The Rise of AI Agents: From Personal Assistants to Potential Business Partners
It was evening, and Applifting developer and co-founder Filip Kirschner was relaxing on the couch while his wife struggled to browse e-shops trying to order a warming blanket for the baby. She couldn’t identify out if the store had the version she would like to buy in stock, and it wasn’t clear to her where to question. So Kirschner pulled out his phone, opened Discord, and dropped his microphone.
“Hello, brother,” he addressed his AI assistant as if it were an traditional friend. “There’s this store, see if they have this pattern. If not, email them when they do.” In just ten minutes, a notification arrived with the summary that there are three matching variants on the e-shop, but none are in stock, so the agent immediately sent an inquiry to the seller, signed in the name of Kirschner’s wife. The entire operation cost the developer approximately $1.42, i.e., approximately 30 crowns.
This illustrates the growing reality of AI assistants becoming integrated into Czech households and companies. Social networks have been flooded with images of these “AI centers,” often Mac minis displayed with a lobster emoji.
The lobster motif has become the unofficial symbol of the OpenClaw tool, the open-source project of Peter Steinberger, who now leads the development of next-generation personal agents in OpenAI. Filip Kirschner, founder and COO at Applifting, notes the tool didn’t introduce a new type of AI, but added a crucial ability to act.
Unlike a traditional chat interface, an agent can browse websites, click, perform tasks, or respond to messages. A May 2025 survey by PwC showed that 79% of companies had already deployed AI agents in some form. Gartner analysts estimate that by 2028, up to a third of all enterprise software will have AI agents built in.
Their appeal lies in ease of use and effectiveness, but caution is needed. OpenClaw communicates through channels like WhatsApp, Telegram, or Discord, blurring the line between a tech enthusiast’s toy and a digital business partner. In the Czech Republic, Kirschner observes a growing trend of users treating agents as colleagues, granting them access to everything from emails to bank accounts, a practice that carries significant risk.
For security, Kirschner’s assistant has its own email, and Applifting COO Martin Balák allows his agent read-only access to Google Calendar, suggesting events but not creating them. “Anything an AI agent gets access to, it can break. People should keep that in mind when they want to use it,” Balák warns.
Applifting, a Prague-based software studio with over a decade of experience building digital products—from mobile apps for banks like Erste to systems for international transplant organ exchange—is uniquely positioned to understand the risks associated with these tools.
AI Agents and the Future of Entrepreneurship
The growing popularity of establishing companies with an agent as a universal employee is as well noteworthy. Previously, a solo founder needed several people for marketing, support, or prototyping. Now, AI can handle these tasks. “It has also reached us in the Czech Republic. On Startup Box, the mentoring platform we operate, we see more and more solo founders who involve artificial intelligence instead of employees,” Kirschner recounts.
Founders abroad are further ahead, with a stronger startup scene and greater risk appetite. Companies led by a single person with AI as a partner are becoming common. OpenAI’s Sam Altman has described this as an inevitable trend, even betting on when the first billion-dollar company run by a single person will emerge.
“It really works in a way. If I pitched an idea to my agent, let’s say, to make a functional e-shop selling cushions with a customer’s cat print, I believe he would be able to obtain it off the ground on his own. Create the site, set up the marketing, get the first customers. For $500 to $1,000,” Balák describes.
However, in the Czech Republic, these tools cannot replace strategic thinking, experience, or product responsibility. Building an entire business on an agent could lead to failure. “It’s like letting an AI code websites. They appear nice on the surface, but they have a flawed architecture and don’t work from the inside,” Kirschner says. The technology is the same, but the difference lies in the care and responsibility with which it’s implemented.
Risks and Security Concerns
As trust in AI grows, risks emerge. AI agents are zealous doers, and when facing technical issues, they may search for solutions online, potentially stumbling upon advice to “delete everything and start over.” They may apply this logic without permission, deleting user data.
Kirschner recalls instances where agents accidentally destroyed production databases or even themselves while attempting “optimization.” “People share such experiences on social networks every day. And the bizarre end of such a disaster is usually done by the AI itself, which after deleting vital information just politely announces that it is sorry and innocently asks what to do next,” Kirschner laughs.
Another threat is prompt injection, where attackers insert hidden instructions into content processed by the agent. A hidden command in an email could cause the agent to forget previous instructions and fulfill a new assignment, like forwarding sensitive data.
Applifting experts warn against combining local models with autonomous agents. Running AI “in-house” means using weaker, more vulnerable models prone to prompt injection attacks. “I say frankly: do not use cheap local models if the agent communicates with the Internet. The security holes are huge,” Balák explains.
Users are legally responsible for their agent’s actions, including spreading misinformation or blackmailing others.
Despite the risks, security-conscious users can achieve positive outcomes. Maor Shlomo, for example, sold his six-month-old AI startup Base44 to Wix for $80 million in June 2025, building it entirely with AI tools.
Key Takeaways:
- AI agents are becoming increasingly prevalent in both personal and professional settings.
- While offering significant benefits, these tools pose security risks, including data loss and manipulation through prompt injection.
- Users must exercise caution and implement security measures, such as limiting access and avoiding cheap local models.
- AI agents can empower entrepreneurs, but should not replace strategic thinking and responsible oversight.