Agents are poised to be the next big thing in AI. Generally speaking, they’re autonomous tools with advanced reasoning and decision-making capabilities. You give them a destination—no map—and they figure out how to get there on their own.
Forward-looking companies are racing to develop and adopt agents. In a recent IBM and Morning Consult survey of 1,000 enterprise AI developers, 99% said they were actively exploring or building AI agents. My SaaS company is tossing our hat in the agent ring, too.
I’m also testing and working with agents daily and encouraging our teams to do the same. Here are my insights and favorite agents in 2025.
Voice Agents
Voice agents are an accessible and powerful entry point to AI. As a16z notes in its 2025 update on AI voice agents: “For consumers, we believe voice will be the first — and perhaps the primary — way people interact with AI. This interaction could take the form of an always-available companion or coach, or by democratizing services, such as language learning, that were previously inaccessible.”
A voice agent can handle customer phone calls 24/7—even during off-hours when calls would typically go to voicemail. It can also proactively follow up on leads that might otherwise fall through the cracks.
One of my favorite voice agents is VoiceOS. With its natural-sounding voice, it brings a distinctly human touch to customer interactions. Conversations with VoiceOS feel like real dialogue: it responds to interruptions, listens closely to your questions, and allows the conversation to shift fluidly between topics. You lead—and VoiceOS follows. It’s also capable of answering questions across a wide range of subjects, making it a versatile assistant.
Knowledge-Based And Persona-Based AI Agents
Imagine having a research assistant along with a library containing all of the references your assistant might need to consult—this is one of the powerful functions of knowledge-based AI agents. One of my favorite tools is Text.cortex. It lets you build your own knowledge base by uploading documents and organizing them into categories. You can give your AI agent tasks and it will check the references, synthesize information, and produce answers that include the documents referenced.
With Text.cortex, you can also create persona-based agents—different AI personalities with distinct roles, for example, a marketing strategist, a legal assistant, a product coach. Each agent can be trained with a specific tone, objective, and area of expertise. It feels like you’re chatting with cross-functional team members, all drawing from the same shared knowledge base.
Another benefit of AI agents for research is continuity. Unlike tools like ChatGPT, which forget conversations between sessions, your Text.cortex agents remember previously uploaded content and interactions. This makes it ideal for people who rely on repeatable and ongoing workflows, like content creators or solopreneurs—in other words, professionals who build things over time.
Deep Research Agents
Remember high school math, when you couldn’t just write the answer, but you also had to show how you got there? That’s what deep research with Gemini feels like. Unlike generative AI, where you ask a question and almost instantly get an answer with maybe a few links to sources, Gemini can function more like a research agent, showing its work as it goes. You see the agent in action: what it’s doing, what sites it’s referencing (often hundreds), and the steps it’s taking in real time. This type of agent is geared toward more advanced, in-depth research tasks, especially those involving academic papers, scientific data, and similar sources.
For example, I once asked Gemini to tackle this task: “Research AI agents in customer service. Create a list of 100, and group them into 3–5 categories.” Gemini got right to work, reviewing a wide range of web sources and highlighting important definitions, characteristics (like being goal-oriented or focused on customer resolution), and ultimately, a thorough report: The Rise of Intelligent Automation: An Examination of AI Agents in Customer Service. The final output included a categorized list, with each agent’s name, provider, and function.
Automation Agents
Automation AI agents are ideal for offloading repetitive, time-consuming tasks that tend to clog up your day. By combining machine learning, predictive analytics, and rule-based logic, these agents handle jobs that don’t need human input—helping teams boost productivity, cut down on errors, and ultimately, lower operating costs.
One standout in this category is Reworkd’s AgentGPT, a top-searched automation agent that runs directly in your web browser. Users can customize their AI agent, giving it a name and a goal, and the agent takes it from there—breaking the goal into tasks, executing them step-by-step, and evaluating progress along the way. Whether it’s scraping data, compiling reports, or ticking off other digital tasks on your to-do list, AgentGPT works autonomously, so you don’t have to micromanage the process.
Final Thoughts
As Reuters recently reported, autonomous agents are likely to “dominate the artificial intelligence agenda”—a sign of how central they’re becoming in the AI landscape. What distinguishes agents, to me, is how useful they are. As Anthropic co-founder and chief scientist Jared Kaplan recently told MIT Technology Review, “If you go back almost 10 years now to [DeepMind’s Go-playing model] AlphaGo, we had AI systems that were superhuman in terms of how well they could play board games. But if all you can work with is a board game, then that’s a very restrictive environment. It’s not actually useful, even if it’s very smart.”
Today’s agents are different. They’re intelligent, dynamic, and—most importantly—practically useful. They help humans do their jobs better in so many ways and free up time to focus on more meaningful work.
Read the full article here
