‍‍‍‍ The development of artificial intelligence agents will fundamentally change the way people work and the face of startups. Over the past year, the number of AI-based agent-based startups has soared from single digits to dozens per month.

In Israel, the number of startups that build AI agents has surged, with the focus being to enable others to integrate and customize these agents to meet the needs of different scenarios.

Many of these companies are leveraging Israel’s advantages in cybersecurity, data science and enterprise software to create AI agents that can solve vertical industry problems, such as medical diagnosis and predictive security.

At the same time, horizontal applications such as workflow automation and personalized customer participation have also emerged.

As people start looking at more of these AI agent-led startups, they are noticed that they are following certain common patterns.

For example, startups initially assisted by general AI are transforming into a complete "artificial intelligence organization."

With every major advancement in the field of AI agents, we are getting closer to the trend we began to predict a few years ago: more and more rely on AI automation operation Companies are beginning to emerge, and humans are only responsible for making key strategic decisions.

This momentum has been going on for several years, but now it feels like a turning point.

OpenAI CEO Ultraman predicts that this year will be the year for artificial intelligence agents to truly join the workforce. By 2027, at least half of companies will launch some form of artificial intelligence agents.

And this is just a starting point.

In the near future, one may see that the entire economy is composed of these AI-first organizations. And if you want to build a truly lasting company, you must see this direction of development.

Maybe after that, companies will hire AI agents, and humans will work with AI agents, or even compete with them.

What else will happen next?

The following are five possible evolutionary stages for artificial intelligence agents:

01.General Chat

The first wave of artificial intelligence collaborators is the basic model (general LLM, such as ChatGPT or Claude). They break through users' user experience and help people understand the broad capabilities of artificial intelligence.

However, artificial intelligence is just a tool, and humans are still leading the way in injecting situations, rationality and empathy into artificial intelligence.

As early adopters said, these general tools are all "owners without a master." This has shifted the landscape of AI startups toward its first evolutionary stage.

02.Domain Expert

General-purpose artificial intelligence can read and write, and in the correct Perform tasks under guidance. However, universal AI tools still perform poorly in super specific industry environments.

Short after the rise of general artificial intelligence, people began to see the emergence of real "experts" in the field of AI.

Artificial intelligence seems to solve problems without excessive human prompts, and chat is still the main interface of these systems, but many companies are already in the chat function. Additional industry-specific features are built on the basis.

Law is one of the examples, and companies such as EvenUp and Darrow have demonstrated the power of artificial intelligence trained in specific legal data corpuses.

These AIs can understand the nuances of legal language and generate professional-level legal materials.

03. Artificial Intelligence Agent (At present)

There are still many excellent companies Carry out business at the expert level in the field of artificial intelligence.

But in the past year or soThere is a clear shift in the intervening process, namely, from a chat-based value proposition to an action-based value proposition.

Universal AI tools and domain experts are true "co-drivers" that can create new connections, generate articles, or provide new material. But humans still need to take action in order for these tools to really work.

From April 2023, people have begun to see that artificial intelligence can perform some more advanced tasks.

The most famous case of artificial intelligence agents is in the field of code generation, such as OpenAI's code interpreter or Cognition's AI programmer Devin.

But this concept has gone far beyond the scope of code generation, but has entered a more complete "job description".

Now there are more and more AI agents that specialize in performing specific tasks. The packaging and combination of these tasks has great potential to translate into real services.

For example, Enso, powered by NFX, is creating a market for AI agents for small and medium-sized enterprises.

Once people continue to improve their ability to complete tasks without a lot of manual supervision Take action under circumstances, and people can no longer look back.

04.Artificial Intelligence Agent Innovation

Once the AI ​​agent can continue to perform tasks, people You will soon see agents with innovative capabilities. If people allow artificial intelligence to generate and explore new directions of knowledge, its value will be raised to a whole new level.

People can look at this problem in the way the human brain solves problems and exerts creativity.

People have task-oriented "if-then" brain presets that help people perform tasks and solve problems.

But we also have an active subconscious mind, a way of thinking that is turned on when you are not focused on solving problems, such as when you are taking a shower or taking a walk.

Have you ever had such an experience, when you took great pains to write or solve the problem, and then easily solved the problem after walking around?

This is the result of your subconscious unrestrained exploration of new creative methods. Most new, creative ideas emerge in this state.

Artificial intelligence innovation agents will be able to conduct this subconscious exploration. They are not bound by the logical statement of "if-then", which can lead to narrow thinking.

Imagine that on Monday you asked a group of AI agents to develop a software function, and by Wednesday you will find that the agent has already based on trial and error experience and market analysis. Improve your initial needs and develop better features.

When the target itself is more abstract (increase sales, improve software performance, and allow users to Like my application), planning goals and setting paths will be the key to the next stage of development of AI agents.

This is also an important factor in making AI agents truly a mature labor force.

Pure automation without critical thinking is a life-saving straw for the lowest hanging fruit in the economy. But it cannot solve the biggest and most valuable problem, creativity is.

The key unlocking lies in trust. People need to be confident in AI agents to make strategic decisions, not just task-oriented decisions.

Some trusts must be established through technology. People need two things: interpretability and infrastructure. These two things themselves may even become industries.

For example, NFX-backed Maisa is perfecting the "proof of work" of artificial intelligence agents, a key factor in building trust in the entire agent ecosystem.

Emcie, another company invested by NFX, is developing the infrastructure needed to create hyper-specific AI agents for individuals and businesses.

This trust will be in the textDevelopment in evolution. The more people see AI making smart decisions and creating better results, the faster the future will come.

The early adopter group will be the key. Small and medium-sized businesses or companies that simply cannot hire people to meet their needs will take the first step, and other parts of the ecosystem will wait and see and follow.

It will touch all industries. For example, the field of education:

05. AI-first organization

With proxy AI workers, AI innovation, and trust and interpretability systems in place, people will eventually see the rise of real AI organizations.

These organizations are a collection of artificial intelligence agents and artificial intelligence innovators that are able to carry out a wide range of actions.

This is the artificial intelligence that people often hear in science fiction novels.

Worst case, you can read this kind of AI in Daemon by Daniel Suarez, or Naomi Kritzer "Improving Life through Algorithms" is found.

These agents can make decisions in complex environments where many potential goals are worth achieving.

The difference here is that artificial intelligence itself will be able to self-select which goals are the best and design a path to achieve them.

AI will take most of the action, and you will fight alongside AI, review and review the route it takes.

It is conceivable that a self-managed supply chain can oversee the entire process from production to delivery and produce automated financial trading companies composed of many artificial intelligence agents.

People don't expect this to happen immediately, it will happen in steps.

With the development of trust and technology, artificial intelligence will begin to undertake increasingly larger tasks. In fact, people are still in the technical window of artificial intelligence proxy systems.

The people who really understand this are still those builders and amateurs who work hard. But soon, an artificial intelligence-led organization will emerge and people will usher in the "ChatGPT" moment. Before ChatGPT came out, how many people really understood the capabilities of artificial intelligence?

If you know where people are going, you'll be one step ahead.

In Israel, the AI ​​agent market is booming, and startups are leveraging local expertise in machine learning, cybersecurity and automation.

People see more and more companies building basic agency platforms for other companies to customize, such as Enso.

The startups here have begun to deal with vertical challenges in areas such as fintech, logistics and healthcare, and position themselves as a rapidly growing AI ecosystem. Important contributor.

AI agents are here, AI innovators are here, and AI organizations are here.

So, now ask yourself, what is preventing these things from entering my field? How to eliminate obstacles? Or, once they are eliminated, how will I be the main beneficiary?

Not every company should focus on building an artificial intelligence proxy infrastructure. But you can understand how the overall economic benefits of your field will change as people release these new labor pools.

You can also think about what psychological impact this will have on the team - what will it look like when managing only AI employees? Or, in turn, what about when humans are managed by AI?

In NFX, people's job is to study how transformative technologies work. These transformations are timely, and as technology changes, certain skills will become more important or weakened.

We also need to deal with psychological changes and the new opportunities that will bring.

Artificial intelligence agents are in the most exciting new life Stage, this is the moment when a great company is born.