How to succeed with AI agents — advice from a strategist at HP

      The swift emergence of AI agents is generating both enthusiasm and concern. Their strength lies in their increasing autonomy to complete tasks. Many AI agents can already tackle multi-step objectives, make decisions, and communicate with external systems with minimal human involvement. Teams of AI agents are starting to collaborate, each taking on a specific role. As their independence grows, they are set to transform numerous business procedures.

      Tech leaders are proclaiming them as the future of the internet. During Microsoft’s Build conference this week, the company announced that we have entered "the era of AI agents." OpenAI CEO Sam Altman participated in the event, describing his lab's new Codex tool as "a genuine agentic coding experience," calling it "one of the most significant transformations in programming that I've ever witnessed."

      Despite the excitement, practical implementations are quickly developing. AI agents are already aiding in various functions, ranging from code production and cybersecurity threat identification to customer service inquiries and marketing strategies.

      Before long, they might evolve into comprehensive executive assistants—managing emails, calendars, and projects. However, to take advantage of these opportunities, individuals must start preparing now.

      Cihangir Kocak is assisting with this preparation. As a principal business and AI strategist at HP, Kocak helps organizations navigate digital transformation. He believes AI agents will open up a new realm of possibilities. "We are moving toward a future where everyone will have an AI agent as an assistant," he states.

      At the TNW Conference this summer, Kocak will lead two sessions on AI agents. On June 19, he will present a keynote addressing their rise. The following day, he will team up with Joost Bos, Senior AI Engineer at Deloitte, for a masterclass titled “Agentic AI: Architecting the Future of Business.”

      In anticipation of the event, he shared some insights.

      1. **Understand what AI agents can do**

      AI agents change large language models (LLMs) from being passive responders to active problem-solvers. Equipped with tools, memory, and clear objectives, they can independently accomplish complex tasks. "Large language models serve as the brains while AI agents function as the hands, meaning they can take action," Kocak explains. "They can autonomously perform tasks for you." Agents can work together, with one sourcing products, another managing logistics, a third constructing your website, and a fourth drafting marketing content. In the future, companies may need their own agents to communicate with others. Your AI assistant could work alongside them to secure the best service for your requirements. Free courses from organizations like Hugging Face, Salesforce, and Microsoft are great starting points to explore the potential.

      2. **Start experimenting**

      Kocak believes AI agents will quickly redefine workplaces. "I anticipate that within five years, everything will be transformed due to AI agents," he asserts. "It could even be less than five years—maybe just two to three years." Numerous companies are already transferring various tasks from humans to AI. In the near future, new hires may need experience with AI agents. "Soon, many of these firms will seek individuals who can work with AI agents," Kocak warns. His advice? "Get hands-on. Experiment with it — but do so mindfully." One tool he suggests is LM Studio, a desktop application for running LLMs locally. However, his primary recommendation is simply to begin. "Do something to familiarize yourself with it. Once you do that, you can proceed to the next phase."

      3. **Find use cases**

      After testing some tools, Kocak recommends pinpointing where they can provide value. He advises seeking tasks where AI can save you time and starting small. "What takes up most of your time? What tasks do you dislike? Identifying these can help you figure out how AI agents can assist you." Kocak utilizes local LLMs for tasks that require privacy and ChatGPT for public applications, such as crafting LinkedIn posts in his voice. "It saves at least half of my time," he notes.

      4. **Focus on the data**

      The true power of AI agents reveals itself when they are personalized with your selected data. Generic tools like ChatGPT can handle broad tasks, but for something specific, agents trained on your data can deliver better results. That internal knowledge can transform a standard agent into a tailored powerhouse. "What differentiates an AI solution is when you provide it with your own data," Kocak asserts. "This creates a solution that can function uniquely."

      5. **Maintain human oversight**

      Even though AI agents can operate independently, human oversight is essential. While agents are potent, they are not infallible. Granting them too much freedom poses risks. "It's advisable to have a human involved," he states. "The future will be a blend of AI agents and humans—that will be the most advantageous combination."

      6. **Stay secure**

      As AI