it is StepmadeA weekly newsletter broke an essential story from Tech World. For more on everything, Follow Hayden FieldStepback comes to ET at 8 am in our customers’ inboxes. Select Stepmade Here,
All this started with Jarvis Yes, that Jarvis the Marvel.
Well, maybe it was not Start With the AI assistant of the Iron Man, but the fictional system certainly helped with the concept of the AI agent. Whenever I have interviewed the people of the AI industry about Agent AI, they often indicate Jarvis in many ways as an example of ideal AI tools – one who knows that whatever you have done before asking can analyze, can analyze and find insight and discovered in large swaths of data, and can offer strategic advice or running on some aspects of your business. People sometimes disagree over the accurate definition of AI agent, but at its core, it is a step beyond chatbots that it is a system that can do multistap, complex tasks on your behalf, which is constantly without the need for communication with you. This essentially needs to complete your own two-do list to meet your favorite final goal. This imagination is close to being a reality in many ways, but when it comes to real utility for everyday users, there are many things that do not work – and perhaps will never work.
The word “AI agent” has been around for a long time, but it especially Started trending in the tech industry in 2023. It was the year concept Of AI agents; The word was on everyone’s lips because people tried to exclude this idea and made it a reality, but you did not see many successful use matters. The next year, 2024, was the year Deployment – People were actually putting the code in the field and seeing what it could do. (Answer, at that time, was not much. And the error was filled with a group of messages.)
I can extend the campaign around AI agents widely for a specific announcement: in February 2024, a Fintech Company, Clarna, a Fintech Company, Said A month later, its AI Assistant (run by OpenI Tech) successfully served as 700 full-time customer service agents and automated two-thirds of the company’s customer service chat. For months, those figures came to me to the conversation of almost every AI industry.
The publicity never died, and in the next months, every big tech CEO used to get a veena on the word in every earning call. Amazon, Meta, Google, Microsoft, and other companies began to talk about their commitment to the creation of executive and successful AI agents of a complete hosting of other companies – and tried to invest their money, where they have to face.
The vision was that one day, an AI agent could do everything from your journey to your journey to generate scenes for your business presentations. Ideal tools can also say, find a good time and place to roam with a group of friends that works with all your calendar, food preferences and dietary restrictions – and then book dinner reservation and create a calendar event for all.
Now let us talk about the “AI coding” of all: Over the years, AI has been coding Lift Agent AI Industry. If you ask someone about real life, successful for AI agents, not-Hangam Now! And ideologically not in a very distant future, they would point to AI coding-and it was a very solid thing that they could indicate. Many engineers use AI agents for coding, and they are purified as very good. It is quite good, in fact, that in Microsoft and Google, Up to 30 percent The code is now being written by AI agents. And for startups such as Openai and Anthropic, which burns through cash at high rates, one of their largest revenue generators is the AI coding tool for customers.
So recently, AI coding has been the main real -life case of AI agents, but obviously, it is not for everyday consumer. Vision, remember, always had a jack-off-all-trade of the AI agent for “every person”. And we are not there yet – but in 2025, we have come closer than before.
In the previous October, Anthropic shut down things by starting “using a computer”, a device that allows the cloud to use a computer like humans – browsing, searching, reaching various platforms, and completing complex functions from the user. The consensus was that equipment was a step ahead for technology, but reviews said that in practice, it left a lot to be desired. By January 2025, Fast-forward, and OpenIAI released to the operator, a version of the same thing, and filled it as a tool for filling the form, ordering groceries, traveling and making memes. Once again, in practice, many users agreed that the device was small, slow, and not always efficient. But then, this was an important step. Next month, Openai issued deep research, an agent AI tool that can compile long research reports on any subject for a user, and that even further things further. Some people said that research reports were more impressive in length than materials, but others were seriously affected. And then in July, Openai added an AI agent product: intensive research and operator in the Chatgpt agent. Was it better than most of the pre-upcoming consumer-consumed agents AI Tool? Absolutely. Was it still difficult to work successfully in practice? Absolutely.
So to reach a long way to reach that vision of an ideal AI agent, but at the same time, we are technically closer than before. This is why tech companies are investing more and more money in the agent AI, through additional calculations, research and development, or talent. Google has recently hired Windsurf’s CEO, Kofounder and some R&D team members, especially to help Google to pursue their AI agent projects. And companies like anthropic and openi are running each other by ladder, roung to introduce these agents in the hands of consumers. (For example, Anthropic announces a chrome extension to the cloud that allows it to work in your browser.)
So in fact, what happens next we will see that AI coding continues to improve (and, unfortunately, potentially change many jobs Entry-level software engineerWe will also see improvement in consumer-support agents products, gradually but of course. And we will see that rapidly used agents for enterprise and government applications have introduced government-specific AI platforms in recent months since anthropic, openiAI and XAI.
Overall, the more false start, start and stop and merge and see the merger and acquisition as the AI agent competition arises (and the promotional bubble balloon continues). One question we all have to ask ourselves because months go ahead: what do we really want an ideological “AI agent” that are able to do for us? Do we want them to just change logistics or more personal, human aspects of life (ie, help write a note for delivery of a wedding toast or flowers)? And how good are they helping with logistics vs. personal goods? (Answer to that last one: It is not very good at this time.)
- In addition to AI’s astronomical environmental cost – especially for large models, which are strengthening AI agent efforts – there is an elephant in the room. And the idea is that “smart AI that can do anything for you” is not always good, especially when people want to use it to do it … bad things. Things such as chemical, biological, radiological and atoms (CBRN) weapons. Top AI companies say they are rapidly worried about that risk. (Of course, they are not worried enough to stop the building.)
- Let’s talk about the regulation of all this. Many people are apprehensive about the implications of AI, but many are not fully aware of the potential threats generated by Uber-Helpful, to target the AI agents in both states and abroad (think: “Vib-Hacing,” romance scams, and more). AI companies say they are ahead of risk with voluntary safety measures they have implemented. But many other people say that this may be a case for the outer intestine.
1 Comment
