
TL; DR: The AI ecosystem is expanding and not only from the model front. AI agents have grown in popularity in the last year, stealing some of the spotlight that once solely shone on AI assistants. As it happens, these AI tools possess more capabilities than your typical chatbot. Agentic solutions are smarter, more proactive, and autonomous. We spoke with Dr. Eli Brosh, Head of AI Research at Wix, about the advances made in AI technology and how AI agents can fit into your business system.
AI assistants helped blaze the path for mass AI adoption amid ChatGPT’s rise. But according to the experts, the newer, evolved AI agents may be the way of the future.
Though AI agents and assistants share some similarities, they differ in one primary way: autonomy. While AI assistants are valuable helpers for simpler interactions, AI agents offer stronger support for more complex workflows.
Unlike assistants, an AI agent can work autonomously to achieve specific goals and perform beyond Q&A scenarios. We spoke with Dr. Eli Brosh, Head of AI Research at Wix, about the rise of AI agents and how they will shape business processes going forward.
“As the AI evolves and specifically the agent evolves, they will actually evolve into proactive long-term companions within the enterprise. So you could set it up for much smarter tasks,” said Eli.
AI agents do what AI assistants can’t: take the initiative to orchestrate complex flows. And this capability enables agentic solutions to be actual extensions of business teams, adding more value to workflows than assistants ever could.
Below, we’ll take a deeper look at the inner workings of an AI agent and its blossoming relationship with humans.
Here Come Newer and Improved AI Agents
Imagine you have Employee A and Employee B. Employee A excels at taking direction, is adept at active listening, and has excellent conversational skills.
Employee B, on the other hand, is more of a go-getter. B takes initiative, can strategize on the fly to help the team, and has the tools and skills to develop solutions without awaiting instructions.

In the AI ecosystem, Employee A would identify as an AI assistant while Employee B would be an AI agent. This example shows the main difference between the two solutions: One is able to think and work on their own without oversight.
Because AI agents are built with persistent memory, they have a greater capacity to learn. Once a user gives it a kickoff prompt, an agent can get to work and learn from previous actions and experiences to improve future responses and decision-making.
An AI assistant doesn’t have the same propensity and limits its actions to the direct prompting of a human.
AI agents’ autonomous nature gives them the potential to impact and reshape business processes on a fundamental level. Not only can teams integrate agents into any part of their business ecosystem, but they can also entrust important tasks to them to improve efficiency.
“They can understand or anticipate your needs. Agents can orchestrate complex flows and actually think about what’s the right or best strategy to help you achieve your goals,” said Eli.
Take web hosting customer service for example. While a chatbot assistant can help with Q&A customer interactions, an intelligent agent can help your hosting company troubleshoot some of the issues a customer may have.
With this capacity, AI agents can be true members of a team. They don’t need to wait on simple Q&A prompts. Instead, agents can use past interactions to be proactive and make suggestions to you and your team, making them ideal for high-level reasoning and support.
AI Models Are Becoming More Specialized
As head AI researcher at Wix, Eli has spent a ton of time reworking and fine-tuning AI models to create the best experience for users.
“We would fine-tune these models to a specific journey or experience to let the model be aware of the exact domain or context at which it operates so you can get much more effective results,” said Eli.
By improving its models, Wix was able to train its AI for specialized experiences. Wix’s AI advancements mirror the shake-up currently taking place in the AI landscape.
Enterprises have relied heavily on general-purpose AI models, such as those of OpenAI. But latency and expense issues make leveraging these models a costly and inefficient endeavor for businesses. Eli sees the AI market pivoting elsewhere.
“What I think will happen is that we’ll see a shift towards introducing these very lightweight, efficient specialized models tailored for specialized needs that you have,” said Eli.
Developers are adding more specialties to AI models, increasing the number of use cases they can execute. For enterprises, this can be pivotal for their business structure. Not only will specialized models help their bottom line, but they will also promote operational efficiency and security.

“We’ll see a lot of these smaller models that are controlled and owned by the enterprise that will complement the larger, general-purpose models with broader capabilities. Together they’ll give you a cost effective solution and create a scalable ecosystem,” said Eli.
As AI models become more specialized, so will AI agents. With specialized data training, agents will have the capabilities to tackle industry-specific tasks and processes, which can make them ideal for different use cases.
In fact, team play is one of the primary highlights of AI agents. Because agents are great at completing tasks within their scope, users can split and delegate responsibilities among their agents. And specialized models will only help agents achieve more accurate results within their specialties.
For example, you can have one agent be in charge of fact-checking while another can cover research. Both will do their jobs well because they were trained specifically for those tasks. In the grand scope, you’ll have a well-oiled machine of agents working together to tackle objectives and provide suggestions to your employees.
A Budding Partnership Between Humans and AI
Will AI eliminate the need for a human workforce? According to Eli, no.
“So I think as AI becomes more powerful, we’ll have a lot of daily workflows that can leverage these agentic or AI based solutions. But the important part is that they will include human oversight,” said Eli.
Humans won’t have to worry about surrendering their workplace to AI robots anytime soon. Better yet, AI and humans are more on course to form a long-term partnership than anything else. Eli describes this budding relationship plainly in the quote below.
“For example, in customer care, the AI may handle the basic, mundane inquiries, while humans can handle the more complex issues,” said Eli.
In web development, this partnership is already an established reality. Developers design and build the overall architecture and system. And the AI takes care of the details and menial tasks.
If regulated and executed properly, the partnership between AI and humans can give way to levels of efficiency and productivity never seen before, with humans at the helm and AI doing what it does best: assist.
“I think we’ll see more of this collaboration between AI and humans. And basically it will give both trust and partnership and combine human creativity, direction, and oversight with the speed and scalability of AI,” said Eli.