vertical SaaS” apps. Vertical SaaS is a hot category because apps can bring AI functions that replace labor. This brings in a different type of pricing, with fewer seats generating more revenue.
The TAISER version is different from normal SaaS because:
- It offers liquid tokens
- It adds functions quickly using AI agent services, rather than Web/microservices
Tokens versus Equity
“Tokenized” AI uses tokens for financing and other types of engagement. The tokens are liquid. You can actively trade them in DeFi. The most recent generation of tokenized AI uses the most radical form of instant liquidity. It issues most of the tokens for trading, before it does anything else.
SaaS companies are funded by non-transferrable equity. They sell it slowly and incrementally. They build value over time by accumulating recurring revenue. Then, they can sell the company or the stock with an exit valuation that is roughly related to the amount of recurring revenue.
SaaS founders do not want people trading their stock. Purchase agreements for startup stock prevent it. They want to avoid problems like:
- Early round investors selling in later rounds and taking money away from company fundraising. The CEO of a capital-consuming startup wants to be a monopoly seller of equity.
- Crashing valuations when investors sell a chunk of stock in an illiquid market.
- Investors that blame the company if they buy or sell without knowing important information. Private companies do not want disclose everything they do, every day.
The current craze for AI tokens is rushing exuberantly past these problems. Are we seeing irrational exuberance? Or, are we seeing what economists call “rational expectations” about the value of liquidity and the cost of transfer restrictions?
Buyers will pay more for unrestricted assets than they will pay for assets with transfer restrictions. Startup investors have suffered through three years (2022–2024) when they couldn’t sell their restricted assets. This increases their motivation to pay for liquid assets.
Sellers can get a higher price now, in exchange for suffering the problems of liquidity later.
TAISER founders can leverage this effect when conditions are favorable.:
- They get more money at founding, which is an advantage when new opportunities emerge.
- They get money faster, which is an advantage in times of rapid change.
- They reduce the burden of disclosure by accepting open source and blockchain business models that are transparent and self-documenting.
- They use investor relations agents to manage some of the problems of liquidity. We are building an increasing number of features to manage AMM pools, sell tokens to fund operating costs, handle block sales from bigger holders, and make disclosures.
Tokenized projects will often fund operations by incrementally selling tokens into a liquid market, rather than by raising a batch of funding. This technique is risky and unreliable because buyers can stop buying at any time. However, it conveys a speed advantage during good times. Selling tokens incrementally, with the help of an investor relations agent, is faster than organizing and selling a funding round.
Speed wins
The fastest builders will win. Is it easier to add and improve a SaaS architecture, or a TAISER architecture? When we add SaaS functions, we usually need to build, test, and maintain APIs that feed structured data to the new function.
In theory, we can rapidly add functions to a TAISER because AI agents have a talent for reading unstructured data. We can drop in a new agent, point it at user data, and tell it to go to work for our existing customers.
Get Involved
Use the Taiser.ai launchpad for non-dilutive token funding. Grow your app by adding functions and users. Follow on Twitter or join the Discord
