Why 2025 Is the Year of AI Agents: Insights from the Travel AI ‘Ask‑Me‑Anything’

In the past three years, large models have moved from generation to understanding, from multimodal to execution, giving AI the ability to handle complex task chains. In 2024 the concept of agents rose rapidly, becoming the new focus for AI applications, leading many in the industry to call 2025 the “Agent Year”.

Today AI is no longer just a chat or writing assistant; it is a digital employee that can understand goals, plan paths, execute tasks, receive feedback, and even collaborate. Data‑intensive processes that previously required repeated human comparison, operation, and integration can now be reliably handled by AI.

In this AI race led by tech giants, Microsoft, Google, and OpenAI dominate the underlying models, while Chinese giants Baidu, Alibaba, and Tencent invest heavily. Although it seems small companies have little chance, the maturation of foundational models allows many startups to consider building products themselves.

For example, in April this year Fliggy launched the AI product “Ask‑Me‑Anything”, essentially an AI travel concierge. Unlike a simple Q&A bot, it acts like a professional travel team: planning routes, checking flights, finding hotels, managing budgets, and delivering a complete travel plan based on user input.

When a user says “I want a seven‑day trip to Kunming this summer”, the agent recognizes the current location, immediately generates a full itinerary with flights, hotels, attractions, and a total budget of ¥3,831, arranging daily schedules and even flight times.

Although the initial version had imperfections—budget estimates were rough and hotel arrangements not fully intelligent—the product demonstrated a significant improvement in reliability and usability, standing out in a market where few agents have progressed beyond the concept stage.

Fliggy’s “Ask‑Me‑Anything” relies on real supply data, integrating real‑time flight and hotel inventory with years of travel experience. As the product evolves, features such as special‑price flight and hotel queries, member assistant roles, hand‑drawn maps, and SUG functions have been added, noticeably enhancing the user experience.

01 Who Closes the Loop Becomes the Default Choice

2025 being called the “Agent Year” is not empty talk. From text generation to multimodal, from Q&A assistants to actionable agents, AI has reached a critical point where it can “take action”. It can now book hotels, find flights, plan schedules, run data pipelines, and even automate internal workflows, turning AI from a tool into an employee with real commercial potential.

The window for AI products to evolve from “functional tools” to “task executors” is opening. Whoever first delivers a tangible, user‑perceived AI Agent product will gain a long‑term advantage.

All software products enjoy a “first‑mover advantage”, but AI amplifies this effect: the more the product is used, the more data it gathers, improving the model and user experience, creating a snowball effect that later entrants struggle to match.

Agent products require not only model tuning and API integration but also sophisticated scheduling logic, task chain execution, and feedback loops—capabilities that cannot be achieved by simply copying code but require real‑world iteration.

We are at a subtle inflection point: technology is mature, but market perception is not yet fixed. A product with a great experience and fast pace can become the default choice in users’ minds.

02 The First Executors Are On Stage

Contrary to the belief that agents have high barriers and are only for big firms, the entry threshold is dropping quickly. Open‑source models, platforms like OpenAI, Claude, DeepSeek, and toolchains such as LangChain, AutoGen, OpenDevin enable rapid assembly of multi‑agent systems. Low‑code automation platforms like Zapier and Make.com also handle much of the integration work.

The key now is finding the right scenario, clarifying task chains, building solid scheduling logic, and quickly testing in the market.

Many recent AI products—Kimi, Notion AI, Draw Things, Rewind, Quivr, Mistral—were built by small teams or individual developers, proving that small firms can compete when they address clear user pain points, hide AI complexity behind the scenes, and deliver polished experiences without over‑optimizing model parameters.

Fliggy, though not a startup, benefits from a lean structure and agile decision‑making, allowing it to launch “Ask‑Me‑Anything” faster than larger competitors.

03 This Is a Whole New Future

AI is opening a new era that reshapes industry structures. Small companies that combine deep domain knowledge, accumulated data, and AI can achieve rapid growth, turning previously high‑end travel planning into a “one‑click” experience with real‑time pricing and comprehensive itineraries.

AI amplifies the strengths of niche players, while large firms focus on scalable, generic scenarios. The “edge cases” and specialized needs become the sweet spot for smaller innovators.

When a small team seizes a niche that big players overlook and executes it exceptionally, AI becomes a growth engine, propelling the company ahead of the curve.

In every vertical, an AI‑driven “new giant” may emerge—not necessarily from Silicon Valley or a unicorn, but from a quiet team that leverages a micro‑innovation and AI to achieve explosive growth.

AI has already blurred the boundaries between companies of any size; anyone who can build a genuinely useful AI product can become the next game‑changing force.