How do AI products survive the bubble and move towards longevity? Introduction A16z gives five suggestions for enterprise AI startups

AI has become the strategic core of enterprises, not just a topical tool, but also a productivity engine. With the Fortune 500 companies trying to integrate AI collaboration into their businesses, enterprise-level AI is ushering in a new era of competition. However, unlike the past SaaS model, to succeed in this changing landscape, AI startups need to be not only fast but also precise and have a long tail.

0/ AI companies behave differently from traditional SaaS companies, and founders consistently ask us how AI companies are adapting and breaking out.

After talking to hundreds of AI companies over the past few years, we think that there are a few emerging principles for… pic.twitter.com/KJ511PjC7t

— Kimberly Tan (@kimberlywtan) June 24, 2025

Venture capital firm a16z, Andreessen Horowitz, in a lengthy article 'From Presentation to Transaction: Enterprise AI Insights,' analyzes how AI startups stand out in the competition. The following is compiled and edited by Chain News.

It's not difficult to show off skills, but it's difficult to create AI products that can last for a long time.

The wave of generative AI has made many products like 'GPT wrappers (GPT wrappers)' easily replicable and ultimately consumed by large models. In reality, it's easy to create a stunning demo that can attract attention, but the real challenge lies in making AI operate stably in actual enterprise scenarios.

From model selection, evaluation criteria, task and workflow design, to deeply embedding products in customer data and business logic, AI startups face extremely tedious and variable engineering challenges.

a16z emphasizes that enterprises not only require accuracy but also value the reliability and liability risks of models in critical areas such as legal or accounting. Therefore, for AI to enter the core systems of enterprises, it cannot rely solely on surface effects but must also achieve long-tail and fault tolerance.

(Long text introduction Redwood Capital's strategic advice to entrepreneurs: How AI can become the next trillion-dollar economy?)

AI new venture competition heats up: 10 times annual revenue growth becomes a new threshold

In the past, achieving an annual recurring revenue of $1 million was considered an early indicator of success, but today it is below the average for AI startups. According to Stripe and market data, many AI companies have seen their annual revenue grow by more than 10 times, with some even reaching millions in revenue within a few months.

a16z believes there are two reasons behind this:

First, many companies have set up AI-specific budgets and actively sought solutions, with a quick purchasing process and sufficient budget; second, AI software typically sells the work results themselves, rather than selling auxiliary tools to help people complete their work.

In other words, AI software directly replaces human labor to complete tasks, shifting towards consuming 'human budget' rather than 'traditional IT budget,' thereby enhancing the overall contract value.

( Is e-commerce dead? How does AI agency take over global consumer decision-making power and rewrite business rules? )

The popularization of technology has led to an explosion in AI applications: even edge scenarios are being seen.

At the same time, AI model pricing is rapidly declining, with OpenAI's o3 model price dropping by as much as 80% in one year. The popularity of tools such as Replit and Cursor allows non-technical users to build AI applications, significantly accelerating the speed of AI product explosion.

This means that whether it's a personalized travel planning assistant, a health monitoring dashboard, or an internal tool to streamline business processes, a large number of them will emerge:

AI is giving birth to many small niche markets and edge application scenarios that used to have small economic benefits and couldn't be scaled, enabling more innovative products to land.

(Interview with Musk: AI superintelligence will explode, entrepreneurs should pursue a life of "useful" rather than "great")

The quickest hand is the winner: being the first mover is a key advantage

AI buyers are not only experiencing information explosion, but also have already developed brand preferences. a16z emphasizes, "Are you the first AI solution to enter this field?" has become the core of whether many enterprises will make a purchase.

In the AI battlefield, speed is a monopoly. Pioneering startups such as code editor (IDE), Cursor, and Harvey, favored by lawyers, have become representative products in specific categories early on, relying on rapid iteration and user penetration. This speed not only surpasses traditional giants but also marginalizes competitors before they can react.

To go far, you still need to build a moat. AI itself is not a panacea.

AI capabilities do not equate to business barriers. In order to avoid being replaced by faster runners, a16z reminds successful startups that they need to build a real "moat".

Become the hub of data: such as Eve or Salient, starting from voice or unstructured data, gradually transforming into a key data system for the industry.

Establish user habits and workflow binding: such as Decagon strengthens the human-machine collaboration process, making it difficult for users to break away.

Deep integration into the corporate ecosystem: such as Tennr, HappyRobot, connecting the old systems of medical and logistics internally, providing irreplaceable value.

Establish trust and collaboration roles: Excellent AI vendors are no longer just tool suppliers, but also partners to help businesses develop AI strategies.

From false heat to real strength, AI startups aim for more precision and depth

In recent years, the AI boom has created unprecedented opportunities for global enterprises, but it has also made the competitive landscape extremely fierce. AI startups that survive not only need to iterate quickly but also need to precisely target demand and create solid moats.

From demo to deals, from ideas to products, from mere tools to partners, the road of AI startups is much longer than imagined, adding value to teams who are fully dedicated to it.

How AI products navigate through bubbles and towards longevity? Introduction a16z's five suggestions for enterprise AI startups first appeared in ChainNews ABMedia.